xref: /freebsd/sys/ofed/include/rdma/ib_verbs.h (revision a2aef24aa3c8458e4036735dd6928b4ef77294e5)
1 /*
2  * Copyright (c) 2004 Mellanox Technologies Ltd.  All rights reserved.
3  * Copyright (c) 2004 Infinicon Corporation.  All rights reserved.
4  * Copyright (c) 2004 Intel Corporation.  All rights reserved.
5  * Copyright (c) 2004 Topspin Corporation.  All rights reserved.
6  * Copyright (c) 2004 Voltaire Corporation.  All rights reserved.
7  * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
8  * Copyright (c) 2005, 2006, 2007 Cisco Systems.  All rights reserved.
9  *
10  * This software is available to you under a choice of one of two
11  * licenses.  You may choose to be licensed under the terms of the GNU
12  * General Public License (GPL) Version 2, available from the file
13  * COPYING in the main directory of this source tree, or the
14  * OpenIB.org BSD license below:
15  *
16  *     Redistribution and use in source and binary forms, with or
17  *     without modification, are permitted provided that the following
18  *     conditions are met:
19  *
20  *      - Redistributions of source code must retain the above
21  *        copyright notice, this list of conditions and the following
22  *        disclaimer.
23  *
24  *      - Redistributions in binary form must reproduce the above
25  *        copyright notice, this list of conditions and the following
26  *        disclaimer in the documentation and/or other materials
27  *        provided with the distribution.
28  *
29  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
30  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
31  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
32  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
33  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
34  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
35  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
36  * SOFTWARE.
37  */
38 
39 #if !defined(IB_VERBS_H)
40 #define IB_VERBS_H
41 
42 #include <linux/types.h>
43 #include <linux/device.h>
44 #include <linux/mm.h>
45 #include <linux/dma-mapping.h>
46 #include <linux/kref.h>
47 #include <linux/list.h>
48 #include <linux/rwsem.h>
49 #include <linux/scatterlist.h>
50 #include <linux/workqueue.h>
51 #include <linux/if_ether.h>
52 #include <linux/mutex.h>
53 
54 #include <asm/uaccess.h>
55 
56 extern struct workqueue_struct *ib_wq;
57 
58 union ib_gid {
59 	u8	raw[16];
60 	struct {
61 		__be64	subnet_prefix;
62 		__be64	interface_id;
63 	} global;
64 };
65 
66 enum rdma_node_type {
67 	/* IB values map to NodeInfo:NodeType. */
68 	RDMA_NODE_IB_CA 	= 1,
69 	RDMA_NODE_IB_SWITCH,
70 	RDMA_NODE_IB_ROUTER,
71 	RDMA_NODE_RNIC,
72 	RDMA_NODE_MIC
73 };
74 
75 enum rdma_transport_type {
76 	RDMA_TRANSPORT_IB,
77 	RDMA_TRANSPORT_IWARP,
78 	RDMA_TRANSPORT_SCIF
79 };
80 
81 enum rdma_transport_type
82 rdma_node_get_transport(enum rdma_node_type node_type) __attribute_const__;
83 
84 enum rdma_link_layer {
85 	IB_LINK_LAYER_UNSPECIFIED,
86 	IB_LINK_LAYER_INFINIBAND,
87 	IB_LINK_LAYER_ETHERNET,
88 	IB_LINK_LAYER_SCIF
89 };
90 
91 enum ib_device_cap_flags {
92 	IB_DEVICE_RESIZE_MAX_WR		= 1,
93 	IB_DEVICE_BAD_PKEY_CNTR		= (1<<1),
94 	IB_DEVICE_BAD_QKEY_CNTR		= (1<<2),
95 	IB_DEVICE_RAW_MULTI		= (1<<3),
96 	IB_DEVICE_AUTO_PATH_MIG		= (1<<4),
97 	IB_DEVICE_CHANGE_PHY_PORT	= (1<<5),
98 	IB_DEVICE_UD_AV_PORT_ENFORCE	= (1<<6),
99 	IB_DEVICE_CURR_QP_STATE_MOD	= (1<<7),
100 	IB_DEVICE_SHUTDOWN_PORT		= (1<<8),
101 	IB_DEVICE_INIT_TYPE		= (1<<9),
102 	IB_DEVICE_PORT_ACTIVE_EVENT	= (1<<10),
103 	IB_DEVICE_SYS_IMAGE_GUID	= (1<<11),
104 	IB_DEVICE_RC_RNR_NAK_GEN	= (1<<12),
105 	IB_DEVICE_SRQ_RESIZE		= (1<<13),
106 	IB_DEVICE_N_NOTIFY_CQ		= (1<<14),
107 	IB_DEVICE_LOCAL_DMA_LKEY	= (1<<15),
108 	IB_DEVICE_RESERVED		= (1<<16), /* old SEND_W_INV */
109 	IB_DEVICE_MEM_WINDOW		= (1<<17),
110 	/*
111 	 * Devices should set IB_DEVICE_UD_IP_SUM if they support
112 	 * insertion of UDP and TCP checksum on outgoing UD IPoIB
113 	 * messages and can verify the validity of checksum for
114 	 * incoming messages.  Setting this flag implies that the
115 	 * IPoIB driver may set NETIF_F_IP_CSUM for datagram mode.
116 	 */
117 	IB_DEVICE_UD_IP_CSUM		= (1<<18),
118 	IB_DEVICE_UD_TSO		= (1<<19),
119 	IB_DEVICE_XRC			= (1<<20),
120 	IB_DEVICE_MEM_MGT_EXTENSIONS	= (1<<21),
121 	IB_DEVICE_BLOCK_MULTICAST_LOOPBACK = (1<<22),
122 	IB_DEVICE_MR_ALLOCATE		= (1<<23),
123 	IB_DEVICE_SHARED_MR             = (1<<24),
124 	IB_DEVICE_QPG			= (1<<25),
125 	IB_DEVICE_UD_RSS		= (1<<26),
126 	IB_DEVICE_UD_TSS		= (1<<27),
127 	IB_DEVICE_CROSS_CHANNEL		= (1<<28),
128 	IB_DEVICE_MANAGED_FLOW_STEERING = (1<<29),
129 	/*
130 	 * Devices can set either IB_DEVICE_MEM_WINDOW_TYPE_2A or
131 	 * IB_DEVICE_MEM_WINDOW_TYPE_2B if it supports type 2A or type 2B
132 	 * memory windows. It can set neither to indicate it doesn't support
133 	 * type 2 windows at all.
134 	 */
135 	IB_DEVICE_MEM_WINDOW_TYPE_2A	= (1<<30),
136 	IB_DEVICE_MEM_WINDOW_TYPE_2B	= (1<<31),
137 	IB_DEVICE_SIGNATURE_HANDOVER	= (1LL<<32)
138 };
139 
140 enum ib_signature_prot_cap {
141 	IB_PROT_T10DIF_TYPE_1 = 1,
142 	IB_PROT_T10DIF_TYPE_2 = 1 << 1,
143 	IB_PROT_T10DIF_TYPE_3 = 1 << 2,
144 };
145 
146 enum ib_signature_guard_cap {
147 	IB_GUARD_T10DIF_CRC	= 1,
148 	IB_GUARD_T10DIF_CSUM	= 1 << 1,
149 };
150 
151 enum ib_atomic_cap {
152 	IB_ATOMIC_NONE,
153 	IB_ATOMIC_HCA,
154 	IB_ATOMIC_GLOB
155 };
156 
157 enum ib_cq_create_flags {
158 	IB_CQ_CREATE_CROSS_CHANNEL	= 1 << 0,
159 	IB_CQ_TIMESTAMP			= 1 << 1,
160 	IB_CQ_TIMESTAMP_TO_SYS_TIME	= 1 << 2
161 };
162 
163 struct ib_device_attr {
164 	u64			fw_ver;
165 	__be64			sys_image_guid;
166 	u64			max_mr_size;
167 	u64			page_size_cap;
168 	u32			vendor_id;
169 	u32			vendor_part_id;
170 	u32			hw_ver;
171 	int			max_qp;
172 	int			max_qp_wr;
173 	u64			device_cap_flags;
174 	int			max_sge;
175 	int			max_sge_rd;
176 	int			max_cq;
177 	int			max_cqe;
178 	int			max_mr;
179 	int			max_pd;
180 	int			max_qp_rd_atom;
181 	int			max_ee_rd_atom;
182 	int			max_res_rd_atom;
183 	int			max_qp_init_rd_atom;
184 	int			max_ee_init_rd_atom;
185 	enum ib_atomic_cap	atomic_cap;
186 	enum ib_atomic_cap	masked_atomic_cap;
187 	int			max_ee;
188 	int			max_rdd;
189 	int			max_mw;
190 	int			max_raw_ipv6_qp;
191 	int			max_raw_ethy_qp;
192 	int			max_mcast_grp;
193 	int			max_mcast_qp_attach;
194 	int			max_total_mcast_qp_attach;
195 	int			max_ah;
196 	int			max_fmr;
197 	int			max_map_per_fmr;
198 	int			max_srq;
199 	int			max_srq_wr;
200 	int			max_srq_sge;
201 	unsigned int		max_fast_reg_page_list_len;
202 	int			max_rss_tbl_sz;
203 	u16			max_pkeys;
204 	u8			local_ca_ack_delay;
205 	int                     comp_mask;
206 	uint64_t                timestamp_mask;
207 	uint64_t                hca_core_clock;
208 	unsigned int		sig_prot_cap;
209 	unsigned int		sig_guard_cap;
210 };
211 
212 enum ib_device_attr_comp_mask {
213 	IB_DEVICE_ATTR_WITH_TIMESTAMP_MASK = 1ULL << 1,
214 	IB_DEVICE_ATTR_WITH_HCA_CORE_CLOCK = 1ULL << 2
215 };
216 
217 enum ib_mtu {
218 	IB_MTU_256  = 1,
219 	IB_MTU_512  = 2,
220 	IB_MTU_1024 = 3,
221 	IB_MTU_2048 = 4,
222 	IB_MTU_4096 = 5
223 };
224 
225 static inline int ib_mtu_enum_to_int(enum ib_mtu mtu)
226 {
227 	switch (mtu) {
228 	case IB_MTU_256:  return  256;
229 	case IB_MTU_512:  return  512;
230 	case IB_MTU_1024: return 1024;
231 	case IB_MTU_2048: return 2048;
232 	case IB_MTU_4096: return 4096;
233 	default: 	  return -1;
234 	}
235 }
236 
237 enum ib_port_state {
238 	IB_PORT_NOP		= 0,
239 	IB_PORT_DOWN		= 1,
240 	IB_PORT_INIT		= 2,
241 	IB_PORT_ARMED		= 3,
242 	IB_PORT_ACTIVE		= 4,
243 	IB_PORT_ACTIVE_DEFER	= 5,
244 	IB_PORT_DUMMY		= -1	/* force enum signed */
245 };
246 
247 enum ib_port_cap_flags {
248 	IB_PORT_SM				= 1 <<  1,
249 	IB_PORT_NOTICE_SUP			= 1 <<  2,
250 	IB_PORT_TRAP_SUP			= 1 <<  3,
251 	IB_PORT_OPT_IPD_SUP                     = 1 <<  4,
252 	IB_PORT_AUTO_MIGR_SUP			= 1 <<  5,
253 	IB_PORT_SL_MAP_SUP			= 1 <<  6,
254 	IB_PORT_MKEY_NVRAM			= 1 <<  7,
255 	IB_PORT_PKEY_NVRAM			= 1 <<  8,
256 	IB_PORT_LED_INFO_SUP			= 1 <<  9,
257 	IB_PORT_SM_DISABLED			= 1 << 10,
258 	IB_PORT_SYS_IMAGE_GUID_SUP		= 1 << 11,
259 	IB_PORT_PKEY_SW_EXT_PORT_TRAP_SUP	= 1 << 12,
260 	IB_PORT_EXTENDED_SPEEDS_SUP             = 1 << 14,
261 	IB_PORT_CM_SUP				= 1 << 16,
262 	IB_PORT_SNMP_TUNNEL_SUP			= 1 << 17,
263 	IB_PORT_REINIT_SUP			= 1 << 18,
264 	IB_PORT_DEVICE_MGMT_SUP			= 1 << 19,
265 	IB_PORT_VENDOR_CLASS_SUP		= 1 << 20,
266 	IB_PORT_DR_NOTICE_SUP			= 1 << 21,
267 	IB_PORT_CAP_MASK_NOTICE_SUP		= 1 << 22,
268 	IB_PORT_BOOT_MGMT_SUP			= 1 << 23,
269 	IB_PORT_LINK_LATENCY_SUP		= 1 << 24,
270 	IB_PORT_CLIENT_REG_SUP			= 1 << 25
271 };
272 
273 enum ib_port_width {
274 	IB_WIDTH_1X	= 1,
275 	IB_WIDTH_4X	= 2,
276 	IB_WIDTH_8X	= 4,
277 	IB_WIDTH_12X	= 8
278 };
279 
280 static inline int ib_width_enum_to_int(enum ib_port_width width)
281 {
282 	switch (width) {
283 	case IB_WIDTH_1X:  return  1;
284 	case IB_WIDTH_4X:  return  4;
285 	case IB_WIDTH_8X:  return  8;
286 	case IB_WIDTH_12X: return 12;
287 	default: 	  return -1;
288 	}
289 }
290 
291 enum ib_port_speed {
292 	IB_SPEED_SDR	= 1,
293 	IB_SPEED_DDR	= 2,
294 	IB_SPEED_QDR	= 4,
295 	IB_SPEED_FDR10	= 8,
296 	IB_SPEED_FDR	= 16,
297 	IB_SPEED_EDR	= 32
298 };
299 
300 struct ib_protocol_stats {
301 	/* TBD... */
302 };
303 
304 struct iw_protocol_stats {
305 	u64	ipInReceives;
306 	u64	ipInHdrErrors;
307 	u64	ipInTooBigErrors;
308 	u64	ipInNoRoutes;
309 	u64	ipInAddrErrors;
310 	u64	ipInUnknownProtos;
311 	u64	ipInTruncatedPkts;
312 	u64	ipInDiscards;
313 	u64	ipInDelivers;
314 	u64	ipOutForwDatagrams;
315 	u64	ipOutRequests;
316 	u64	ipOutDiscards;
317 	u64	ipOutNoRoutes;
318 	u64	ipReasmTimeout;
319 	u64	ipReasmReqds;
320 	u64	ipReasmOKs;
321 	u64	ipReasmFails;
322 	u64	ipFragOKs;
323 	u64	ipFragFails;
324 	u64	ipFragCreates;
325 	u64	ipInMcastPkts;
326 	u64	ipOutMcastPkts;
327 	u64	ipInBcastPkts;
328 	u64	ipOutBcastPkts;
329 
330 	u64	tcpRtoAlgorithm;
331 	u64	tcpRtoMin;
332 	u64	tcpRtoMax;
333 	u64	tcpMaxConn;
334 	u64	tcpActiveOpens;
335 	u64	tcpPassiveOpens;
336 	u64	tcpAttemptFails;
337 	u64	tcpEstabResets;
338 	u64	tcpCurrEstab;
339 	u64	tcpInSegs;
340 	u64	tcpOutSegs;
341 	u64	tcpRetransSegs;
342 	u64	tcpInErrs;
343 	u64	tcpOutRsts;
344 };
345 
346 union rdma_protocol_stats {
347 	struct ib_protocol_stats	ib;
348 	struct iw_protocol_stats	iw;
349 };
350 
351 /* Define bits for the various functionality this port needs to be supported by
352  * the core.
