xref: /illumos-gate/usr/src/uts/common/sys/ib/ibtl/ibti_common.h (revision 1a1a84a324206b6b1f5f704ab166c4ebf78aed76)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #ifndef	_SYS_IB_IBTL_IBTI_COMMON_H
27 #define	_SYS_IB_IBTL_IBTI_COMMON_H
28 
29 /*
30  * ibti_common.h
31  *
32  * This file contains the shared/common transport data types and function
33  * prototypes.
34  */
35 #include <sys/ib/ibtl/ibtl_types.h>
36 #include <sys/ib/ibtl/ibti_cm.h>
37 #include <sys/isa_defs.h>
38 #include <sys/byteorder.h>
39 
40 #ifdef __cplusplus
41 extern "C" {
42 #endif
43 
44 /*
45  * Max number of paths that can be requested in an ibt_get_paths() call,
46  * if IBT_PATH_PERF or IBT_PATH_AVAIL flag (ibt_path_flags_t) is set.
47  */
48 #define	IBT_MAX_SPECIAL_PATHS	2
49 
50 /*
51  * The name of DDI Event, generated when the properties of IOC device
52  * node properties were modified.
53  */
54 #define	IB_PROP_UPDATE_EVENT	"SUNW,IB:IB_PROP_UPDATE"
55 
56 
57 /* Transport Interface version */
58 typedef enum ibt_version_e {
59 	IBTI_V1 = 1,
60 	IBTI_V2 = 2,		/* FMR Support */
61 	IBTI_V3 = 3,
62 	IBTI_V_CURR = IBTI_V3
63 } ibt_version_t;
64 
65 /*
66  * Driver class type. Identifies a type of client driver so that
67  * "IBTF Policy" decisions can be made on a driver class basis.
68  * The last class should always be IBT_CLNT_NUM, and any new classes added
69  * must be defined before IBT_CLNT_NUM. The class values must be above 0.
70  * Any class values below or equal to 0 shall be invalid
71  */
72 typedef enum ibt_clnt_class_e {
73 	IBT_STORAGE_DEV = 0x1,	/* SCSI, FC, etc.. */
74 	IBT_NETWORK_DEV,	/* Network driver with associated client H/W */
75 	IBT_GENERIC_DEV,	/* Generic client H/W device driver */
76 	IBT_NETWORK,		/* Network driver with no associated */
77 				/* client H/W, e.g., IPoIB */
78 	IBT_GENERIC,		/* A generic IB driver not */
79 				/* associated with client H/W */
80 	IBT_USER,		/* A user application IBT interface driver */
81 	IBT_IBMA,		/* The IBMA Module */
82 	IBT_CM,			/* The CM Module */
83 	IBT_DM,			/* The DM Module */
84 	IBT_CLASS_NUM		/* Place holder for class count */
85 } ibt_clnt_class_t;
86 
87 #define	IBT_TEST_DEV	999	/* Place holder for modules that test IBTL */
88 
89 #define	IBT_CLNT_DEVICE_CLASS(class)	((class) == IBT_STORAGE_DEV || \
90 					(class) == IBT_NETWORK_DEV || \
91 					(class) == IBT_GENERIC_DEV)
92 
93 #define	IBT_CLNT_GENERAL_CLASS(class)	((class) == IBT_NETWORK || \
94 					(class) == IBT_GENERIC || \
95 					(class) == IBT_USER)
96 
97 #define	IBT_CLNT_MGMT_CLASS(class)	((class) == IBT_IBMA || \
98 					(class) == IBT_CM || \
99 					(class) == IBT_DM || \
100 					(class) == IBT_TEST_DEV)
101 /*
102  * Event record & status returns for asynchronous events and errors.
103  */
104 typedef struct ibt_async_event_s {
105 	uint64_t		ev_fma_ena;		/* FMA Error data */
106 	ibt_channel_hdl_t	ev_chan_hdl;		/* Channel handle */
107 	ibt_cq_hdl_t		ev_cq_hdl;		/* CQ handle */
108 	ib_guid_t		ev_hca_guid;		/* HCA node GUID */
109 	ibt_srq_hdl_t		ev_srq_hdl;		/* SRQ handle */
110 	ibt_port_change_t	ev_port_flags;		/* Port Change flags */
111 	uint8_t			ev_port;		/* HCA port */
112 } ibt_async_event_t;
113 
114 /*
115  * IBT Client Callback function typedefs.
116  *
117  * ibt_async_handler_t
118  *	Pointer to an async event/error handler function.  This function is
119  *	called when an async event/error is detected on a HCA that is being
120  *	used by the IBT client driver that registered the function.
121  */
122 typedef void (*ibt_async_handler_t)(void *clnt_private,
123     ibt_hca_hdl_t hca_hdl, ibt_async_code_t code, ibt_async_event_t *event);
124 
125 /*
126  * IBT Client Memory Error Callback function typedefs.
127  *
128  * ibt_memory_handler_t
129  *	Pointer to an memory event/error handler function.
130  */
131 typedef void (*ibt_memory_handler_t)(void *clnt_private,
132     ibt_hca_hdl_t hca_hdl, ibt_mem_code_t code, ibt_mem_data_t *data);
133 
134 /*
135  * Define a client module information structure. All clients MUST
136  * define a global of type ibt_clnt_modinfo_t. A pointer to this global
137  * is passed into the IBTF by a client when calling ibt_attach().
138  * This struct must persist during the life of the client.
139  *
140  * The client's mi_async_handler is called when an async event/error is
141  * detected on a HCA that is being used by this client.
142  */
143 typedef struct ibt_clnt_modinfo_s {
144 	ibt_version_t		mi_ibt_version;		/* TI version */
145 	ibt_clnt_class_t	mi_clnt_class;		/* Type of client */
146 	ibt_async_handler_t	mi_async_handler;	/* Async Handler */
147 	ibt_memory_handler_t	mi_reserved;		/* Memory handler */
148 	char			*mi_clnt_name;		/* Client Name. */
149 } ibt_clnt_modinfo_t;
150 
151 
152 /*
153  * Definitions for use with ibt_register_subnet_notices()
154  */
155 typedef enum ibt_subnet_event_code_e {
156 	IBT_SM_EVENT_MCG_CREATED = 1,
157 	IBT_SM_EVENT_MCG_DELETED = 2,
158 	IBT_SM_EVENT_AVAILABLE	 = 3,
159 	IBT_SM_EVENT_UNAVAILABLE = 4,
160 	IBT_SM_EVENT_GID_AVAIL	 = 5,
161 	IBT_SM_EVENT_GID_UNAVAIL = 6
162 } ibt_subnet_event_code_t;
163 
164 typedef struct ibt_subnet_event_s {
165 	ib_gid_t sm_notice_gid;
166 } ibt_subnet_event_t;
167 
168 typedef void (*ibt_sm_notice_handler_t)(void *private, ib_gid_t gid,
169     ibt_subnet_event_code_t code, ibt_subnet_event_t *event);
170 
171 
172 /*
173  * MTU Request type.
174  */
175 typedef struct ibt_mtu_req_s {
176 	ib_mtu_t	r_mtu;		/* Requested MTU */
177 	ibt_selector_t	r_selector;	/* Qualifier for r_mtu */
178 } ibt_mtu_req_t;
179 
180 
181 /*
182  * Qflags, used by ibt_resize_queues().
183  */
184 typedef enum ibt_qflags_e {
185 	IBT_SEND_Q	= 1 << 0,	/* Op applies to the Send Q */
186 	IBT_RECV_Q	= 1 << 1	/* Op applies to the Recv Q */
187 } ibt_qflags_t;
188 
189 /*
190  * CQ priorities
191  * The IBTF will attempt to implement a coarse 3 level priority scheme
192  * (IBT_CQ_LOW, IBT_CQ_MEDIUM, IBT_CQ_HIGH) based on the class of client
193  * driver. The requested priority is not guaranteed. If a CI implementation
194  * has the ability to implement priority CQs, then the IBTF will take advantage
195  * of that when calling the CI to create a CQ by passing a priority indicator
196  * to the CI.