353  */
354 /* Management                           0x00000FFF */
355 #define RDMA_CORE_CAP_IB_MAD            0x00000001
356 #define RDMA_CORE_CAP_IB_SMI            0x00000002
357 #define RDMA_CORE_CAP_IB_CM             0x00000004
358 #define RDMA_CORE_CAP_IW_CM             0x00000008
359 #define RDMA_CORE_CAP_IB_SA             0x00000010
360 #define RDMA_CORE_CAP_OPA_MAD           0x00000020
361 
362 /* Address format                       0x000FF000 */
363 #define RDMA_CORE_CAP_AF_IB             0x00001000
364 #define RDMA_CORE_CAP_ETH_AH            0x00002000
365 #define RDMA_CORE_CAP_OPA_AH            0x00004000
366 
367 /* Protocol                             0xFFF00000 */
368 #define RDMA_CORE_CAP_PROT_IB           0x00100000
369 #define RDMA_CORE_CAP_PROT_ROCE         0x00200000
370 #define RDMA_CORE_CAP_PROT_IWARP        0x00400000
371 #define RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP 0x00800000
372 #define RDMA_CORE_CAP_PROT_RAW_PACKET   0x01000000
373 #define RDMA_CORE_CAP_PROT_USNIC        0x02000000
374 
375 #define RDMA_CORE_PORT_IBA_IB          (RDMA_CORE_CAP_PROT_IB  \
376 					| RDMA_CORE_CAP_IB_MAD \
377 					| RDMA_CORE_CAP_IB_SMI \
378 					| RDMA_CORE_CAP_IB_CM  \
379 					| RDMA_CORE_CAP_IB_SA  \
380 					| RDMA_CORE_CAP_AF_IB)
381 #define RDMA_CORE_PORT_IBA_ROCE        (RDMA_CORE_CAP_PROT_ROCE \
382 					| RDMA_CORE_CAP_IB_MAD  \
383 					| RDMA_CORE_CAP_IB_CM   \
384 					| RDMA_CORE_CAP_AF_IB   \
385 					| RDMA_CORE_CAP_ETH_AH)
386 #define RDMA_CORE_PORT_IBA_ROCE_UDP_ENCAP			\
387 					(RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP \
388 					| RDMA_CORE_CAP_IB_MAD  \
389 					| RDMA_CORE_CAP_IB_CM   \
390 					| RDMA_CORE_CAP_AF_IB   \
391 					| RDMA_CORE_CAP_ETH_AH)
392 #define RDMA_CORE_PORT_IWARP           (RDMA_CORE_CAP_PROT_IWARP \
393 					| RDMA_CORE_CAP_IW_CM)
394 #define RDMA_CORE_PORT_INTEL_OPA       (RDMA_CORE_PORT_IBA_IB  \
395 					| RDMA_CORE_CAP_OPA_MAD)
396 
397 #define RDMA_CORE_PORT_RAW_PACKET	(RDMA_CORE_CAP_PROT_RAW_PACKET)
398 
399 #define RDMA_CORE_PORT_USNIC		(RDMA_CORE_CAP_PROT_USNIC)
400 
401 struct ib_port_attr {
402 	enum ib_port_state	state;
403 	enum ib_mtu		max_mtu;
404 	enum ib_mtu		active_mtu;
405 	int			gid_tbl_len;
406 	u32			port_cap_flags;
407 	u32			max_msg_sz;
408 	u32			bad_pkey_cntr;
409 	u32			qkey_viol_cntr;
410 	u16			pkey_tbl_len;
411 	u16			lid;
412 	u16			sm_lid;
413 	u8			lmc;
414 	u8			max_vl_num;
415 	u8			sm_sl;
416 	u8			subnet_timeout;
417 	u8			init_type_reply;
418 	u8			active_width;
419 	u8			active_speed;
420 	u8                      phys_state;
421 };
422 
423 enum ib_device_modify_flags {
424 	IB_DEVICE_MODIFY_SYS_IMAGE_GUID	= 1 << 0,
425 	IB_DEVICE_MODIFY_NODE_DESC	= 1 << 1
426 };
427 
428 struct ib_device_modify {
429 	u64	sys_image_guid;
430 	char	node_desc[64];
431 };
432 
433 enum ib_port_modify_flags {
434 	IB_PORT_SHUTDOWN		= 1,
435 	IB_PORT_INIT_TYPE		= (1<<2),
436 	IB_PORT_RESET_QKEY_CNTR		= (1<<3)
437 };
438 
439 struct ib_port_modify {
440 	u32	set_port_cap_mask;
441 	u32	clr_port_cap_mask;
442 	u8	init_type;
443 };
444 
445 enum ib_event_type {
446 	IB_EVENT_CQ_ERR,
447 	IB_EVENT_QP_FATAL,
448 	IB_EVENT_QP_REQ_ERR,
449 	IB_EVENT_QP_ACCESS_ERR,
450 	IB_EVENT_COMM_EST,
451 	IB_EVENT_SQ_DRAINED,
452 	IB_EVENT_PATH_MIG,
453 	IB_EVENT_PATH_MIG_ERR,
454 	IB_EVENT_DEVICE_FATAL,
455 	IB_EVENT_PORT_ACTIVE,
456 	IB_EVENT_PORT_ERR,
457 	IB_EVENT_LID_CHANGE,
458 	IB_EVENT_PKEY_CHANGE,
459 	IB_EVENT_SM_CHANGE,
460 	IB_EVENT_SRQ_ERR,
461 	IB_EVENT_SRQ_LIMIT_REACHED,
462 	IB_EVENT_QP_LAST_WQE_REACHED,
463 	IB_EVENT_CLIENT_REREGISTER,
464 	IB_EVENT_GID_CHANGE,
465 };
466 
467 struct ib_event {
468 	struct ib_device	*device;
469 	union {
470 		struct ib_cq	*cq;
471 		struct ib_qp	*qp;
472 		struct ib_srq	*srq;
473 		u8		port_num;
474 	} element;
475 	enum ib_event_type	event;
476 };
477 
478 struct ib_event_handler {
479 	struct ib_device *device;
480 	void            (*handler)(struct ib_event_handler *, struct ib_event *);
481 	struct list_head  list;
482 };
483 
484 #define INIT_IB_EVENT_HANDLER(_ptr, _device, _handler)		\
485 	do {							\
486 		(_ptr)->device  = _device;			\
487 		(_ptr)->handler = _handler;			\
488 		INIT_LIST_HEAD(&(_ptr)->list);			\
489 	} while (0)
490 
491 struct ib_global_route {
492 	union ib_gid	dgid;
493 	u32		flow_label;
494 	u8		sgid_index;
495 	u8		hop_limit;
496 	u8		traffic_class;
497 };
498 
499 struct ib_grh {
500 	__be32		version_tclass_flow;
501 	__be16		paylen;
502 	u8		next_hdr;
503 	u8		hop_limit;
504 	union ib_gid	sgid;
505 	union ib_gid	dgid;
506 };
507 
508 enum {
509 	IB_MULTICAST_QPN = 0xffffff
510 };
511 
512 #define IB_LID_PERMISSIVE	cpu_to_be16(0xFFFF)
513 
514 enum ib_ah_flags {
515 	IB_AH_GRH	= 1
516 };
517 
518 enum ib_rate {
519 	IB_RATE_PORT_CURRENT = 0,
520 	IB_RATE_2_5_GBPS = 2,
521 	IB_RATE_5_GBPS   = 5,
522 	IB_RATE_10_GBPS  = 3,
523 	IB_RATE_20_GBPS  = 6,
524 	IB_RATE_30_GBPS  = 4,
525 	IB_RATE_40_GBPS  = 7,
526 	IB_RATE_60_GBPS  = 8,
527 	IB_RATE_80_GBPS  = 9,
528 	IB_RATE_120_GBPS = 10,
529 	IB_RATE_14_GBPS  = 11,
530 	IB_RATE_56_GBPS  = 12,
531 	IB_RATE_112_GBPS = 13,
532 	IB_RATE_168_GBPS = 14,
533 	IB_RATE_25_GBPS  = 15,
534 	IB_RATE_100_GBPS = 16,
535 	IB_RATE_200_GBPS = 17,
536 	IB_RATE_300_GBPS = 18
537 };
538 
539 enum ib_mr_create_flags {
540 		IB_MR_SIGNATURE_EN = 1,
541 };
542 
543 /**
544  * ib_mr_init_attr - Memory region init attributes passed to routine
545  *     ib_create_mr.
546  * @max_reg_descriptors: max number of registration descriptors that
547  *     may be used with registration work requests.
548  * @flags: MR creation flags bit mask.
549  */
550 struct ib_mr_init_attr {
551 	int	    max_reg_descriptors;
552 	u32	    flags;
553 };
554 
555 /**
556  * ib_rate_to_mult - Convert the IB rate enum to a multiple of the
557  * base rate of 2.5 Gbit/sec.  For example, IB_RATE_5_GBPS will be
558  * converted to 2, since 5 Gbit/sec is 2 * 2.5 Gbit/sec.
559  * @rate: rate to convert.
560  */
561 int ib_rate_to_mult(enum ib_rate rate) __attribute_const__;
562 
563 /**
564  * ib_rate_to_mbps - Convert the IB rate enum to Mbps.
565  * For example, IB_RATE_2_5_GBPS will be converted to 2500.
566  * @rate: rate to convert.
567  */
568 int ib_rate_to_mbps(enum ib_rate rate) __attribute_const__;
569 
570 struct ib_cq_init_attr {
571 	int	cqe;
572 	int	comp_vector;
573 	u32	flags;
574 };
575 
576 enum ib_signature_type {
577 	IB_SIG_TYPE_T10_DIF,
578 };
579 
580 /**
581  * T10-DIF Signature types
582  * T10-DIF types are defined by SCSI
583  * specifications.
584  */
585 enum ib_t10_dif_type {
586 	IB_T10DIF_NONE,
587 	IB_T10DIF_TYPE1,
588 	IB_T10DIF_TYPE2,
589 	IB_T10DIF_TYPE3
590 };
591 
592 /**
593  * Signature T10-DIF block-guard types
594  * IB_T10DIF_CRC: Corresponds to T10-PI mandated CRC checksum rules.
595  * IB_T10DIF_CSUM: Corresponds to IP checksum rules.
596  */
597 enum ib_t10_dif_bg_type {
598 	IB_T10DIF_CRC,
599 	IB_T10DIF_CSUM
600 };
601 
602 /**
603  * struct ib_t10_dif_domain - Parameters specific for T10-DIF
604  *     domain.
605  * @type: T10-DIF type (0|1|2|3)
606  * @bg_type: T10-DIF block guard type (CRC|CSUM)
607  * @pi_interval: protection information interval.
608  * @bg: seed of guard computation.
609  * @app_tag: application tag of guard block
610  * @ref_tag: initial guard block reference tag.
611  * @type3_inc_reftag: T10-DIF type 3 does not state
612  *     about the reference tag, it is the user
613  *     choice to increment it or not.
614  */
615 struct ib_t10_dif_domain {
616 	enum ib_t10_dif_type	type;
617 	enum ib_t10_dif_bg_type bg_type;
618 	u32			pi_interval;
619 	u16			bg;
620 	u16			app_tag;
621 	u32			ref_tag;
622 	bool			type3_inc_reftag;
623 };
624 
625 /**
626  * struct ib_sig_domain - Parameters for signature domain
627  * @sig_type: specific signauture type
628  * @sig: union of all signature domain attributes that may
629  *     be used to set domain layout.
630  */
631 struct ib_sig_domain {
632 	enum ib_signature_type sig_type;
633 	union {
634 		struct ib_t10_dif_domain dif;
635 	} sig;
636 };
637 
638 /**
639  * struct ib_sig_attrs - Parameters for signature handover operation
640  * @check_mask: bitmask for signature byte check (8 bytes)
641  * @mem: memory domain layout desciptor.
642  * @wire: wire domain layout desciptor.
643  */
644 struct ib_sig_attrs {
645 	u8			check_mask;
646 	struct ib_sig_domain	mem;
647 	struct ib_sig_domain	wire;
648 };
649 
650 enum ib_sig_err_type {
651 	IB_SIG_BAD_GUARD,
652 	IB_SIG_BAD_REFTAG,
653 	IB_SIG_BAD_APPTAG,
654 };
655 
656 /**
657  * struct ib_sig_err - signature error descriptor
658  */
659 struct ib_sig_err {
660 	enum ib_sig_err_type	err_type;
661 	u32			expected;
662 	u32			actual;
663 	u64			sig_err_offset;
664 	u32			key;
665 };
666 
667 enum ib_mr_status_check {
668 	IB_MR_CHECK_SIG_STATUS = 1,
669 };
670 
671 /**
672  * struct ib_mr_status - Memory region status container
673  *
674  * @fail_status: Bitmask of MR checks status. For each
675  *     failed check a corresponding status bit is set.
676  * @sig_err: Additional info for IB_MR_CEHCK_SIG_STATUS
677  *     failure.
678  */
679 struct ib_mr_status {
680 	u32		    fail_status;
681 	struct ib_sig_err   sig_err;
682 };
683 
684 /**
685  * mult_to_ib_rate - Convert a multiple of 2.5 Gbit/sec to an IB rate
686  * enum.
687  * @mult: multiple to convert.
688  */
689 enum ib_rate mult_to_ib_rate(int mult) __attribute_const__;
690 
691 struct ib_ah_attr {
692 	struct ib_global_route	grh;
693 	u16			dlid;
694 	u8			sl;
695 	u8			src_path_bits;
696 	u8			static_rate;
697 	u8			ah_flags;
698 	u8			port_num;
699 	u8			dmac[6];
700 	u16			vlan_id;
701 };
702 
703 enum ib_wc_status {
704 	IB_WC_SUCCESS,
705 	IB_WC_LOC_LEN_ERR,
706 	IB_WC_LOC_QP_OP_ERR,
707 	IB_WC_LOC_EEC_OP_ERR,
708 	IB_WC_LOC_PROT_ERR,
709 	IB_WC_WR_FLUSH_ERR,
710 	IB_WC_MW_BIND_ERR,
711 	IB_WC_BAD_RESP_ERR,
712 	IB_WC_LOC_ACCESS_ERR,
713 	IB_WC_REM_INV_REQ_ERR,
714 	IB_WC_REM_ACCESS_ERR,
715 	IB_WC_REM_OP_ERR,
716 	IB_WC_RETRY_EXC_ERR,
717 	IB_WC_RNR_RETRY_EXC_ERR,
718 	IB_WC_LOC_RDD_VIOL_ERR,
719 	IB_WC_REM_INV_RD_REQ_ERR,
720 	IB_WC_REM_ABORT_ERR,
721 	IB_WC_INV_EECN_ERR,
722 	IB_WC_INV_EEC_STATE_ERR,
723 	IB_WC_FATAL_ERR,
724 	IB_WC_RESP_TIMEOUT_ERR,
725 	IB_WC_GENERAL_ERR
726 };
727 
728 enum ib_wc_opcode {
729 	IB_WC_SEND,
730 	IB_WC_RDMA_WRITE,
731 	IB_WC_RDMA_READ,
732 	IB_WC_COMP_SWAP,
733 	IB_WC_FETCH_ADD,
734 	IB_WC_BIND_MW,
735 	IB_WC_LSO,
736 	IB_WC_LOCAL_INV,
737 	IB_WC_FAST_REG_MR,
738 	IB_WC_MASKED_COMP_SWAP,
739 	IB_WC_MASKED_FETCH_ADD,
740 /*
741  * Set value of IB_WC_RECV so consumers can test if a completion is a
742  * receive by testing (opcode & IB_WC_RECV).
743  */
744 	IB_WC_RECV			= 1 << 7,
745 	IB_WC_RECV_RDMA_WITH_IMM
746 };
747 
748 enum ib_wc_flags {
749 	IB_WC_GRH		= 1,
750 	IB_WC_WITH_IMM		= (1<<1),
751 	IB_WC_WITH_INVALIDATE	= (1<<2),
752 	IB_WC_IP_CSUM_OK	= (1<<3),
753 	IB_WC_WITH_SL		= (1<<4),
754 	IB_WC_WITH_SLID		= (1<<5),
755 	IB_WC_WITH_TIMESTAMP	= (1<<6),
756 	IB_WC_WITH_SMAC		= (1<<7),
757 	IB_WC_WITH_VLAN		= (1<<8),
758 };
759 
760 struct ib_wc {
761 	u64			wr_id;
762 	enum ib_wc_status	status;
763 	enum ib_wc_opcode	opcode;
764 	u32			vendor_err;
765 	u32			byte_len;
766 	struct ib_qp	       *qp;
767 	union {
768 		__be32		imm_data;
769 		u32		invalidate_rkey;
770 	} ex;
771 	u32			src_qp;
772 	int			wc_flags;
773 	u16			pkey_index;
774 	u16			slid;
775 	u8			sl;
776 	u8			dlid_path_bits;
777 	u8			port_num;	/* valid only for DR SMPs on switches */
778 	int			csum_ok;
779 	struct {
780 		uint64_t	timestamp; /* timestamp = 0 indicates error*/
781 	} ts;
782 	u8			smac[6];
783 	u16			vlan_id;
784 };
785 
786 enum ib_cq_notify_flags {
787 	IB_CQ_SOLICITED			= 1 << 0,
788 	IB_CQ_NEXT_COMP			= 1 << 1,
789 	IB_CQ_SOLICITED_MASK		= IB_CQ_SOLICITED | IB_CQ_NEXT_COMP,
790 	IB_CQ_REPORT_MISSED_EVENTS	= 1 << 2,
791 };
792 
793 enum ib_srq_type {
794 	IB_SRQT_BASIC,
795 	IB_SRQT_XRC
796 };
797 
798 enum ib_srq_attr_mask {
799 	IB_SRQ_MAX_WR	= 1 << 0,
800 	IB_SRQ_LIMIT	= 1 << 1,
801 };
802 
803 struct ib_srq_attr {
804 	u32	max_wr;
805 	u32	max_sge;
806 	u32	srq_limit;
807 };
808 
809 struct ib_srq_init_attr {
810 	void		      (*event_handler)(struct ib_event *, void *);
811 	void		       *srq_context;
812 	struct ib_srq_attr	attr;
813 	enum ib_srq_type	srq_type;
814 
815 	union {
816 		struct {
817 			struct ib_xrcd *xrcd;
818 			struct ib_cq   *cq;
819 		} xrc;
820 	} ext;
821 };
822 
823 struct ib_qp_cap {
824 	u32	max_send_wr;
825 	u32	max_recv_wr;
826 	u32	max_send_sge;
827 	u32	max_recv_sge;
828 	u32	max_inline_data;
829 	u32	qpg_tss_mask_sz;
830 };
831 
832 enum ib_sig_type {
833 	IB_SIGNAL_ALL_WR,
834 	IB_SIGNAL_REQ_WR
835 };
836 
837 enum ib_qp_type {
838 	/*
839 	 * IB_QPT_SMI and IB_QPT_GSI have to be the first two entries
840 	 * here (and in that order) since the MAD layer uses them as
841 	 * indices into a 2-entry table.