197  */
198 typedef enum ibt_cq_priority_e {
199 	IBT_CQ_DEFAULT		= 0x0,
200 	IBT_CQ_LOW		= 0x1,
201 	IBT_CQ_MEDIUM		= 0x2,
202 	IBT_CQ_HIGH		= 0x3,
203 	IBT_CQ_OPAQUE_1		= 0x4,
204 	IBT_CQ_OPAQUE_2		= 0x5,
205 	IBT_CQ_OPAQUE_3		= 0x6,
206 	IBT_CQ_OPAQUE_4		= 0x7,
207 	IBT_CQ_OPAQUE_5		= 0x8,
208 	IBT_CQ_OPAQUE_6		= 0x9,
209 	IBT_CQ_OPAQUE_7		= 0xA,
210 	IBT_CQ_OPAQUE_8		= 0xB,
211 	IBT_CQ_OPAQUE_9		= 0xC,
212 	IBT_CQ_OPAQUE_10	= 0xD,
213 	IBT_CQ_OPAQUE_11	= 0xE,
214 	IBT_CQ_OPAQUE_12	= 0xF,
215 	IBT_CQ_OPAQUE_13	= 0x10,
216 	IBT_CQ_OPAQUE_14	= 0x11,
217 	IBT_CQ_OPAQUE_15	= 0x12,
218 	IBT_CQ_OPAQUE_16	= 0x13
219 } ibt_cq_priority_t;
220 
221 /*
222  * Attributes when creating a Completion Queue Scheduling Handle.
223  */
224 typedef struct ibt_cq_sched_attr_s {
225 	ibt_cq_sched_flags_t	cqs_flags;
226 	ibt_cq_priority_t	cqs_priority;
227 	uint_t			cqs_load;
228 	ibt_sched_hdl_t		cqs_affinity_hdl;
229 } ibt_cq_sched_attr_t;
230 
231 
232 /*
233  * ibt_cq_handler_t
234  *	Pointer to a work request completion handler function.  This function
235  *	is called when a WR completes on a CQ that is being used by the IBTF
236  *	client driver that registered the function.
237  */
238 typedef void (*ibt_cq_handler_t)(ibt_cq_hdl_t ibt_cq, void *arg);
239 
240 /* default CQ handler ID */
241 #define	IBT_CQ_HID_DEFAULT	(1)
242 
243 /*
244  * Service Data and flags.
245  *	(IBTA Spec Release 1.1, Vol-1 Ref: 15.2.5.14.4)
246  *
247  * The ServiceData8.1 (sb_data8[0]) through ServiceData64.2 (sb_data64[1])
248  * components together constitutes a 64-byte area in which any data may be
249  * placed. It is intended to be a convenient way for a service to provide its
250  * clients with some initial data.
251  *
252  * In addition, this 64-byte area is formally divided into a total of 30
253  * components, 16 8-bit (uint8_t) components, then 8 16-bit (uint16_t)
254  * components, then 6 32-bit (uint32_t) components, then 2 64-bit (uint64_t)
255  * components,  thereby assigning ComponentMask bits (ibt_srv_data_flags_t) to
256  * variously-sized segments of the data. All data are in host endian format.
257  * This allows query operations (ibt_get_paths()) to be used which match
258  * parts of the Service Data, making it possible, for example, for
259  * service-specific parts of the ServiceData to serve as a binary-coded
260  * extension to the ServiceName for purposes of lookup.
261  */
262 typedef enum ibt_srv_data_flags_e {
263 	IBT_NO_SDATA	= 0,
264 
265 	IBT_SDATA8_0	= (1 << 0),
266 	IBT_SDATA8_1	= (1 << 1),
267 	IBT_SDATA8_2	= (1 << 2),
268 	IBT_SDATA8_3	= (1 << 3),
269 	IBT_SDATA8_4	= (1 << 4),
270 	IBT_SDATA8_5	= (1 << 5),
271 	IBT_SDATA8_6	= (1 << 6),
272 	IBT_SDATA8_7	= (1 << 7),
273 	IBT_SDATA8_8	= (1 << 8),
274 	IBT_SDATA8_9	= (1 << 9),
275 	IBT_SDATA8_10	= (1 << 10),
276 	IBT_SDATA8_11	= (1 << 11),
277 	IBT_SDATA8_12	= (1 << 12),
278 	IBT_SDATA8_13	= (1 << 13),
279 	IBT_SDATA8_14	= (1 << 14),
280 	IBT_SDATA8_15	= (1 << 15),
281 
282 	IBT_SDATA16_0	= (1 << 16),
283 	IBT_SDATA16_1	= (1 << 17),
284 	IBT_SDATA16_2	= (1 << 18),
285 	IBT_SDATA16_3	= (1 << 19),
286 	IBT_SDATA16_4	= (1 << 20),
287 	IBT_SDATA16_5	= (1 << 21),
288 	IBT_SDATA16_6	= (1 << 22),
289 	IBT_SDATA16_7	= (1 << 23),
290 
291 	IBT_SDATA32_0	= (1 << 24),
292 	IBT_SDATA32_1	= (1 << 25),
293 	IBT_SDATA32_2	= (1 << 26),
294 	IBT_SDATA32_3	= (1 << 27),
295 
296 	IBT_SDATA64_0	= (1 << 28),
297 	IBT_SDATA64_1	= (1 << 29),
298 
299 	IBT_SDATA_ALL	= 0x3FFFFFFF
300 } ibt_srv_data_flags_t;
301 
302 typedef struct ibt_srv_data_s {
303 	uint8_t		s_data8[16];	/* 8-bit service data fields. */
304 	uint16_t	s_data16[8];	/* 16-bit service data fields. */
305 	uint32_t	s_data32[4];	/* 32-bit service data fields. */
306 	uint64_t	s_data64[2];	/* 64-bit service data fields. */
307 } ibt_srv_data_t;
308 
309 /*
310  * Path flags, used in ibt_get_paths()
311  */
312 typedef enum ibt_path_flags_e {
313 	IBT_PATH_NO_FLAGS	= 0,
314 	IBT_PATH_APM		= 1 << 0,	/* APM is desired. */
315 	IBT_PATH_AVAIL		= 1 << 2,
316 	IBT_PATH_PERF		= 1 << 3,
317 	IBT_PATH_MULTI_SVC_DEST	= 1 << 4,	/* Multiple ServiceRecords */
318 	IBT_PATH_HOP		= 1 << 5,	/* pa_hop is specified. */
319 	IBT_PATH_PKEY		= 1 << 6	/* pa_pkey is specified. */
320 } ibt_path_flags_t;
321 
322 /*
323  * Path attributes.
324  *
325  * The ibt_path_attr_t structure is used to specify required attributes in a
326  * path from the requesting (source) node to a specified destination node.
327  * Attributes that are don't care should be set to NULL or '0'.
328  * A destination must be specified, where a destination can be defined as
329  * one of the following:
330  *
331  *	o Service Name
332  *	o Service ID (SID)
333  *	o Array of DGIDs.
334  *	o Service Name and Array of DGIDs.
335  */
336 typedef struct ibt_path_attr_s {
337 	ib_gid_t		*pa_dgids;	/* Array of DGIDs */
338 	ib_gid_t		pa_sgid;
339 	ib_guid_t		pa_hca_guid;
340 	char			*pa_sname;	/* ASCII Service name  */
341 						/* NULL Terminated */
342 	ib_svc_id_t		pa_sid;		/* Service ID */
343 	ibt_srv_data_flags_t	pa_sd_flags;	/* Service Data flags. */
344 	ibt_srv_data_t		pa_sdata;	/* Service Data */
345 	uint8_t			pa_hca_port_num;
346 	uint8_t			pa_num_dgids;	/* size of pa_dgids array */
347 	uint8_t			pa_sl:4;
348 	ibt_mtu_req_t		pa_mtu;
349 	ibt_srate_req_t		pa_srate;
350 	ibt_pkt_lt_req_t	pa_pkt_lt;	/* Packet Life Time Request */
351 	uint_t			pa_flow:20;
352 	uint8_t			pa_hop;		/* IBT_PATH_HOP */
353 	uint8_t			pa_tclass;
354 	ib_pkey_t		pa_pkey;	/* IBT_PATH_PKEY */
355 } ibt_path_attr_t;
356 
357 /*
358  * Path Information.