842 	 */
843 	IB_QPT_SMI,
844 	IB_QPT_GSI,
845 
846 	IB_QPT_RC,
847 	IB_QPT_UC,
848 	IB_QPT_UD,
849 	IB_QPT_RAW_IPV6,
850 	IB_QPT_RAW_ETHERTYPE,
851 	IB_QPT_RAW_PACKET = 8,
852 	IB_QPT_XRC_INI = 9,
853 	IB_QPT_XRC_TGT,
854 	IB_QPT_DC_INI,
855 	IB_QPT_MAX,
856 	/* Reserve a range for qp types internal to the low level driver.
857 	 * These qp types will not be visible at the IB core layer, so the
858 	 * IB_QPT_MAX usages should not be affected in the core layer
859 	 */
860 	IB_QPT_RESERVED1 = 0x1000,
861 	IB_QPT_RESERVED2,
862 	IB_QPT_RESERVED3,
863 	IB_QPT_RESERVED4,
864 	IB_QPT_RESERVED5,
865 	IB_QPT_RESERVED6,
866 	IB_QPT_RESERVED7,
867 	IB_QPT_RESERVED8,
868 	IB_QPT_RESERVED9,
869 	IB_QPT_RESERVED10,
870 };
871 
872 enum ib_qp_create_flags {
873 	IB_QP_CREATE_IPOIB_UD_LSO		= 1 << 0,
874 	IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK	= 1 << 1,
875 	IB_QP_CREATE_CROSS_CHANNEL		= 1 << 2,
876 	IB_QP_CREATE_MANAGED_SEND		= 1 << 3,
877 	IB_QP_CREATE_MANAGED_RECV		= 1 << 4,
878 	IB_QP_CREATE_NETIF_QP			= 1 << 5,
879 	IB_QP_CREATE_SIGNATURE_EN		= 1 << 6,
880 	/* reserve bits 26-31 for low level drivers' internal use */
881 	IB_QP_CREATE_RESERVED_START		= 1 << 26,
882 	IB_QP_CREATE_RESERVED_END		= 1 << 31,
883 };
884 
885 enum ib_qpg_type {
886 	IB_QPG_NONE	= 0,
887 	IB_QPG_PARENT	= (1<<0),
888 	IB_QPG_CHILD_RX = (1<<1),
889 	IB_QPG_CHILD_TX = (1<<2)
890 };
891 
892 struct ib_qpg_init_attrib {
893 	u32 tss_child_count;
894 	u32 rss_child_count;
895 };
896 
897 struct ib_qp_init_attr {
898 	void                  (*event_handler)(struct ib_event *, void *);
899 	void		       *qp_context;
900 	struct ib_cq	       *send_cq;
901 	struct ib_cq	       *recv_cq;
902 	struct ib_srq	       *srq;
903 	struct ib_xrcd	       *xrcd;     /* XRC TGT QPs only */
904 	struct ib_qp_cap	cap;
905 	union {
906 		struct ib_qp *qpg_parent; /* see qpg_type */
907 		struct ib_qpg_init_attrib parent_attrib;
908 	};
909 	enum ib_sig_type	sq_sig_type;
910 	enum ib_qp_type		qp_type;
911 	enum ib_qp_create_flags	create_flags;
912 	enum ib_qpg_type	qpg_type;
913 	u8			port_num; /* special QP types only */
914 };
915 
916 enum {
917 	IB_DCT_CREATE_FLAG_RCV_INLINE		= 1 << 0,
918 	IB_DCT_CREATE_FLAGS_MASK		= IB_DCT_CREATE_FLAG_RCV_INLINE,
919 };
920 
921 struct ib_dct_init_attr {
922 	struct ib_pd	       *pd;
923 	struct ib_cq	       *cq;
924 	struct ib_srq	       *srq;
925 	u64			dc_key;
926 	u8			port;
927 	u32			access_flags;
928 	u8			min_rnr_timer;
929 	u8			tclass;
930 	u32			flow_label;
931 	enum ib_mtu		mtu;
932 	u8			pkey_index;
933 	u8			gid_index;
934 	u8			hop_limit;
935 	u32			create_flags;
936 };
937 
938 struct ib_dct_attr {
939 	u64			dc_key;
940 	u8			port;
941 	u32			access_flags;
942 	u8			min_rnr_timer;
943 	u8			tclass;
944 	u32			flow_label;
945 	enum ib_mtu		mtu;
946 	u8			pkey_index;
947 	u8			gid_index;
948 	u8			hop_limit;
949 	u32			key_violations;
950 	u8			state;
951 };
952 
953 struct ib_qp_open_attr {
954 	void                  (*event_handler)(struct ib_event *, void *);
955 	void		       *qp_context;
956 	u32			qp_num;
957 	enum ib_qp_type		qp_type;
958 };
959 
960 enum ib_rnr_timeout {
961 	IB_RNR_TIMER_655_36 =  0,
962 	IB_RNR_TIMER_000_01 =  1,
963 	IB_RNR_TIMER_000_02 =  2,
964 	IB_RNR_TIMER_000_03 =  3,
965 	IB_RNR_TIMER_000_04 =  4,
966 	IB_RNR_TIMER_000_06 =  5,
967 	IB_RNR_TIMER_000_08 =  6,
968 	IB_RNR_TIMER_000_12 =  7,
969 	IB_RNR_TIMER_000_16 =  8,
970 	IB_RNR_TIMER_000_24 =  9,
971 	IB_RNR_TIMER_000_32 = 10,
972 	IB_RNR_TIMER_000_48 = 11,
973 	IB_RNR_TIMER_000_64 = 12,
974 	IB_RNR_TIMER_000_96 = 13,
975 	IB_RNR_TIMER_001_28 = 14,
976 	IB_RNR_TIMER_001_92 = 15,
977 	IB_RNR_TIMER_002_56 = 16,
978 	IB_RNR_TIMER_003_84 = 17,
979 	IB_RNR_TIMER_005_12 = 18,
980 	IB_RNR_TIMER_007_68 = 19,
981 	IB_RNR_TIMER_010_24 = 20,
982 	IB_RNR_TIMER_015_36 = 21,
983 	IB_RNR_TIMER_020_48 = 22,
984 	IB_RNR_TIMER_030_72 = 23,
985 	IB_RNR_TIMER_040_96 = 24,
986 	IB_RNR_TIMER_061_44 = 25,
987 	IB_RNR_TIMER_081_92 = 26,
988 	IB_RNR_TIMER_122_88 = 27,
989 	IB_RNR_TIMER_163_84 = 28,
990 	IB_RNR_TIMER_245_76 = 29,
991 	IB_RNR_TIMER_327_68 = 30,
992 	IB_RNR_TIMER_491_52 = 31
993 };
994 
995 enum ib_qp_attr_mask {
996 	IB_QP_STATE			= 1,
997 	IB_QP_CUR_STATE			= (1<<1),
998 	IB_QP_EN_SQD_ASYNC_NOTIFY	= (1<<2),
999 	IB_QP_ACCESS_FLAGS		= (1<<3),
1000 	IB_QP_PKEY_INDEX		= (1<<4),
1001 	IB_QP_PORT			= (1<<5),
1002 	IB_QP_QKEY			= (1<<6),
1003 	IB_QP_AV			= (1<<7),
1004 	IB_QP_PATH_MTU			= (1<<8),
1005 	IB_QP_TIMEOUT			= (1<<9),
1006 	IB_QP_RETRY_CNT			= (1<<10),
1007 	IB_QP_RNR_RETRY			= (1<<11),
1008 	IB_QP_RQ_PSN			= (1<<12),
1009 	IB_QP_MAX_QP_RD_ATOMIC		= (1<<13),
1010 	IB_QP_ALT_PATH			= (1<<14),
1011 	IB_QP_MIN_RNR_TIMER		= (1<<15),
1012 	IB_QP_SQ_PSN			= (1<<16),
1013 	IB_QP_MAX_DEST_RD_ATOMIC	= (1<<17),
1014 	IB_QP_PATH_MIG_STATE		= (1<<18),
1015 	IB_QP_CAP			= (1<<19),
1016 	IB_QP_DEST_QPN			= (1<<20),
1017 	IB_QP_GROUP_RSS			= (1<<21),
1018 	IB_QP_DC_KEY			= (1<<22),
1019 	IB_QP_SMAC			= (1<<23),
1020 	IB_QP_ALT_SMAC			= (1<<24),
1021 	IB_QP_VID			= (1<<25),
1022 	IB_QP_ALT_VID			= (1<<26)
1023 };
1024 
1025 enum ib_qp_state {
1026 	IB_QPS_RESET,
1027 	IB_QPS_INIT,
1028 	IB_QPS_RTR,
1029 	IB_QPS_RTS,
1030 	IB_QPS_SQD,
1031 	IB_QPS_SQE,
1032 	IB_QPS_ERR,
1033 	IB_QPS_DUMMY = -1	/* force enum signed */
1034 };
1035 
1036 enum ib_mig_state {
1037 	IB_MIG_MIGRATED,
1038 	IB_MIG_REARM,
1039 	IB_MIG_ARMED
1040 };
1041 
1042 enum ib_mw_type {
1043 	IB_MW_TYPE_1 = 1,
1044 	IB_MW_TYPE_2 = 2
1045 };
1046 
1047 struct ib_qp_attr {
1048 	enum ib_qp_state	qp_state;
1049 	enum ib_qp_state	cur_qp_state;
1050 	enum ib_mtu		path_mtu;
1051 	enum ib_mig_state	path_mig_state;
1052 	u32			qkey;
1053 	u32			rq_psn;
1054 	u32			sq_psn;
1055 	u32			dest_qp_num;
1056 	int			qp_access_flags;
1057 	struct ib_qp_cap	cap;
1058 	struct ib_ah_attr	ah_attr;
1059 	struct ib_ah_attr	alt_ah_attr;
1060 	u16			pkey_index;
1061 	u16			alt_pkey_index;
1062 	u8			en_sqd_async_notify;
1063 	u8			sq_draining;
1064 	u8			max_rd_atomic;
1065 	u8			max_dest_rd_atomic;
1066 	u8			min_rnr_timer;
1067 	u8			port_num;
1068 	u8			timeout;
1069 	u8			retry_cnt;
1070 	u8			rnr_retry;
1071 	u8			alt_port_num;
1072 	u8			alt_timeout;
1073 	u8                      smac[ETH_ALEN];
1074 	u8                      alt_smac[ETH_ALEN];
1075 	u16                     vlan_id;
1076 	u16                     alt_vlan_id;
1077 
1078 };
1079 
1080 struct ib_qp_attr_ex {
1081 	enum ib_qp_state	qp_state;
1082 	enum ib_qp_state	cur_qp_state;
1083 	enum ib_mtu		path_mtu;
1084 	enum ib_mig_state	path_mig_state;
1085 	u32			qkey;
1086 	u32			rq_psn;
1087 	u32			sq_psn;
1088 	u32			dest_qp_num;
1089 	int			qp_access_flags;
1090 	struct ib_qp_cap	cap;
1091 	struct ib_ah_attr	ah_attr;
1092 	struct ib_ah_attr	alt_ah_attr;
1093 	u16			pkey_index;
1094 	u16			alt_pkey_index;
1095 	u8			en_sqd_async_notify;
1096 	u8			sq_draining;
1097 	u8			max_rd_atomic;
1098 	u8			max_dest_rd_atomic;
1099 	u8			min_rnr_timer;
1100 	u8			port_num;
1101 	u8			timeout;
1102 	u8			retry_cnt;
1103 	u8			rnr_retry;
1104 	u8			alt_port_num;
1105 	u8			alt_timeout;
1106 	u64			dct_key;
1107 };
1108 
1109 enum ib_wr_opcode {
1110 	IB_WR_RDMA_WRITE,
1111 	IB_WR_RDMA_WRITE_WITH_IMM,
1112 	IB_WR_SEND,
1113 	IB_WR_SEND_WITH_IMM,
1114 	IB_WR_RDMA_READ,
1115 	IB_WR_ATOMIC_CMP_AND_SWP,
1116 	IB_WR_ATOMIC_FETCH_AND_ADD,
1117 	IB_WR_LSO,
1118 	IB_WR_SEND_WITH_INV,
1119 	IB_WR_RDMA_READ_WITH_INV,
1120 	IB_WR_LOCAL_INV,
1121 	IB_WR_FAST_REG_MR,
1122 	IB_WR_MASKED_ATOMIC_CMP_AND_SWP,
1123 	IB_WR_MASKED_ATOMIC_FETCH_AND_ADD,
1124 	IB_WR_BIND_MW,
1125 	IB_WR_REG_SIG_MR,
1126 	/* reserve values for low level drivers' internal use.
1127 	 * These values will not be used at all in the ib core layer.
1128 	 */
1129 	IB_WR_RESERVED1 = 0xf0,
1130 	IB_WR_RESERVED2,
1131 	IB_WR_RESERVED3,
1132 	IB_WR_RESERVED4,
1133 	IB_WR_RESERVED5,
1134 	IB_WR_RESERVED6,
1135 	IB_WR_RESERVED7,
1136 	IB_WR_RESERVED8,
1137 	IB_WR_RESERVED9,
1138 	IB_WR_RESERVED10,
1139 };
1140 
1141 enum ib_send_flags {
1142 	IB_SEND_FENCE		= 1,
1143 	IB_SEND_SIGNALED	= (1<<1),
1144 	IB_SEND_SOLICITED	= (1<<2),
1145 	IB_SEND_INLINE		= (1<<3),
1146 	IB_SEND_IP_CSUM		= (1<<4),
1147 
1148 	/* reserve bits 26-31 for low level drivers' internal use */
1149 	IB_SEND_RESERVED_START	= (1 << 26),
1150 	IB_SEND_RESERVED_END	= (1 << 31),
1151 	IB_SEND_UMR_UNREG       = (1<<5)
1152 };
1153 
1154 struct ib_sge {
1155 	u64	addr;
1156 	u32	length;
1157 	u32	lkey;
1158 };
1159 
1160 struct ib_fast_reg_page_list {
1161 	struct ib_device       *device;
1162 	u64		       *page_list;
1163 	unsigned int		max_page_list_len;
1164 };
1165 
1166 /**
1167  * struct ib_mw_bind_info - Parameters for a memory window bind operation.
1168  * @mr: A memory region to bind the memory window to.
1169  * @addr: The address where the memory window should begin.
1170  * @length: The length of the memory window, in bytes.
1171  * @mw_access_flags: Access flags from enum ib_access_flags for the window.
1172  *
1173  * This struct contains the shared parameters for type 1 and type 2
1174  * memory window bind operations.
1175  */
1176 struct ib_mw_bind_info {
1177 	struct ib_mr   *mr;
1178 	u64		addr;
1179 	u64		length;
1180 	int		mw_access_flags;
1181 };
1182 
1183 struct ib_send_wr {
1184 	struct ib_send_wr      *next;
1185 	u64			wr_id;
1186 	struct ib_sge	       *sg_list;
1187 	int			num_sge;
1188 	enum ib_wr_opcode	opcode;
1189 	int			send_flags;
1190 	union {
1191 		__be32		imm_data;
1192 		u32		invalidate_rkey;
1193 	} ex;
1194 	union {
1195 		struct {
1196 			u64	remote_addr;
1197 			u32	rkey;
1198 		} rdma;
1199 		struct {
1200 			u64	remote_addr;
1201 			u64	compare_add;
1202 			u64	swap;
1203 			u64	compare_add_mask;
1204 			u64	swap_mask;
1205 			u32	rkey;
1206 		} atomic;
1207 		struct {
1208 			struct ib_ah *ah;
1209 			void   *header;
1210 			int     hlen;
1211 			int     mss;
1212 			u32	remote_qpn;
1213 			u32	remote_qkey;
1214 			u16	pkey_index; /* valid for GSI only */
1215 			u8	port_num;   /* valid for DR SMPs on switch only */
1216 		} ud;
1217 		struct {
1218 			u64				iova_start;
1219 			struct ib_fast_reg_page_list   *page_list;
1220 			unsigned int			page_shift;
1221 			unsigned int			page_list_len;
1222 			u32				length;
1223 			int				access_flags;
1224 			u32				rkey;
1225 		} fast_reg;
1226 		struct {
1227 			int		npages;
1228 			int		access_flags;
1229 			u32		mkey;
1230 			struct ib_pd   *pd;
1231 			u64		virt_addr;
1232 			u64		length;
1233 			int		page_shift;
1234 		} umr;
1235 		struct {
1236 			struct ib_mw            *mw;
1237 			/* The new rkey for the memory window. */
1238 			u32                      rkey;
1239 			struct ib_mw_bind_info   bind_info;
1240 		} bind_mw;
1241 		struct {
1242 			struct ib_sig_attrs    *sig_attrs;
1243 			struct ib_mr	       *sig_mr;
1244 			int			access_flags;
1245 			struct ib_sge	       *prot;
1246 		} sig_handover;
1247 	} wr;
1248 	u32			xrc_remote_srq_num;	/* XRC TGT QPs only */
1249 };
1250 
1251 struct ib_recv_wr {
1252 	struct ib_recv_wr      *next;
1253 	u64			wr_id;
1254 	struct ib_sge	       *sg_list;
1255 	int			num_sge;
1256 };
1257 
1258 enum ib_access_flags {
1259 	IB_ACCESS_LOCAL_WRITE	= 1,
1260 	IB_ACCESS_REMOTE_WRITE	= (1<<1),
1261 	IB_ACCESS_REMOTE_READ	= (1<<2),
1262 	IB_ACCESS_REMOTE_ATOMIC	= (1<<3),
1263 	IB_ACCESS_MW_BIND	= (1<<4),
1264 	IB_ACCESS_ALLOCATE_MR	= (1<<5),
1265 	IB_ZERO_BASED		= (1<<13)
1266 };
1267 
1268 struct ib_phys_buf {
1269 	u64      addr;
1270 	u64      size;
1271 };
1272 
1273 struct ib_mr_attr {
1274 	struct ib_pd	*pd;
1275 	u64		device_virt_addr;
1276 	u64		size;
1277 	int		mr_access_flags;
1278 	u32		lkey;
1279 	u32		rkey;
1280 };
1281 
1282 enum ib_mr_rereg_flags {
1283 	IB_MR_REREG_TRANS	= 1,
1284 	IB_MR_REREG_PD		= (1<<1),
1285 	IB_MR_REREG_ACCESS	= (1<<2)
1286 };
1287 
1288 /**
1289  * struct ib_mw_bind - Parameters for a type 1 memory window bind operation.
1290  * @wr_id:      Work request id.
1291  * @send_flags: Flags from ib_send_flags enum.
1292  * @bind_info:  More parameters of the bind operation.
1293  */
1294 struct ib_mw_bind {
1295 	u64		wr_id;
1296 	int		send_flags;
1297 	struct ib_mw_bind_info bind_info;
1298 };
1299 
1300 struct ib_fmr_attr {
1301 	int	max_pages;
1302 	int	max_maps;
1303 	u8	page_shift;
1304 };
1305 
1306 struct ib_ucontext {
1307 	struct ib_device       *device;
1308 	struct list_head	pd_list;
1309 	struct list_head	mr_list;
1310 	struct list_head	mw_list;
1311 	struct list_head	cq_list;
1312 	struct list_head	qp_list;
1313 	struct list_head	srq_list;
1314 	struct list_head	ah_list;
1315 	struct list_head	xrcd_list;
1316 	struct list_head	rule_list;
1317 	struct list_head	dct_list;
1318 	int			closing;
1319 	void		 *peer_mem_private_data;
1320 	char		 *peer_mem_name;
1321 };
1322 
1323 struct ib_uobject {
1324 	u64			user_handle;	/* handle given to us by userspace */
1325 	struct ib_ucontext     *context;	/* associated user context */
1326 	void		       *object;		/* containing object */
1327 	struct list_head	list;		/* link to context's list */
1328 	int			id;		/* index into kernel idr */
1329 	struct kref		ref;
1330 	struct rw_semaphore	mutex;		/* protects .live */
1331 	int			live;
1332 };
1333 
1334 struct ib_udata;
1335 struct ib_udata_ops {
1336 	int		      (*copy_from)(void *dest, struct ib_udata *udata,
1337 					   size_t len);
1338 	int		      (*copy_to)(struct ib_udata *udata, void *src,
1339 					 size_t len);
1340 };
1341 
1342 struct ib_udata {
1343 	struct ib_udata_ops    *ops;
1344 	void __user *inbuf;
1345 	void __user *outbuf;
1346 	size_t       inlen;
1347 	size_t       outlen;
1348 };
1349 
1350 struct ib_pd {
1351 	struct ib_device       *device;
1352 	struct ib_uobject      *uobject;
1353 	atomic_t          	usecnt; /* count all resources */
1354 };
1355 
1356 struct ib_xrcd {
1357 	struct ib_device       *device;
1358 	atomic_t		usecnt; /* count all exposed resources */
1359 	struct inode	       *inode;
1360 
1361 	struct mutex		tgt_qp_mutex;
1362 	struct list_head	tgt_qp_list;
1363 };
1364 
1365 struct ib_ah {
1366 	struct ib_device	*device;
1367 	struct ib_pd		*pd;
1368 	struct ib_uobject	*uobject;
1369 };
1370 
1371 enum ib_cq_attr_mask {
1372 	IB_CQ_MODERATION               = (1 << 0),
1373 	IB_CQ_CAP_FLAGS                = (1 << 1)
1374 };
1375 
1376 enum ib_cq_cap_flags {
1377 	IB_CQ_IGNORE_OVERRUN           = (1 << 0)
1378 };
1379 
1380 struct ib_cq_attr {
1381 	struct {
1382 		u16     cq_count;
1383 		u16     cq_period;
1384 	} moderation;
1385 	u32     cq_cap_flags;
1386 };
1387 
1388 typedef void (*ib_comp_handler)(struct ib_cq *cq, void *cq_context);
1389 
1390 struct ib_cq {
1391 	struct ib_device       *device;
1392 	struct ib_uobject      *uobject;
1393 	ib_comp_handler   	comp_handler;
1394 	void                  (*event_handler)(struct ib_event *, void *);
1395 	void                   *cq_context;
1396 	int               	cqe;
1397 	atomic_t          	usecnt; /* count number of work queues */
1398 };
1399 
1400 struct ib_srq {
1401 	struct ib_device       *device;
1402 	struct ib_pd	       *pd;
1403 	struct ib_uobject      *uobject;
1404 	void		      (*event_handler)(struct ib_event *, void *);
1405 	void		       *srq_context;
1406 	enum ib_srq_type	srq_type;
1407 	atomic_t		usecnt;
1408 
1409 	union {
1410 		struct {
1411 			struct ib_xrcd *xrcd;
1412 			struct ib_cq   *cq;
1413 			u32		srq_num;
1414 		} xrc;
1415 	} ext;
1416 };
1417 
1418 struct ib_qp {
1419 	struct ib_device       *device;
1420 	struct ib_pd	       *pd;
1421 	struct ib_cq	       *send_cq;
1422 	struct ib_cq	       *recv_cq;
1423 	struct ib_srq	       *srq;
1424 	struct ib_xrcd	       *xrcd; /* XRC TGT QPs only */
1425 	struct list_head	xrcd_list;
1426 	/* count times opened, mcast attaches, flow attaches */
1427 	atomic_t		usecnt;
1428 	struct list_head	open_list;
1429 	struct ib_qp           *real_qp;
1430 	struct ib_uobject      *uobject;
1431 	void                  (*event_handler)(struct ib_event *, void *);
1432 	void		       *qp_context;
1433 	u32			qp_num;
1434 	enum ib_qp_type		qp_type;
1435 	enum ib_qpg_type	qpg_type;
1436 	u8			port_num;
1437 };
1438 
1439 struct ib_dct {
1440 	struct ib_device       *device;
1441 	struct ib_uobject      *uobject;
1442 	struct ib_pd	       *pd;
1443 	struct ib_cq	       *cq;
1444 	struct ib_srq	       *srq;
1445 	u32			dct_num;
1446 };
1447 
1448 struct ib_mr {
1449 	struct ib_device  *device;
1450 	struct ib_pd	  *pd;
1451 	struct ib_uobject *uobject;
1452 	u32		   lkey;
1453 	u32		   rkey;
1454 	atomic_t	   usecnt; /* count number of MWs */
1455 };
1456 
1457 struct ib_mw {
1458 	struct ib_device	*device;
1459 	struct ib_pd		*pd;
1460 	struct ib_uobject	*uobject;
1461 	u32			rkey;
1462 	enum ib_mw_type         type;
1463 };
1464 
1465 struct ib_fmr {
1466 	struct ib_device	*device;
1467 	struct ib_pd		*pd;
1468 	struct list_head	list;
1469 	u32			lkey;
1470 	u32			rkey;
1471 };
1472 
1473 /* Supported steering options */
1474 enum ib_flow_attr_type {
1475 	/* steering according to rule specifications */
1476 	IB_FLOW_ATTR_NORMAL		= 0x0,
1477 	/* default unicast and multicast rule -
1478 	 * receive all Eth traffic which isn't steered to any QP
1479 	 */
1480 	IB_FLOW_ATTR_ALL_DEFAULT	= 0x1,
1481 	/* default multicast rule -
1482 	 * receive all Eth multicast traffic which isn't steered to any QP
1483 	 */
1484 	IB_FLOW_ATTR_MC_DEFAULT		= 0x2,
1485 	/* sniffer rule - receive all port traffic */
1486 	IB_FLOW_ATTR_SNIFFER		= 0x3
1487 };
1488 
1489 /* Supported steering header types */
1490 enum ib_flow_spec_type {
1491 	/* L2 headers*/
1492 	IB_FLOW_SPEC_ETH	= 0x20,
1493 	IB_FLOW_SPEC_IB		= 0x21,
1494 	/* L3 header*/
1495 	IB_FLOW_SPEC_IPV4	= 0x30,
1496 	/* L4 headers*/
1497 	IB_FLOW_SPEC_TCP	= 0x40,
1498 	IB_FLOW_SPEC_UDP	= 0x41
1499 };
1500 
1501 #define IB_FLOW_SPEC_SUPPORT_LAYERS 4
1502 
1503 /* Flow steering rule priority is set according to it's domain.
1504  * Lower domain value means higher priority.
1505  */
1506 enum ib_flow_domain {
1507 	IB_FLOW_DOMAIN_USER,
1508 	IB_FLOW_DOMAIN_ETHTOOL,
1509 	IB_FLOW_DOMAIN_RFS,
1510 	IB_FLOW_DOMAIN_NIC,
1511 	IB_FLOW_DOMAIN_NUM /* Must be last */
1512 };
1513 
1514 enum ib_flow_flags {
1515 	IB_FLOW_ATTR_FLAGS_ALLOW_LOOP_BACK = 1
1516 };
1517 
1518 struct ib_flow_eth_filter {
1519 	u8	dst_mac[6];
1520 	u8	src_mac[6];
1521 	__be16	ether_type;
1522 	__be16	vlan_tag;
1523 };
1524 
1525 struct ib_flow_spec_eth {
1526 	enum ib_flow_spec_type	  type;
1527 	u16			  size;
1528 	struct ib_flow_eth_filter val;
1529 	struct ib_flow_eth_filter mask;
1530 };
1531 
1532 struct ib_flow_ib_filter {
1533 	__be32	l3_type_qpn;
1534 	u8	dst_gid[16];
1535 };
1536 
1537 struct ib_flow_spec_ib {
1538 	enum ib_flow_spec_type	 type;
1539 	u16			 size;
1540 	struct ib_flow_ib_filter val;
1541 	struct ib_flow_ib_filter mask;
1542 };
1543 
1544 struct ib_flow_ipv4_filter {
1545 	__be32 src_ip;
1546 	__be32 dst_ip;
1547 };
1548 
1549 struct ib_flow_spec_ipv4 {
1550 	enum ib_flow_spec_type	   type;
1551 	u16			   size;
1552 	struct ib_flow_ipv4_filter val;
1553 	struct ib_flow_ipv4_filter mask;
1554 };
1555 
1556 struct ib_flow_tcp_udp_filter {
1557 	__be16 dst_port;
1558 	__be16	src_port;
1559 };
1560 
1561 struct ib_flow_spec_tcp_udp {
1562 	enum ib_flow_spec_type	      type;
1563 	u16			      size;
1564 	struct ib_flow_tcp_udp_filter val;
1565 	struct ib_flow_tcp_udp_filter mask;
1566 };
1567 
1568 union ib_flow_spec {
1569 	struct {
1570 		enum ib_flow_spec_type	type;
1571 		u16			size;
1572 	};
1573 	struct ib_flow_spec_ib ib;
1574 	struct ib_flow_spec_eth eth;
1575 	struct ib_flow_spec_ipv4 ipv4;
1576 	struct ib_flow_spec_tcp_udp tcp_udp;
1577 };
1578 
1579 struct ib_flow_attr {
1580 	enum ib_flow_attr_type type;
1581 	u16	     size;
1582 	u16	     priority;
1583 	u8	     num_of_specs;
1584 	u8	     port;
1585 	u32	     flags;
1586 	/* Following are the optional layers according to user request
1587 	 * struct ib_flow_spec_xxx
1588 	 * struct ib_flow_spec_yyy
1589 	 */
1590 };
1591 
1592 struct ib_flow {
1593 	struct ib_qp		*qp;
1594 	struct ib_uobject	*uobject;
1595 };
1596 
1597 struct ib_mad;
1598 struct ib_grh;
1599 
1600 enum ib_process_mad_flags {
1601 	IB_MAD_IGNORE_MKEY	= 1,
1602 	IB_MAD_IGNORE_BKEY	= 2,
1603 	IB_MAD_IGNORE_ALL	= IB_MAD_IGNORE_MKEY | IB_MAD_IGNORE_BKEY
1604 };
1605 
1606 enum ib_mad_result {
1607 	IB_MAD_RESULT_FAILURE  = 0,      /* (!SUCCESS is the important flag) */
1608 	IB_MAD_RESULT_SUCCESS  = 1 << 0, /* MAD was successfully processed   */
1609 	IB_MAD_RESULT_REPLY    = 1 << 1, /* Reply packet needs to be sent    */
1610 	IB_MAD_RESULT_CONSUMED = 1 << 2  /* Packet consumed: stop processing */
1611 };
1612 
1613 #define IB_DEVICE_NAME_MAX 64
1614 
1615 struct ib_cache {
1616 	rwlock_t                lock;
1617 	struct ib_event_handler event_handler;
1618 	struct ib_pkey_cache  **pkey_cache;
1619 	struct ib_gid_cache   **gid_cache;
1620 	u8                     *lmc_cache;
1621 };
1622 
1623 enum verbs_values_mask {
1624 	IBV_VALUES_HW_CLOCK = 1 << 0
1625 };
1626 
1627 struct ib_device_values {
1628 	int values_mask;
1629 	uint64_t hwclock;
1630 };
1631 
1632 struct ib_dma_mapping_ops {
1633 	int		(*mapping_error)(struct ib_device *dev,
1634 					 u64 dma_addr);
1635 	u64		(*map_single)(struct ib_device *dev,
1636 				      void *ptr, size_t size,
1637 				      enum dma_data_direction direction);
1638 	void		(*unmap_single)(struct ib_device *dev,
1639 					u64 addr, size_t size,
1640 					enum dma_data_direction direction);
1641 	u64		(*map_page)(struct ib_device *dev,
1642 				    struct page *page, unsigned long offset,
1643 				    size_t size,
1644 				    enum dma_data_direction direction);
1645 	void		(*unmap_page)(struct ib_device *dev,
1646 				      u64 addr, size_t size,
1647 				      enum dma_data_direction direction);
1648 	int		(*map_sg)(struct ib_device *dev,
1649 				  struct scatterlist *sg, int nents,
1650 				  enum dma_data_direction direction);
1651 	void		(*unmap_sg)(struct ib_device *dev,
1652 				    struct scatterlist *sg, int nents,
1653 				    enum dma_data_direction direction);
1654 	u64		(*dma_address)(struct ib_device *dev,
1655 				       struct scatterlist *sg);
1656 	unsigned int	(*dma_len)(struct ib_device *dev,
1657 				   struct scatterlist *sg);
1658 	void		(*sync_single_for_cpu)(struct ib_device *dev,
1659 					       u64 dma_handle,
1660 					       size_t size,
1661 					       enum dma_data_direction dir);
1662 	void		(*sync_single_for_device)(struct ib_device *dev,
1663 						  u64 dma_handle,
1664 						  size_t size,
1665 						  enum dma_data_direction dir);
1666 	void		*(*alloc_coherent)(struct ib_device *dev,
1667 					   size_t size,
1668 					   u64 *dma_handle,
1669 					   gfp_t flag);
1670 	void		(*free_coherent)(struct ib_device *dev,
1671 					 size_t size, void *cpu_addr,
1672 					 u64 dma_handle);
1673 };
1674 
1675 struct iw_cm_verbs;
1676 
1677 struct ib_port_immutable {
1678 	int                           pkey_tbl_len;
1679 	int                           gid_tbl_len;
1680 	u32                           core_cap_flags;
1681 	u32                           max_mad_size;
1682 };
1683 
1684 struct ib_exp_device_attr;
1685 struct ib_exp_qp_init_attr;
1686 
1687 struct ib_device {
1688 	struct device                *dma_device;
1689 
1690 	char                          name[IB_DEVICE_NAME_MAX];
1691 
1692 	struct list_head              event_handler_list;
1693 	spinlock_t                    event_handler_lock;
1694 
1695 	spinlock_t                    client_data_lock;
1696 	struct list_head              core_list;
1697 	struct list_head              client_data_list;
1698 
1699 	struct ib_cache               cache;
1700 	/**
1701 	 * port_immutable is indexed by port number
1702 	 */
1703 	struct ib_port_immutable     *port_immutable;
1704 
1705 	int			      num_comp_vectors;
1706 
1707 	struct iw_cm_verbs	     *iwcm;
1708 
1709 	int		           (*get_protocol_stats)(struct ib_device *device,
1710 							 union rdma_protocol_stats *stats);
1711 	int		           (*query_device)(struct ib_device *device,
1712 						   struct ib_device_attr *device_attr);
1713 	int		           (*query_port)(struct ib_device *device,
1714 						 u8 port_num,
1715 						 struct ib_port_attr *port_attr);
1716 	enum rdma_link_layer	   (*get_link_layer)(struct ib_device *device,
1717 						     u8 port_num);
1718 	/* When calling get_netdev, the HW vendor's driver should return the
1719 	 * net device of device @device at port @port_num. The function
1720 	 * is called in rtnl_lock. The HW vendor's device driver must guarantee
1721 	 * to return NULL before the net device has reached
1722 	 * NETDEV_UNREGISTER_FINAL state.
1723 	 */
1724 	struct net_device	*(*get_netdev)(struct ib_device *device,
1725 					       u8 port_num);
1726 	int		           (*query_gid)(struct ib_device *device,
1727 						u8 port_num, int index,
1728 						union ib_gid *gid);
1729 	int		           (*query_pkey)(struct ib_device *device,
1730 						 u8 port_num, u16 index, u16 *pkey);
1731 	int		           (*modify_device)(struct ib_device *device,
1732 						    int device_modify_mask,
1733 						    struct ib_device_modify *device_modify);
1734 	int		           (*modify_port)(struct ib_device *device,
1735 						  u8 port_num, int port_modify_mask,
1736 						  struct ib_port_modify *port_modify);
1737 	struct ib_ucontext *       (*alloc_ucontext)(struct ib_device *device,
1738 						     struct ib_udata *udata);
1739 	int                        (*dealloc_ucontext)(struct ib_ucontext *context);
1740 	int                        (*mmap)(struct ib_ucontext *context,
1741 					   struct vm_area_struct *vma);
1742 	struct ib_pd *             (*alloc_pd)(struct ib_device *device,
1743 					       struct ib_ucontext *context,
1744 					       struct ib_udata *udata);
1745 	int                        (*dealloc_pd)(struct ib_pd *pd);
1746 	struct ib_ah *             (*create_ah)(struct ib_pd *pd,
1747 						struct ib_ah_attr *ah_attr);
1748 	int                        (*modify_ah)(struct ib_ah *ah,
1749 						struct ib_ah_attr *ah_attr);
1750 	int                        (*query_ah)(struct ib_ah *ah,
1751 					       struct ib_ah_attr *ah_attr);
1752 	int                        (*destroy_ah)(struct ib_ah *ah);
1753 	struct ib_srq *            (*create_srq)(struct ib_pd *pd,
1754 						 struct ib_srq_init_attr *srq_init_attr,
1755 						 struct ib_udata *udata);
1756 	int                        (*modify_srq)(struct ib_srq *srq,
1757 						 struct ib_srq_attr *srq_attr,
1758 						 enum ib_srq_attr_mask srq_attr_mask,
1759 						 struct ib_udata *udata);
1760 	int                        (*query_srq)(struct ib_srq *srq,
1761 						struct ib_srq_attr *srq_attr);
1762 	int                        (*destroy_srq)(struct ib_srq *srq);
1763 	int                        (*post_srq_recv)(struct ib_srq *srq,
1764 						    struct ib_recv_wr *recv_wr,
1765 						    struct ib_recv_wr **bad_recv_wr);
1766 	struct ib_qp *             (*create_qp)(struct ib_pd *pd,
1767 						struct ib_qp_init_attr *qp_init_attr,
1768 						struct ib_udata *udata);
1769 	int                        (*modify_qp)(struct ib_qp *qp,
1770 						struct ib_qp_attr *qp_attr,
1771 						int qp_attr_mask,
1772 						struct ib_udata *udata);
1773 	int                        (*query_qp)(struct ib_qp *qp,
1774 					       struct ib_qp_attr *qp_attr,
1775 					       int qp_attr_mask,
1776 					       struct ib_qp_init_attr *qp_init_attr);
1777 	int                        (*destroy_qp)(struct ib_qp *qp);
1778 	int                        (*post_send)(struct ib_qp *qp,
1779 						struct ib_send_wr *send_wr,
1780 						struct ib_send_wr **bad_send_wr);
1781 	int                        (*post_recv)(struct ib_qp *qp,
1782 						struct ib_recv_wr *recv_wr,
1783 						struct ib_recv_wr **bad_recv_wr);
1784 	struct ib_cq *             (*create_cq)(struct ib_device *device,
1785 						struct ib_cq_init_attr *attr,
1786 						struct ib_ucontext *context,
1787 						struct ib_udata *udata);
1788 	int                        (*modify_cq)(struct ib_cq *cq,
1789 						struct ib_cq_attr *cq_attr,
1790 						int cq_attr_mask);
1791 	int                        (*destroy_cq)(struct ib_cq *cq);
1792 	int                        (*resize_cq)(struct ib_cq *cq, int cqe,
1793 						struct ib_udata *udata);
1794 	int                        (*poll_cq)(struct ib_cq *cq, int num_entries,
1795 					      struct ib_wc *wc);
1796 	int                        (*peek_cq)(struct ib_cq *cq, int wc_cnt);
1797 	int                        (*req_notify_cq)(struct ib_cq *cq,
1798 						    enum ib_cq_notify_flags flags);
1799 	int                        (*req_ncomp_notif)(struct ib_cq *cq,
1800 						      int wc_cnt);
1801 	struct ib_mr *             (*get_dma_mr)(struct ib_pd *pd,
1802 						 int mr_access_flags);
1803 	struct ib_mr *             (*reg_phys_mr)(struct ib_pd *pd,
1804 						  struct ib_phys_buf *phys_buf_array,
1805 						  int num_phys_buf,
1806 						  int mr_access_flags,
1807 						  u64 *iova_start);
1808 	struct ib_mr *             (*reg_user_mr)(struct ib_pd *pd,
1809 						  u64 start, u64 length,
1810 						  u64 virt_addr,
1811 						  int mr_access_flags,
1812 						  struct ib_udata *udata,
1813 							int mr_id);
1814 	int                        (*query_mr)(struct ib_mr *mr,
1815 					       struct ib_mr_attr *mr_attr);
1816 	int                        (*dereg_mr)(struct ib_mr *mr);
1817 	int                        (*destroy_mr)(struct ib_mr *mr);
1818 	struct ib_mr *		   (*create_mr)(struct ib_pd *pd,
1819 						struct ib_mr_init_attr *mr_init_attr);
1820 	struct ib_mr *		   (*alloc_fast_reg_mr)(struct ib_pd *pd,
1821 					       int max_page_list_len);
1822 	struct ib_fast_reg_page_list * (*alloc_fast_reg_page_list)(struct ib_device *device,
1823 								   int page_list_len);
1824 	void			   (*free_fast_reg_page_list)(struct ib_fast_reg_page_list *page_list);
1825 	int                        (*rereg_phys_mr)(struct ib_mr *mr,
1826 						    int mr_rereg_mask,
1827 						    struct ib_pd *pd,
1828 						    struct ib_phys_buf *phys_buf_array,
1829 						    int num_phys_buf,
1830 						    int mr_access_flags,
1831 						    u64 *iova_start);
1832 	struct ib_mw *             (*alloc_mw)(struct ib_pd *pd,
1833 					       enum ib_mw_type type);
1834 	int                        (*bind_mw)(struct ib_qp *qp,
1835 					      struct ib_mw *mw,
1836 					      struct ib_mw_bind *mw_bind);
1837 	int                        (*dealloc_mw)(struct ib_mw *mw);
1838 	struct ib_fmr *	           (*alloc_fmr)(struct ib_pd *pd,
1839 						int mr_access_flags,
1840 						struct ib_fmr_attr *fmr_attr);
1841 	int		           (*map_phys_fmr)(struct ib_fmr *fmr,
1842 						   u64 *page_list, int list_len,
1843 						   u64 iova);
1844 	int		           (*unmap_fmr)(struct list_head *fmr_list);
1845 	int		           (*dealloc_fmr)(struct ib_fmr *fmr);
1846 	int                        (*attach_mcast)(struct ib_qp *qp,
1847 						   union ib_gid *gid,
1848 						   u16 lid);
1849 	int                        (*detach_mcast)(struct ib_qp *qp,
1850 						   union ib_gid *gid,
1851 						   u16 lid);
1852 	int                        (*process_mad)(struct ib_device *device,
1853 						  int process_mad_flags,
1854 						  u8 port_num,
1855 						  struct ib_wc *in_wc,
1856 						  struct ib_grh *in_grh,
1857 						  struct ib_mad *in_mad,
1858 						  struct ib_mad *out_mad);
1859 	struct ib_xrcd *	   (*alloc_xrcd)(struct ib_device *device,
1860 						 struct ib_ucontext *ucontext,
1861 						 struct ib_udata *udata);
1862 	int			   (*dealloc_xrcd)(struct ib_xrcd *xrcd);
1863 	struct ib_flow *	   (*create_flow)(struct ib_qp *qp,
1864 						  struct ib_flow_attr
1865 						  *flow_attr,
1866 						  int domain);
1867 	int			   (*destroy_flow)(struct ib_flow *flow_id);
1868 	int			   (*check_mr_status)(struct ib_mr *mr, u32 check_mask,
1869 						      struct ib_mr_status *mr_status);
1870 
1871 	unsigned long		   (*get_unmapped_area)(struct file *file,
1872 					unsigned long addr,
1873 					unsigned long len, unsigned long pgoff,
1874 					unsigned long flags);
1875 	int			   (*ioctl)(struct ib_ucontext *context,
1876 					    unsigned int cmd,
1877 					    unsigned long arg);
1878 	int			   (*query_values)(struct ib_device *device,
1879 					    int q_values,
1880 					    struct ib_device_values *values);
1881 	struct ib_dma_mapping_ops   *dma_ops;
1882 
1883 	struct module               *owner;
1884 	struct device                dev;
1885 	struct kobject               *ports_parent;
1886 	struct list_head             port_list;
1887 
1888 	enum {
1889 		IB_DEV_UNINITIALIZED,
1890 		IB_DEV_REGISTERED,
1891 		IB_DEV_UNREGISTERED
1892 	}                            reg_state;
1893 
1894 	int			     uverbs_abi_ver;
1895 	u64			     uverbs_cmd_mask;
1896 	u64			     uverbs_ex_cmd_mask;
1897 
1898 	char			     node_desc[64];
1899 	__be64			     node_guid;
1900 	u32			     local_dma_lkey;
1901 	u8                           node_type;
1902 	u8                           phys_port_cnt;
1903 	int			     cmd_perf;
1904 	u64			     cmd_avg;
1905 	u32			     cmd_n;
1906 	spinlock_t		     cmd_perf_lock;
1907 
1908 
1909 	/**
1910 	 * The following mandatory functions are used only at device
1911 	 * registration.  Keep functions such as these at the end of this
1912 	 * structure to avoid cache line misses when accessing struct ib_device
1913 	 * in fast paths.
1914 	 */
1915 	int (*get_port_immutable)(struct ib_device *, u8, struct ib_port_immutable *);
1916 
1917 	/*
1918 	 * Experimental data and functions
1919 	 */
1920 	int			(*exp_query_device)(struct ib_device *device,
1921 						    struct ib_exp_device_attr *device_attr);
1922 	struct ib_qp *		(*exp_create_qp)(struct ib_pd *pd,
1923 						 struct ib_exp_qp_init_attr *qp_init_attr,
1924 						 struct ib_udata *udata);
1925 	struct ib_dct *		(*exp_create_dct)(struct ib_pd *pd,
1926 					      struct ib_dct_init_attr *attr,
1927 					      struct ib_udata *udata);
1928 	int			(*exp_destroy_dct)(struct ib_dct *dct);
1929 	int			(*exp_query_dct)(struct ib_dct *dct, struct ib_dct_attr *attr);
1930 
1931 	u64			uverbs_exp_cmd_mask;
1932 };
1933 
1934 struct ib_client {
1935 	char  *name;
1936 	void (*add)   (struct ib_device *);
1937 	void (*remove)(struct ib_device *);
1938 
1939 	struct list_head list;
1940 };
1941 
1942 struct ib_device *ib_alloc_device(size_t size);
1943 void ib_dealloc_device(struct ib_device *device);
1944 
1945 int ib_register_device(struct ib_device *device,
1946 		       int (*port_callback)(struct ib_device *,
1947 					    u8, struct kobject *));
1948 void ib_unregister_device(struct ib_device *device);
1949 
1950 int ib_register_client   (struct ib_client *client);
1951 void ib_unregister_client(struct ib_client *client);
1952 
1953 void *ib_get_client_data(struct ib_device *device, struct ib_client *client);
1954 void  ib_set_client_data(struct ib_device *device, struct ib_client *client,
1955 			 void *data);
1956 
1957 static inline int ib_copy_from_udata(void *dest, struct ib_udata *udata, size_t len)
1958 {
1959 	return udata->ops->copy_from(dest, udata, len);
1960 }
1961 
1962 static inline int ib_copy_to_udata(struct ib_udata *udata, void *src, size_t len)
1963 {
1964 	return udata->ops->copy_to(udata, src, len);
1965 }
1966 
1967 /**
1968  * ib_modify_qp_is_ok - Check that the supplied attribute mask
1969  * contains all required attributes and no attributes not allowed for
1970  * the given QP state transition.
1971  * @cur_state: Current QP state
1972  * @next_state: Next QP state
1973  * @type: QP type
1974  * @mask: Mask of supplied QP attributes
1975  * @ll : link layer of port
1976  *
1977  * This function is a helper function that a low-level driver's
1978  * modify_qp method can use to validate the consumer's input.  It
1979  * checks that cur_state and next_state are valid QP states, that a
1980  * transition from cur_state to next_state is allowed by the IB spec,
1981  * and that the attribute mask supplied is allowed for the transition.
1982  */
1983 int ib_modify_qp_is_ok(enum ib_qp_state cur_state, enum ib_qp_state next_state,
1984 		       enum ib_qp_type type, enum ib_qp_attr_mask mask,
1985 		       enum rdma_link_layer ll);
1986 
1987 int ib_register_event_handler  (struct ib_event_handler *event_handler);
1988 int ib_unregister_event_handler(struct ib_event_handler *event_handler);
1989 void ib_dispatch_event(struct ib_event *event);
1990 
1991 int ib_query_device(struct ib_device *device,
1992 		    struct ib_device_attr *device_attr);
1993 
1994 int ib_query_port(struct ib_device *device,
1995 		  u8 port_num, struct ib_port_attr *port_attr);
1996 
1997 enum rdma_link_layer rdma_port_get_link_layer(struct ib_device *device,
1998 					       u8 port_num);
1999 
2000 static inline bool rdma_protocol_ib(const struct ib_device *device, u8 port_num)
2001 {
2002 	return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_IB;
2003 }
2004 
2005 static inline bool rdma_protocol_roce(const struct ib_device *device, u8 port_num)
2006 {
2007 	return device->port_immutable[port_num].core_cap_flags &
2008 		(RDMA_CORE_CAP_PROT_ROCE | RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP);
2009 }
2010 
2011 static inline bool rdma_protocol_roce_udp_encap(const struct ib_device *device, u8 port_num)
2012 {
2013 	return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP;
2014 }
2015 
2016 static inline bool rdma_protocol_roce_eth_encap(const struct ib_device *device, u8 port_num)
2017 {
2018 	return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_ROCE;
2019 }
2020 
2021 static inline bool rdma_protocol_iwarp(const struct ib_device *device, u8 port_num)
2022 {
2023 	return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_IWARP;
2024 }
2025 
2026 static inline bool rdma_ib_or_roce(const struct ib_device *device, u8 port_num)
2027 {
2028 	return rdma_protocol_ib(device, port_num) ||
2029 		rdma_protocol_roce(device, port_num);
2030 }
2031 
2032 /**
2033  * rdma_cap_ib_mad - Check if the port of a device supports Infiniband
2034  * Management Datagrams.
2035  * @device: Device to check
2036  * @port_num: Port number to check
2037  *
2038  * Management Datagrams (MAD) are a required part of the InfiniBand
2039  * specification and are supported on all InfiniBand devices.  A slightly
2040  * extended version are also supported on OPA interfaces.
2041  *
2042  * Return: true if the port supports sending/receiving of MAD packets.
2043  */
2044 static inline bool rdma_cap_ib_mad(const struct ib_device *device, u8 port_num)
2045 {
2046 	return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IB_MAD;
2047 }
2048 
2049 /**
2050  * rdma_cap_opa_mad - Check if the port of device provides support for OPA
2051  * Management Datagrams.
2052  * @device: Device to check
2053  * @port_num: Port number to check
2054  *
2055  * Intel OmniPath devices extend and/or replace the InfiniBand Management
2056  * datagrams with their own versions.  These OPA MADs share many but not all of
2057  * the characteristics of InfiniBand MADs.
2058  *
2059  * OPA MADs differ in the following ways:
2060  *
2061  *    1) MADs are variable size up to 2K
2062  *       IBTA defined MADs remain fixed at 256 bytes
2063  *    2) OPA SMPs must carry valid PKeys
2064  *    3) OPA SMP packets are a different format
2065  *
2066  * Return: true if the port supports OPA MAD packet formats.
2067  */
2068 static inline bool rdma_cap_opa_mad(struct ib_device *device, u8 port_num)
2069 {
2070 	return (device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_OPA_MAD)
2071 		== RDMA_CORE_CAP_OPA_MAD;
2072 }
2073 
2074 /**
2075  * rdma_cap_ib_smi - Check if the port of a device provides an Infiniband
2076  * Subnet Management Agent (SMA) on the Subnet Management Interface (SMI).
2077  * @device: Device to check
2078  * @port_num: Port number to check
2079  *
2080  * Each InfiniBand node is required to provide a Subnet Management Agent
2081  * that the subnet manager can access.  Prior to the fabric being fully
2082  * configured by the subnet manager, the SMA is accessed via a well known
2083  * interface called the Subnet Management Interface (SMI).  This interface
2084  * uses directed route packets to communicate with the SM to get around the
2085  * chicken and egg problem of the SM needing to know what's on the fabric
2086  * in order to configure the fabric, and needing to configure the fabric in
2087  * order to send packets to the devices on the fabric.  These directed
2088  * route packets do not need the fabric fully configured in order to reach
2089  * their destination.  The SMI is the only method allowed to send
2090  * directed route packets on an InfiniBand fabric.
2091  *
2092  * Return: true if the port provides an SMI.
2093  */
2094 static inline bool rdma_cap_ib_smi(const struct ib_device *device, u8 port_num)
2095 {
2096 	return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IB_SMI;
2097 }
2098 
2099 /**
2100  * rdma_cap_ib_cm - Check if the port of device has the capability Infiniband
2101  * Communication Manager.
2102  * @device: Device to check
2103  * @port_num: Port number to check
2104  *
2105  * The InfiniBand Communication Manager is one of many pre-defined General
2106  * Service Agents (GSA) that are accessed via the General Service
2107  * Interface (GSI).  It's role is to facilitate establishment of connections
2108  * between nodes as well as other management related tasks for established
2109  * connections.
2110  *
2111  * Return: true if the port supports an IB CM (this does not guarantee that
2112  * a CM is actually running however).
2113  */
2114 static inline bool rdma_cap_ib_cm(const struct ib_device *device, u8 port_num)
2115 {
2116 	return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IB_CM;
2117 }
2118 
2119 /**
2120  * rdma_cap_iw_cm - Check if the port of device has the capability IWARP
2121  * Communication Manager.
2122  * @device: Device to check
2123  * @port_num: Port number to check
2124  *
2125  * Similar to above, but specific to iWARP connections which have a different
2126  * managment protocol than InfiniBand.
2127  *
2128  * Return: true if the port supports an iWARP CM (this does not guarantee that
2129  * a CM is actually running however).
2130  */
2131 static inline bool rdma_cap_iw_cm(const struct ib_device *device, u8 port_num)
2132 {
2133 	return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IW_CM;
2134 }
2135 
2136 /**
2137  * rdma_cap_ib_sa - Check if the port of device has the capability Infiniband
2138  * Subnet Administration.
2139  * @device: Device to check
2140  * @port_num: Port number to check
2141  *
2142  * An InfiniBand Subnet Administration (SA) service is a pre-defined General
2143  * Service Agent (GSA) provided by the Subnet Manager (SM).  On InfiniBand
2144  * fabrics, devices should resolve routes to other hosts by contacting the
2145  * SA to query the proper route.
2146  *
2147  * Return: true if the port should act as a client to the fabric Subnet
2148  * Administration interface.  This does not imply that the SA service is
2149  * running locally.
2150  */
2151 static inline bool rdma_cap_ib_sa(const struct ib_device *device, u8 port_num)
2152 {
2153 	return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IB_SA;
2154 }
2155 
2156 /**
2157  * rdma_cap_ib_mcast - Check if the port of device has the capability Infiniband
2158  * Multicast.
2159  * @device: Device to check
2160  * @port_num: Port number to check
2161  *
2162  * InfiniBand multicast registration is more complex than normal IPv4 or
2163  * IPv6 multicast registration.  Each Host Channel Adapter must register
2164  * with the Subnet Manager when it wishes to join a multicast group.  It
2165  * should do so only once regardless of how many queue pairs it subscribes
2166  * to this group.  And it should leave the group only after all queue pairs
2167  * attached to the group have been detached.
2168  *
2169  * Return: true if the port must undertake the additional adminstrative
2170  * overhead of registering/unregistering with the SM and tracking of the
2171  * total number of queue pairs attached to the multicast group.
2172  */
2173 static inline bool rdma_cap_ib_mcast(const struct ib_device *device, u8 port_num)
2174 {
2175 	return rdma_cap_ib_sa(device, port_num);
2176 }
2177 
2178 /**
2179  * rdma_cap_af_ib - Check if the port of device has the capability
2180  * Native Infiniband Address.
2181  * @device: Device to check
2182  * @port_num: Port number to check
2183  *
2184  * InfiniBand addressing uses a port's GUID + Subnet Prefix to make a default
2185  * GID.  RoCE uses a different mechanism, but still generates a GID via
2186  * a prescribed mechanism and port specific data.
2187  *
2188  * Return: true if the port uses a GID address to identify devices on the
2189  * network.
2190  */
2191 static inline bool rdma_cap_af_ib(const struct ib_device *device, u8 port_num)
2192 {
2193 	return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_AF_IB;
2194 }
2195 
2196 /**
2197  * rdma_cap_eth_ah - Check if the port of device has the capability
2198  * Ethernet Address Handle.
2199  * @device: Device to check
2200  * @port_num: Port number to check
2201  *
2202  * RoCE is InfiniBand over Ethernet, and it uses a well defined technique
2203  * to fabricate GIDs over Ethernet/IP specific addresses native to the
2204  * port.  Normally, packet headers are generated by the sending host
2205  * adapter, but when sending connectionless datagrams, we must manually
2206  * inject the proper headers for the fabric we are communicating over.
2207  *
2208  * Return: true if we are running as a RoCE port and must force the
2209  * addition of a Global Route Header built from our Ethernet Address
2210  * Handle into our header list for connectionless packets.
2211  */
2212 static inline bool rdma_cap_eth_ah(const struct ib_device *device, u8 port_num)
2213 {
2214 	return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_ETH_AH;
2215 }
2216 
2217 /**
2218  * rdma_max_mad_size - Return the max MAD size required by this RDMA Port.
2219  *
2220  * @device: Device
2221  * @port_num: Port number
2222  *
2223  * This MAD size includes the MAD headers and MAD payload.  No other headers
2224  * are included.
2225  *
2226  * Return the max MAD size required by the Port.  Will return 0 if the port
2227  * does not support MADs
2228  */
2229 static inline size_t rdma_max_mad_size(const struct ib_device *device, u8 port_num)
2230 {
2231 	return device->port_immutable[port_num].max_mad_size;
2232 }
2233 
2234 /*
2235  * Check if the device supports READ W/ INVALIDATE.
2236  */
2237 static inline bool rdma_cap_read_inv(struct ib_device *dev, u32 port_num)
2238 {
2239 	/*
2240 	 * iWarp drivers must support READ W/ INVALIDATE.  No other protocol
2241 	 * has support for it yet.
2242 	 */
2243 	return rdma_protocol_iwarp(dev, port_num);
2244 }
2245 
2246 int ib_query_gid(struct ib_device *device,
2247 		 u8 port_num, int index, union ib_gid *gid);
2248 
2249 int ib_query_pkey(struct ib_device *device,
2250 		  u8 port_num, u16 index, u16 *pkey);
2251 
2252 int ib_modify_device(struct ib_device *device,
2253 		     int device_modify_mask,
2254 		     struct ib_device_modify *device_modify);
2255 
2256 int ib_modify_port(struct ib_device *device,
2257 		   u8 port_num, int port_modify_mask,
2258 		   struct ib_port_modify *port_modify);
2259 
2260 int ib_find_gid(struct ib_device *device, union ib_gid *gid,
2261 		u8 *port_num, u16 *index);
2262 
2263 int ib_find_pkey(struct ib_device *device,
2264 		 u8 port_num, u16 pkey, u16 *index);
2265 
2266 /**
2267  * ib_alloc_pd - Allocates an unused protection domain.
2268  * @device: The device on which to allocate the protection domain.
2269  *
2270  * A protection domain object provides an association between QPs, shared
2271  * receive queues, address handles, memory regions, and memory windows.
2272  */
2273 struct ib_pd *ib_alloc_pd(struct ib_device *device);
2274 
2275 /**
2276  * ib_dealloc_pd - Deallocates a protection domain.
2277  * @pd: The protection domain to deallocate.
2278  */
2279 int ib_dealloc_pd(struct ib_pd *pd);
2280 
2281 /**
2282  * ib_create_ah - Creates an address handle for the given address vector.
2283  * @pd: The protection domain associated with the address handle.
2284  * @ah_attr: The attributes of the address vector.
2285  *
2286  * The address handle is used to reference a local or global destination
2287  * in all UD QP post sends.
2288  */
2289 struct ib_ah *ib_create_ah(struct ib_pd *pd, struct ib_ah_attr *ah_attr);
2290 
2291 /**
2292  * ib_init_ah_from_wc - Initializes address handle attributes from a
2293  *   work completion.
2294  * @device: Device on which the received message arrived.
2295  * @port_num: Port on which the received message arrived.
2296  * @wc: Work completion associated with the received message.
2297  * @grh: References the received global route header.  This parameter is
2298  *   ignored unless the work completion indicates that the GRH is valid.
2299  * @ah_attr: Returned attributes that can be used when creating an address
2300  *   handle for replying to the message.
2301  */
2302 int ib_init_ah_from_wc(struct ib_device *device, u8 port_num, struct ib_wc *wc,
2303 		       struct ib_grh *grh, struct ib_ah_attr *ah_attr);
2304 
2305 /**
2306  * ib_create_ah_from_wc - Creates an address handle associated with the
2307  *   sender of the specified work completion.
2308  * @pd: The protection domain associated with the address handle.
2309  * @wc: Work completion information associated with a received message.
2310  * @grh: References the received global route header.  This parameter is
2311  *   ignored unless the work completion indicates that the GRH is valid.
2312  * @port_num: The outbound port number to associate with the address.
2313  *
2314  * The address handle is used to reference a local or global destination
2315  * in all UD QP post sends.
2316  */
2317 struct ib_ah *ib_create_ah_from_wc(struct ib_pd *pd, struct ib_wc *wc,
2318 				   struct ib_grh *grh, u8 port_num);
2319 
2320 /**
2321  * ib_modify_ah - Modifies the address vector associated with an address
2322  *   handle.
2323  * @ah: The address handle to modify.
2324  * @ah_attr: The new address vector attributes to associate with the
2325  *   address handle.
2326  */
2327 int ib_modify_ah(struct ib_ah *ah, struct ib_ah_attr *ah_attr);
2328 
2329 /**
2330  * ib_query_ah - Queries the address vector associated with an address
2331  *   handle.
2332  * @ah: The address handle to query.
2333  * @ah_attr: The address vector attributes associated with the address
2334  *   handle.
2335  */
2336 int ib_query_ah(struct ib_ah *ah, struct ib_ah_attr *ah_attr);
2337 
2338 /**
2339  * ib_destroy_ah - Destroys an address handle.
2340  * @ah: The address handle to destroy.
2341  */
2342 int ib_destroy_ah(struct ib_ah *ah);
2343 
2344 /**
2345  * ib_create_srq - Creates a SRQ associated with the specified protection
2346  *   domain.
2347  * @pd: The protection domain associated with the SRQ.
2348  * @srq_init_attr: A list of initial attributes required to create the
2349  *   SRQ.  If SRQ creation succeeds, then the attributes are updated to
2350  *   the actual capabilities of the created SRQ.
2351  *
2352  * srq_attr->max_wr and srq_attr->max_sge are read the determine the
2353  * requested size of the SRQ, and set to the actual values allocated
2354  * on return.  If ib_create_srq() succeeds, then max_wr and max_sge
2355  * will always be at least as large as the requested values.
2356  */
2357 struct ib_srq *ib_create_srq(struct ib_pd *pd,
2358 			     struct ib_srq_init_attr *srq_init_attr);
2359 
2360 /**
2361  * ib_modify_srq - Modifies the attributes for the specified SRQ.
2362  * @srq: The SRQ to modify.
2363  * @srq_attr: On input, specifies the SRQ attributes to modify.  On output,
2364  *   the current values of selected SRQ attributes are returned.
2365  * @srq_attr_mask: A bit-mask used to specify which attributes of the SRQ
2366  *   are being modified.
2367  *
2368  * The mask may contain IB_SRQ_MAX_WR to resize the SRQ and/or
2369  * IB_SRQ_LIMIT to set the SRQ's limit and request notification when
2370  * the number of receives queued drops below the limit.
2371  */
2372 int ib_modify_srq(struct ib_srq *srq,
2373 		  struct ib_srq_attr *srq_attr,
2374 		  enum ib_srq_attr_mask srq_attr_mask);
2375 
2376 /**
2377  * ib_query_srq - Returns the attribute list and current values for the
2378  *   specified SRQ.
2379  * @srq: The SRQ to query.
2380  * @srq_attr: The attributes of the specified SRQ.
2381  */
2382 int ib_query_srq(struct ib_srq *srq,
2383 		 struct ib_srq_attr *srq_attr);
2384 
2385 /**
2386  * ib_destroy_srq - Destroys the specified SRQ.
2387  * @srq: The SRQ to destroy.
2388  */
2389 int ib_destroy_srq(struct ib_srq *srq);
2390 
2391 /**
2392  * ib_post_srq_recv - Posts a list of work requests to the specified SRQ.
2393  * @srq: The SRQ to post the work request on.
2394  * @recv_wr: A list of work requests to post on the receive queue.
2395  * @bad_recv_wr: On an immediate failure, this parameter will reference
2396  *   the work request that failed to be posted on the QP.
2397  */
2398 static inline int ib_post_srq_recv(struct ib_srq *srq,
2399 				   struct ib_recv_wr *recv_wr,
2400 				   struct ib_recv_wr **bad_recv_wr)
2401 {
2402 	return srq->device->post_srq_recv(srq, recv_wr, bad_recv_wr);
2403 }
2404 
2405 /**
2406  * ib_create_qp - Creates a QP associated with the specified protection
2407  *   domain.
2408  * @pd: The protection domain associated with the QP.
2409  * @qp_init_attr: A list of initial attributes required to create the
2410  *   QP.  If QP creation succeeds, then the attributes are updated to
2411  *   the actual capabilities of the created QP.
2412  */
2413 struct ib_qp *ib_create_qp(struct ib_pd *pd,
2414 			   struct ib_qp_init_attr *qp_init_attr);
2415 
2416 /**
2417  * ib_modify_qp - Modifies the attributes for the specified QP and then
2418  *   transitions the QP to the given state.
2419  * @qp: The QP to modify.
2420  * @qp_attr: On input, specifies the QP attributes to modify.  On output,
2421  *   the current values of selected QP attributes are returned.
2422  * @qp_attr_mask: A bit-mask used to specify which attributes of the QP
2423  *   are being modified.
2424  */
2425 int ib_modify_qp(struct ib_qp *qp,
2426 		 struct ib_qp_attr *qp_attr,
2427 		 int qp_attr_mask);
2428 
2429 /**
2430  * ib_query_qp - Returns the attribute list and current values for the
2431  *   specified QP.
2432  * @qp: The QP to query.
2433  * @qp_attr: The attributes of the specified QP.
2434  * @qp_attr_mask: A bit-mask used to select specific attributes to query.
2435  * @qp_init_attr: Additional attributes of the selected QP.
2436  *
2437  * The qp_attr_mask may be used to limit the query to gathering only the
2438  * selected attributes.
2439  */
2440 int ib_query_qp(struct ib_qp *qp,
2441 		struct ib_qp_attr *qp_attr,
2442 		int qp_attr_mask,
2443 		struct ib_qp_init_attr *qp_init_attr);
2444 
2445 /**
2446  * ib_destroy_qp - Destroys the specified QP.
2447  * @qp: The QP to destroy.
2448  */
2449 int ib_destroy_qp(struct ib_qp *qp);
2450 
2451 /**
2452  * ib_open_qp - Obtain a reference to an existing sharable QP.
2453  * @xrcd - XRC domain
2454  * @qp_open_attr: Attributes identifying the QP to open.
2455  *
2456  * Returns a reference to a sharable QP.
2457  */
2458 struct ib_qp *ib_open_qp(struct ib_xrcd *xrcd,
2459 			 struct ib_qp_open_attr *qp_open_attr);
2460 
2461 /**
2462  * ib_close_qp - Release an external reference to a QP.
2463  * @qp: The QP handle to release
2464  *
2465  * The opened QP handle is released by the caller.  The underlying
2466  * shared QP is not destroyed until all internal references are released.
2467  */
2468 int ib_close_qp(struct ib_qp *qp);
2469 
2470 /**
2471  * ib_post_send - Posts a list of work requests to the send queue of
2472  *   the specified QP.
2473  * @qp: The QP to post the work request on.
2474  * @send_wr: A list of work requests to post on the send queue.
2475  * @bad_send_wr: On an immediate failure, this parameter will reference
2476  *   the work request that failed to be posted on the QP.
2477  *
2478  * While IBA Vol. 1 section 11.4.1.1 specifies that if an immediate
2479  * error is returned, the QP state shall not be affected,
2480  * ib_post_send() will return an immediate error after queueing any
2481  * earlier work requests in the list.
2482  */
2483 static inline int ib_post_send(struct ib_qp *qp,
2484 			       struct ib_send_wr *send_wr,
2485 			       struct ib_send_wr **bad_send_wr)
2486 {
2487 	return qp->device->post_send(qp, send_wr, bad_send_wr);
2488 }
2489 
2490 /**
2491  * ib_post_recv - Posts a list of work requests to the receive queue of
2492  *   the specified QP.
2493  * @qp: The QP to post the work request on.
2494  * @recv_wr: A list of work requests to post on the receive queue.
2495  * @bad_recv_wr: On an immediate failure, this parameter will reference
2496  *   the work request that failed to be posted on the QP.
2497  */
2498 static inline int ib_post_recv(struct ib_qp *qp,
2499 			       struct ib_recv_wr *recv_wr,
2500 			       struct ib_recv_wr **bad_recv_wr)
2501 {
2502 	return qp->device->post_recv(qp, recv_wr, bad_recv_wr);
2503 }
2504 
2505 /**
2506  * ib_create_cq - Creates a CQ on the specified device.
2507  * @device: The device on which to create the CQ.
2508  * @comp_handler: A user-specified callback that is invoked when a
2509  *   completion event occurs on the CQ.
2510  * @event_handler: A user-specified callback that is invoked when an
2511  *   asynchronous event not associated with a completion occurs on the CQ.
2512  * @cq_context: Context associated with the CQ returned to the user via
2513  *   the associated completion and event handlers.
2514  * @cqe: The minimum size of the CQ.
2515  * @comp_vector - Completion vector used to signal completion events.
2516  *     Must be >= 0 and < context->num_comp_vectors.
2517  *
2518  * Users can examine the cq structure to determine the actual CQ size.
2519  */
2520 struct ib_cq *ib_create_cq(struct ib_device *device,
2521 			   ib_comp_handler comp_handler,
2522 			   void (*event_handler)(struct ib_event *, void *),
2523 			   void *cq_context, int cqe, int comp_vector);
2524 
2525 /**
2526  * ib_resize_cq - Modifies the capacity of the CQ.
2527  * @cq: The CQ to resize.
2528  * @cqe: The minimum size of the CQ.
2529  *
2530  * Users can examine the cq structure to determine the actual CQ size.
2531  */
2532 int ib_resize_cq(struct ib_cq *cq, int cqe);
2533 
2534 /**
2535  * ib_modify_cq - Modifies the attributes for the specified CQ and then
2536  *   transitions the CQ to the given state.
2537  * @cq: The CQ to modify.
2538  * @cq_attr: specifies the CQ attributes to modify.
2539  * @cq_attr_mask: A bit-mask used to specify which attributes of the CQ
2540  *   are being modified.
2541  */
2542 int ib_modify_cq(struct ib_cq *cq,
2543 		 struct ib_cq_attr *cq_attr,
2544 		 int cq_attr_mask);
2545 
2546 /**
2547  * ib_destroy_cq - Destroys the specified CQ.
2548  * @cq: The CQ to destroy.
2549  */
2550 int ib_destroy_cq(struct ib_cq *cq);
2551 
2552 /**
2553  * ib_poll_cq - poll a CQ for completion(s)
2554  * @cq:the CQ being polled
2555  * @num_entries:maximum number of completions to return
2556  * @wc:array of at least @num_entries &struct ib_wc where completions
2557  *   will be returned
2558  *
2559  * Poll a CQ for (possibly multiple) completions.  If the return value
2560  * is < 0, an error occurred.  If the return value is >= 0, it is the
2561  * number of completions returned.  If the return value is
2562  * non-negative and < num_entries, then the CQ was emptied.
2563  */
2564 static inline int ib_poll_cq(struct ib_cq *cq, int num_entries,
2565 			     struct ib_wc *wc)
2566 {
2567 	return cq->device->poll_cq(cq, num_entries, wc);
2568 }
2569 
2570 /**
2571  * ib_peek_cq - Returns the number of unreaped completions currently
2572  *   on the specified CQ.
2573  * @cq: The CQ to peek.
2574  * @wc_cnt: A minimum number of unreaped completions to check for.
2575  *
2576  * If the number of unreaped completions is greater than or equal to wc_cnt,
2577  * this function returns wc_cnt, otherwise, it returns the actual number of
2578  * unreaped completions.
2579  */
2580 int ib_peek_cq(struct ib_cq *cq, int wc_cnt);
2581 
2582 /**
2583  * ib_req_notify_cq - Request completion notification on a CQ.
2584  * @cq: The CQ to generate an event for.
2585  * @flags:
2586  *   Must contain exactly one of %IB_CQ_SOLICITED or %IB_CQ_NEXT_COMP
2587  *   to request an event on the next solicited event or next work
2588  *   completion at any type, respectively. %IB_CQ_REPORT_MISSED_EVENTS
2589  *   may also be |ed in to request a hint about missed events, as
2590  *   described below.
2591  *
2592  * Return Value:
2593  *    < 0 means an error occurred while requesting notification
2594  *   == 0 means notification was requested successfully, and if
2595  *        IB_CQ_REPORT_MISSED_EVENTS was passed in, then no events
2596  *        were missed and it is safe to wait for another event.  In
2597  *        this case is it guaranteed that any work completions added
2598  *        to the CQ since the last CQ poll will trigger a completion
2599  *        notification event.
2600  *    > 0 is only returned if IB_CQ_REPORT_MISSED_EVENTS was passed
2601  *        in.  It means that the consumer must poll the CQ again to
2602  *        make sure it is empty to avoid missing an event because of a
2603  *        race between requesting notification and an entry being
2604  *        added to the CQ.  This return value means it is possible
2605  *        (but not guaranteed) that a work completion has been added
2606  *        to the CQ since the last poll without triggering a
2607  *        completion notification event.
2608  */
2609 static inline int ib_req_notify_cq(struct ib_cq *cq,
2610 				   enum ib_cq_notify_flags flags)
2611 {
2612 	return cq->device->req_notify_cq(cq, flags);
2613 }
2614 
2615 /**
2616  * ib_req_ncomp_notif - Request completion notification when there are
2617  *   at least the specified number of unreaped completions on the CQ.
2618  * @cq: The CQ to generate an event for.
2619  * @wc_cnt: The number of unreaped completions that should be on the
2620  *   CQ before an event is generated.
2621  */
2622 static inline int ib_req_ncomp_notif(struct ib_cq *cq, int wc_cnt)
2623 {
2624 	return cq->device->req_ncomp_notif ?
2625 		cq->device->req_ncomp_notif(cq, wc_cnt) :
2626 		-ENOSYS;
2627 }
2628 
2629 /**
2630  * ib_get_dma_mr - Returns a memory region for system memory that is
2631  *   usable for DMA.
2632  * @pd: The protection domain associated with the memory region.
2633  * @mr_access_flags: Specifies the memory access rights.
2634  *
2635  * Note that the ib_dma_*() functions defined below must be used
2636  * to create/destroy addresses used with the Lkey or Rkey returned
2637  * by ib_get_dma_mr().
2638  */
2639 struct ib_mr *ib_get_dma_mr(struct ib_pd *pd, int mr_access_flags);
2640 
2641 /**
2642  * ib_dma_mapping_error - check a DMA addr for error
2643  * @dev: The device for which the dma_addr was created
2644  * @dma_addr: The DMA address to check
2645  */
2646 static inline int ib_dma_mapping_error(struct ib_device *dev, u64 dma_addr)
2647 {
2648 	if (dev->dma_ops)
2649 		return dev->dma_ops->mapping_error(dev, dma_addr);
2650 	return dma_mapping_error(dev->dma_device, dma_addr);
2651 }
2652 
2653 /**
2654  * ib_dma_map_single - Map a kernel virtual address to DMA address
2655  * @dev: The device for which the dma_addr is to be created
2656  * @cpu_addr: The kernel virtual address
2657  * @size: The size of the region in bytes
2658  * @direction: The direction of the DMA
2659  */
2660 static inline u64 ib_dma_map_single(struct ib_device *dev,
2661 				    void *cpu_addr, size_t size,
2662 				    enum dma_data_direction direction)
2663 {
2664 	if (dev->dma_ops)
2665 		return dev->dma_ops->map_single(dev, cpu_addr, size, direction);
2666 	return dma_map_single(dev->dma_device, cpu_addr, size, direction);
2667 }
2668 
2669 /**
2670  * ib_dma_unmap_single - Destroy a mapping created by ib_dma_map_single()
2671  * @dev: The device for which the DMA address was created
2672  * @addr: The DMA address
2673  * @size: The size of the region in bytes
2674  * @direction: The direction of the DMA
2675  */
2676 static inline void ib_dma_unmap_single(struct ib_device *dev,
2677 				       u64 addr, size_t size,
2678 				       enum dma_data_direction direction)
2679 {
2680 	if (dev->dma_ops)
2681 		dev->dma_ops->unmap_single(dev, addr, size, direction);
2682 	else
2683 		dma_unmap_single(dev->dma_device, addr, size, direction);
2684 }
2685 
2686 static inline u64 ib_dma_map_single_attrs(struct ib_device *dev,
2687 					  void *cpu_addr, size_t size,
2688 					  enum dma_data_direction direction,
2689 					  struct dma_attrs *attrs)
2690 {
2691 	return dma_map_single_attrs(dev->dma_device, cpu_addr, size,
2692 				    direction, attrs);
2693 }
2694 
2695 static inline void ib_dma_unmap_single_attrs(struct ib_device *dev,
2696 					     u64 addr, size_t size,
2697 					     enum dma_data_direction direction,
2698 					     struct dma_attrs *attrs)
2699 {
2700 	return dma_unmap_single_attrs(dev->dma_device, addr, size,
2701 				      direction, attrs);
2702 }
2703 
2704 /**
2705  * ib_dma_map_page - Map a physical page to DMA address
2706  * @dev: The device for which the dma_addr is to be created
2707  * @page: The page to be mapped
2708  * @offset: The offset within the page
2709  * @size: The size of the region in bytes
2710  * @direction: The direction of the DMA
2711  */
2712 static inline u64 ib_dma_map_page(struct ib_device *dev,
2713 				  struct page *page,
2714 				  unsigned long offset,
2715 				  size_t size,
2716 					 enum dma_data_direction direction)
2717 {
2718 	if (dev->dma_ops)
2719 		return dev->dma_ops->map_page(dev, page, offset, size, direction);
2720 	return dma_map_page(dev->dma_device, page, offset, size, direction);
2721 }
2722 
2723 /**
2724  * ib_dma_unmap_page - Destroy a mapping created by ib_dma_map_page()
2725  * @dev: The device for which the DMA address was created
2726  * @addr: The DMA address
2727  * @size: The size of the region in bytes
2728  * @direction: The direction of the DMA
2729  */
2730 static inline void ib_dma_unmap_page(struct ib_device *dev,
2731 				     u64 addr, size_t size,
2732 				     enum dma_data_direction direction)
2733 {
2734 	if (dev->dma_ops)
2735 		dev->dma_ops->unmap_page(dev, addr, size, direction);
2736 	else
2737 		dma_unmap_page(dev->dma_device, addr, size, direction);
2738 }
2739 
2740 /**
2741  * ib_dma_map_sg - Map a scatter/gather list to DMA addresses
2742  * @dev: The device for which the DMA addresses are to be created
2743  * @sg: The array of scatter/gather entries
2744  * @nents: The number of scatter/gather entries
2745  * @direction: The direction of the DMA
2746  */
2747 static inline int ib_dma_map_sg(struct ib_device *dev,
2748 				struct scatterlist *sg, int nents,
2749 				enum dma_data_direction direction)
2750 {
2751 	if (dev->dma_ops)
2752 		return dev->dma_ops->map_sg(dev, sg, nents, direction);
2753 	return dma_map_sg(dev->dma_device, sg, nents, direction);
2754 }
2755 
2756 /**
2757  * ib_dma_unmap_sg - Unmap a scatter/gather list of DMA addresses
2758  * @dev: The device for which the DMA addresses were created
2759  * @sg: The array of scatter/gather entries
2760  * @nents: The number of scatter/gather entries
2761  * @direction: The direction of the DMA
2762  */
2763 static inline void ib_dma_unmap_sg(struct ib_device *dev,
2764 				   struct scatterlist *sg, int nents,
2765 				   enum dma_data_direction direction)
2766 {
2767 	if (dev->dma_ops)
2768 		dev->dma_ops->unmap_sg(dev, sg, nents, direction);
2769 	else
2770 		dma_unmap_sg(dev->dma_device, sg, nents, direction);
2771 }
2772 
2773 static inline int ib_dma_map_sg_attrs(struct ib_device *dev,
2774 				      struct scatterlist *sg, int nents,
2775 				      enum dma_data_direction direction,
2776 				      struct dma_attrs *attrs)
2777 {
2778 	return dma_map_sg_attrs(dev->dma_device, sg, nents, direction, attrs);
2779 }
2780 
2781 static inline void ib_dma_unmap_sg_attrs(struct ib_device *dev,
2782 					 struct scatterlist *sg, int nents,
2783 					 enum dma_data_direction direction,
2784 					 struct dma_attrs *attrs)
2785 {
2786 	dma_unmap_sg_attrs(dev->dma_device, sg, nents, direction, attrs);
2787 }
2788 /**
2789  * ib_sg_dma_address - Return the DMA address from a scatter/gather entry
2790  * @dev: The device for which the DMA addresses were created
2791  * @sg: The scatter/gather entry
2792  */
2793 static inline u64 ib_sg_dma_address(struct ib_device *dev,
2794 				    struct scatterlist *sg)
2795 {
2796 	if (dev->dma_ops)
2797 		return dev->dma_ops->dma_address(dev, sg);
2798 	return sg_dma_address(sg);
2799 }
2800 
2801 /**
2802  * ib_sg_dma_len - Return the DMA length from a scatter/gather entry
2803  * @dev: The device for which the DMA addresses were created
2804  * @sg: The scatter/gather entry
2805  */
2806 static inline unsigned int ib_sg_dma_len(struct ib_device *dev,
2807 					 struct scatterlist *sg)
2808 {
2809 	if (dev->dma_ops)
2810 		return dev->dma_ops->dma_len(dev, sg);
2811 	return sg_dma_len(sg);
2812 }
2813 
2814 /**
2815  * ib_dma_sync_single_for_cpu - Prepare DMA region to be accessed by CPU
2816  * @dev: The device for which the DMA address was created
2817  * @addr: The DMA address
2818  * @size: The size of the region in bytes
2819  * @dir: The direction of the DMA
2820  */
2821 static inline void ib_dma_sync_single_for_cpu(struct ib_device *dev,
2822 					      u64 addr,
2823 					      size_t size,
2824 					      enum dma_data_direction dir)
2825 {
2826 	if (dev->dma_ops)
2827 		dev->dma_ops->sync_single_for_cpu(dev, addr, size, dir);
2828 	else
2829 		dma_sync_single_for_cpu(dev->dma_device, addr, size, dir);
2830 }
2831 
2832 /**
2833  * ib_dma_sync_single_for_device - Prepare DMA region to be accessed by device
2834  * @dev: The device for which the DMA address was created
2835  * @addr: The DMA address
2836  * @size: The size of the region in bytes
2837  * @dir: The direction of the DMA
2838  */
2839 static inline void ib_dma_sync_single_for_device(struct ib_device *dev,
2840 						 u64 addr,
2841 						 size_t size,
2842 						 enum dma_data_direction dir)
2843 {
2844 	if (dev->dma_ops)
2845 		dev->dma_ops->sync_single_for_device(dev, addr, size, dir);
2846 	else
2847 		dma_sync_single_for_device(dev->dma_device, addr, size, dir);
2848 }
2849 
2850 /**
2851  * ib_dma_alloc_coherent - Allocate memory and map it for DMA
2852  * @dev: The device for which the DMA address is requested
2853  * @size: The size of the region to allocate in bytes
2854  * @dma_handle: A pointer for returning the DMA address of the region
2855  * @flag: memory allocator flags
2856  */
2857 static inline void *ib_dma_alloc_coherent(struct ib_device *dev,
2858 					   size_t size,
2859 					   u64 *dma_handle,
2860 					   gfp_t flag)
2861 {
2862 	if (dev->dma_ops)
2863 		return dev->dma_ops->alloc_coherent(dev, size, dma_handle, flag);
2864 	else {
2865 		dma_addr_t handle;
2866 		void *ret;
2867 
2868 		ret = dma_alloc_coherent(dev->dma_device, size, &handle, flag);
2869 		*dma_handle = handle;
2870 		return ret;
2871 	}
2872 }
2873 
2874 /**
2875  * ib_dma_free_coherent - Free memory allocated by ib_dma_alloc_coherent()
2876  * @dev: The device for which the DMA addresses were allocated
2877  * @size: The size of the region
2878  * @cpu_addr: the address returned by ib_dma_alloc_coherent()
2879  * @dma_handle: the DMA address returned by ib_dma_alloc_coherent()
2880  */
2881 static inline void ib_dma_free_coherent(struct ib_device *dev,
2882 					size_t size, void *cpu_addr,
2883 					u64 dma_handle)
2884 {
2885 	if (dev->dma_ops)
2886 		dev->dma_ops->free_coherent(dev, size, cpu_addr, dma_handle);
2887 	else
2888 		dma_free_coherent(dev->dma_device, size, cpu_addr, dma_handle);
2889 }
2890 
2891 /**
2892  * ib_reg_phys_mr - Prepares a virtually addressed memory region for use
2893  *   by an HCA.
2894  * @pd: The protection domain associated assigned to the registered region.
2895  * @phys_buf_array: Specifies a list of physical buffers to use in the
2896  *   memory region.
2897  * @num_phys_buf: Specifies the size of the phys_buf_array.
2898  * @mr_access_flags: Specifies the memory access rights.
2899  * @iova_start: The offset of the region's starting I/O virtual address.
2900  */
2901 struct ib_mr *ib_reg_phys_mr(struct ib_pd *pd,
2902 			     struct ib_phys_buf *phys_buf_array,
2903 			     int num_phys_buf,
2904 			     int mr_access_flags,
2905 			     u64 *iova_start);
2906 
2907 /**
2908  * ib_rereg_phys_mr - Modifies the attributes of an existing memory region.
2909  *   Conceptually, this call performs the functions deregister memory region
2910  *   followed by register physical memory region.  Where possible,
2911  *   resources are reused instead of deallocated and reallocated.
2912  * @mr: The memory region to modify.
2913  * @mr_rereg_mask: A bit-mask used to indicate which of the following
2914  *   properties of the memory region are being modified.
2915  * @pd: If %IB_MR_REREG_PD is set in mr_rereg_mask, this field specifies
2916  *   the new protection domain to associated with the memory region,
2917  *   otherwise, this parameter is ignored.
2918  * @phys_buf_array: If %IB_MR_REREG_TRANS is set in mr_rereg_mask, this
2919  *   field specifies a list of physical buffers to use in the new
2920  *   translation, otherwise, this parameter is ignored.
2921  * @num_phys_buf: If %IB_MR_REREG_TRANS is set in mr_rereg_mask, this
2922  *   field specifies the size of the phys_buf_array, otherwise, this
2923  *   parameter is ignored.
2924  * @mr_access_flags: If %IB_MR_REREG_ACCESS is set in mr_rereg_mask, this
2925  *   field specifies the new memory access rights, otherwise, this
2926  *   parameter is ignored.
2927  * @iova_start: The offset of the region's starting I/O virtual address.
2928  */
2929 int ib_rereg_phys_mr(struct ib_mr *mr,
2930 		     int mr_rereg_mask,
2931 		     struct ib_pd *pd,
2932 		     struct ib_phys_buf *phys_buf_array,
2933 		     int num_phys_buf,
2934 		     int mr_access_flags,
2935 		     u64 *iova_start);
2936 
2937 /**
2938  * ib_query_mr - Retrieves information about a specific memory region.
2939  * @mr: The memory region to retrieve information about.
2940  * @mr_attr: The attributes of the specified memory region.
2941  */
2942 int ib_query_mr(struct ib_mr *mr, struct ib_mr_attr *mr_attr);
2943 
2944 /**
2945  * ib_dereg_mr - Deregisters a memory region and removes it from the
2946  *   HCA translation table.
2947  * @mr: The memory region to deregister.
2948  *
2949  * This function can fail, if the memory region has memory windows bound to it.
2950  */
2951 int ib_dereg_mr(struct ib_mr *mr);
2952 
2953 
2954 /**
2955  * ib_create_mr - Allocates a memory region that may be used for
2956  *     signature handover operations.
2957  * @pd: The protection domain associated with the region.
2958  * @mr_init_attr: memory region init attributes.
2959  */
2960 struct ib_mr *ib_create_mr(struct ib_pd *pd,
2961 			   struct ib_mr_init_attr *mr_init_attr);
2962 
2963 /**
2964  * ib_destroy_mr - Destroys a memory region that was created using
2965  *     ib_create_mr and removes it from HW translation tables.
2966  * @mr: The memory region to destroy.
2967  *
2968  * This function can fail, if the memory region has memory windows bound to it.
2969  */
2970 int ib_destroy_mr(struct ib_mr *mr);
2971 
2972 /**
2973  * ib_alloc_fast_reg_mr - Allocates memory region usable with the
2974  *   IB_WR_FAST_REG_MR send work request.
2975  * @pd: The protection domain associated with the region.
2976  * @max_page_list_len: requested max physical buffer list length to be
2977  *   used with fast register work requests for this MR.
2978  */
2979 struct ib_mr *ib_alloc_fast_reg_mr(struct ib_pd *pd, int max_page_list_len);
2980 
2981 /**
2982  * ib_alloc_fast_reg_page_list - Allocates a page list array
2983  * @device - ib device pointer.
2984  * @page_list_len - size of the page list array to be allocated.
2985  *
2986  * This allocates and returns a struct ib_fast_reg_page_list * and a
2987  * page_list array that is at least page_list_len in size.  The actual
2988  * size is returned in max_page_list_len.  The caller is responsible
2989  * for initializing the contents of the page_list array before posting
2990  * a send work request with the IB_WC_FAST_REG_MR opcode.
2991  *
2992  * The page_list array entries must be translated using one of the
2993  * ib_dma_*() functions just like the addresses passed to
2994  * ib_map_phys_fmr().  Once the ib_post_send() is issued, the struct
2995  * ib_fast_reg_page_list must not be modified by the caller until the
2996  * IB_WC_FAST_REG_MR work request completes.
2997  */
2998 struct ib_fast_reg_page_list *ib_alloc_fast_reg_page_list(
2999 				struct ib_device *device, int page_list_len);
3000 
3001 /**
3002  * ib_free_fast_reg_page_list - Deallocates a previously allocated
3003  *   page list array.
3004  * @page_list - struct ib_fast_reg_page_list pointer to be deallocated.
3005  */
3006 void ib_free_fast_reg_page_list(struct ib_fast_reg_page_list *page_list);
3007 
3008 /**
3009  * ib_update_fast_reg_key - updates the key portion of the fast_reg MR
3010  *   R_Key and L_Key.
3011  * @mr - struct ib_mr pointer to be updated.
3012  * @newkey - new key to be used.
3013  */
3014 static inline void ib_update_fast_reg_key(struct ib_mr *mr, u8 newkey)
3015 {
3016 	mr->lkey = (mr->lkey & 0xffffff00) | newkey;
3017 	mr->rkey = (mr->rkey & 0xffffff00) | newkey;
3018 }
3019 
3020 /**
3021  * ib_inc_rkey - increments the key portion of the given rkey. Can be used
3022  * for calculating a new rkey for type 2 memory windows.
3023  * @rkey - the rkey to increment.
3024  */
3025 static inline u32 ib_inc_rkey(u32 rkey)
3026 {
3027 	const u32 mask = 0x000000ff;
3028 	return ((rkey + 1) & mask) | (rkey & ~mask);
3029 }
3030 
3031 /**
3032  * ib_alloc_mw - Allocates a memory window.
3033  * @pd: The protection domain associated with the memory window.
3034  * @type: The type of the memory window (1 or 2).
3035  */
3036 struct ib_mw *ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type);
3037 
3038 /**
3039  * ib_bind_mw - Posts a work request to the send queue of the specified
3040  *   QP, which binds the memory window to the given address range and
3041  *   remote access attributes.
3042  * @qp: QP to post the bind work request on.
3043  * @mw: The memory window to bind.
3044  * @mw_bind: Specifies information about the memory window, including
3045  *   its address range, remote access rights, and associated memory region.
3046  *
3047  * If there is no immediate error, the function will update the rkey member
3048  * of the mw parameter to its new value. The bind operation can still fail
3049  * asynchronously.
3050  */
3051 static inline int ib_bind_mw(struct ib_qp *qp,
3052 			     struct ib_mw *mw,
3053 			     struct ib_mw_bind *mw_bind)
3054 {
3055 	/* XXX reference counting in corresponding MR? */
3056 	return mw->device->bind_mw ?
3057 		mw->device->bind_mw(qp, mw, mw_bind) :
3058 		-ENOSYS;
3059 }
3060 
3061 /**
3062  * ib_dealloc_mw - Deallocates a memory window.
3063  * @mw: The memory window to deallocate.
3064  */
3065 int ib_dealloc_mw(struct ib_mw *mw);
3066 
3067 /**
3068  * ib_alloc_fmr - Allocates a unmapped fast memory region.
3069  * @pd: The protection domain associated with the unmapped region.
3070  * @mr_access_flags: Specifies the memory access rights.
3071  * @fmr_attr: Attributes of the unmapped region.
3072  *
3073  * A fast memory region must be mapped before it can be used as part of
3074  * a work request.
3075  */
3076 struct ib_fmr *ib_alloc_fmr(struct ib_pd *pd,
3077 			    int mr_access_flags,
3078 			    struct ib_fmr_attr *fmr_attr);
3079 
3080 /**
3081  * ib_map_phys_fmr - Maps a list of physical pages to a fast memory region.
3082  * @fmr: The fast memory region to associate with the pages.
3083  * @page_list: An array of physical pages to map to the fast memory region.
3084  * @list_len: The number of pages in page_list.
3085  * @iova: The I/O virtual address to use with the mapped region.
3086  */
3087 static inline int ib_map_phys_fmr(struct ib_fmr *fmr,
3088 				  u64 *page_list, int list_len,
3089 				  u64 iova)
3090 {
3091 	return fmr->device->map_phys_fmr(fmr, page_list, list_len, iova);
3092 }
3093 
3094 /**
3095  * ib_unmap_fmr - Removes the mapping from a list of fast memory regions.
3096  * @fmr_list: A linked list of fast memory regions to unmap.
3097  */
3098 int ib_unmap_fmr(struct list_head *fmr_list);
3099 
3100 /**
3101  * ib_dealloc_fmr - Deallocates a fast memory region.
3102  * @fmr: The fast memory region to deallocate.
3103  */
3104 int ib_dealloc_fmr(struct ib_fmr *fmr);
3105 
3106 /**
3107  * ib_attach_mcast - Attaches the specified QP to a multicast group.
3108  * @qp: QP to attach to the multicast group.  The QP must be type
3109  *   IB_QPT_UD.
3110  * @gid: Multicast group GID.
3111  * @lid: Multicast group LID in host byte order.
3112  *
3113  * In order to send and receive multicast packets, subnet
3114  * administration must have created the multicast group and configured
3115  * the fabric appropriately.  The port associated with the specified
3116  * QP must also be a member of the multicast group.
3117  */
3118 int ib_attach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid);
3119 
3120 /**
3121  * ib_detach_mcast - Detaches the specified QP from a multicast group.
3122  * @qp: QP to detach from the multicast group.
3123  * @gid: Multicast group GID.
3124  * @lid: Multicast group LID in host byte order.
3125  */
3126 int ib_detach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid);
3127 
3128 /**
3129  * ib_alloc_xrcd - Allocates an XRC domain.
3130  * @device: The device on which to allocate the XRC domain.
3131  */
3132 struct ib_xrcd *ib_alloc_xrcd(struct ib_device *device);
3133 
3134 /**
3135  * ib_dealloc_xrcd - Deallocates an XRC domain.
3136  * @xrcd: The XRC domain to deallocate.
3137  */
3138 int ib_dealloc_xrcd(struct ib_xrcd *xrcd);
3139 
3140 struct ib_flow *ib_create_flow(struct ib_qp *qp,
3141 			       struct ib_flow_attr *flow_attr, int domain);
3142 int ib_destroy_flow(struct ib_flow *flow_id);
3143 
3144 struct ib_dct *ib_create_dct(struct ib_pd *pd, struct ib_dct_init_attr *attr,
3145 			     struct ib_udata *udata);
3146 int ib_destroy_dct(struct ib_dct *dct);
3147 int ib_query_dct(struct ib_dct *dct, struct ib_dct_attr *attr);
3148 
3149 int ib_query_values(struct ib_device *device,
3150                    int q_values, struct ib_device_values *values);
3151 
3152 static inline void ib_active_speed_enum_to_rate(u8 active_speed,
3153 						int *rate,
3154 						char **speed)
3155 {
3156 	switch (active_speed) {
3157 	case IB_SPEED_DDR:
3158 		*speed = " DDR";
3159 		*rate = 50;
3160 		break;
3161 	case IB_SPEED_QDR:
3162 		*speed = " QDR";
3163 		*rate = 100;
3164 		break;
3165 	case IB_SPEED_FDR10:
3166 		*speed = " FDR10";
3167 		*rate = 100;
3168 		break;
3169 	case IB_SPEED_FDR:
3170 		*speed = " FDR";
3171 		*rate = 140;
3172 		break;
3173 	case IB_SPEED_EDR:
3174 		*speed = " EDR";
3175 		*rate = 250;
3176 		break;
3177 	case IB_SPEED_SDR:
3178 	default:		/* default to SDR for invalid rates */
3179 		*rate = 25;
3180 		break;
3181 	}
3182 
3183 }
3184 
3185 static inline int ib_check_mr_access(int flags)
3186 {
3187 	/*
3188 	 * Local write permission is required if remote write or
3189 	 * remote atomic permission is also requested.
3190 	 */
3191 	if (flags & (IB_ACCESS_REMOTE_ATOMIC | IB_ACCESS_REMOTE_WRITE) &&
3192 	    !(flags & IB_ACCESS_LOCAL_WRITE))
3193 		return -EINVAL;
3194 
3195 	return 0;
3196 }
3197 
3198 /**
3199  * ib_check_mr_status: lightweight check of MR status.
3200  *     This routine may provide status checks on a selected
3201  *     ib_mr. first use is for signature status check.
3202  *
3203  * @mr: A memory region.
3204  * @check_mask: Bitmask of which checks to perform from
3205  *     ib_mr_status_check enumeration.
3206  * @mr_status: The container of relevant status checks.
3207  *     failed checks will be indicated in the status bitmask
3208  *     and the relevant info shall be in the error item.
3209  */
3210 int ib_check_mr_status(struct ib_mr *mr, u32 check_mask,
3211 		       struct ib_mr_status *mr_status);
3212 
3213 #endif /* IB_VERBS_H */
3214