359  *
360  * The ibt_get_paths() performs SA Path record lookups to select a path(s) to
361  * a given destination(s), details of selected path(s) are returned in this
362  * structure.
363  *
364  * The ibt_path_info_t contains all the attributes of the best path(s), as
365  * as determined by IBTL, to the specified destination(s), including the
366  * local HCA and HCA port to use to access the fabric.
367  *
368  * The Service ID (pi_sid) and Service Data (pi_sdata) are returned only for
369  * lookups based on Service ID or/and Service Name.
370  */
371 typedef struct ibt_path_info_s {
372 	ib_guid_t	pi_hca_guid;		/* Local HCA GUID; 0 implies */
373 						/* this record is invalid */
374 	ib_svc_id_t	pi_sid;			/* Service ID */
375 	ibt_srv_data_t	pi_sdata;		/* Service Data */
376 
377 	ibt_cep_path_t	pi_prim_cep_path;	/* Contains CEP adds info */
378 	ibt_cep_path_t	pi_alt_cep_path;	/* RC & UC Only, valid if */
379 						/* cep_hca_port_num is not */
380 						/* '0' */
381 	ib_mtu_t	pi_path_mtu;		/* Common path MTU */
382 	ib_time_t	pi_prim_pkt_lt;
383 	ib_time_t	pi_alt_pkt_lt;
384 } ibt_path_info_t;
385 
386 /*
387  * Optional Alternate Path attributes.
388  *
389  * The ibt_alt_path_attr_t structure is used to specify additional optional
390  * attributes when requesting an alternate path for an existing channel.
391  *
392  * Attributes that are don't care should be set to NULL or '0'.
393  */
394 typedef struct ibt_alt_path_attr_s {
395 	ib_gid_t		apa_sgid;
396 	ib_gid_t		apa_dgid;
397 	ibt_srate_req_t		apa_srate;
398 	ibt_pkt_lt_req_t	apa_pkt_lt;	/* Packet Life Time Request */
399 	uint_t			apa_flow:20;
400 	uint8_t			apa_sl:4;
401 	uint8_t			apa_hop;
402 	uint8_t			apa_tclass;
403 } ibt_alt_path_attr_t;
404 
405 /*
406  * Path Information for Alternate Path - input to ibt_set_alt_path().
407  */
408 typedef struct ibt_alt_path_info_s {
409 	ibt_cep_path_t	ap_alt_cep_path;	/* RC & UC Only, valid if */
410 						/* cep_hca_port_num is not */
411 						/* '0' */
412 	ib_time_t	ap_alt_pkt_lt;
413 } ibt_alt_path_info_t;
414 
415 /*
416  * Open Channel flags, Used in ibt_open_rc_channel call
417  */
418 typedef enum ibt_chan_open_flags_e {
419 	IBT_OCHAN_NO_FLAGS		= 0,
420 	IBT_OCHAN_REDIRECTED		= 1 << 0,
421 	IBT_OCHAN_PORT_REDIRECTED	= 1 << 1,
422 	IBT_OCHAN_DUP			= 1 << 2,
423 	IBT_OCHAN_PORT_FIXED		= 1 << 3,
424 	IBT_OCHAN_OPAQUE1		= 1 << 4,
425 	IBT_OCHAN_OPAQUE2		= 1 << 5,
426 	IBT_OCHAN_OPAQUE3		= 1 << 6,
427 	IBT_OCHAN_OPAQUE4		= 1 << 7,
428 	IBT_OCHAN_OPAQUE5		= 1 << 8
429 } ibt_chan_open_flags_t;
430 
431 /*
432  * Arguments for ibt_open_rc_channel().
433  *
434  * oc_priv_data should be NULL or point to a buffer allocated by the caller,
435  * the size of which should be in oc_priv_data_len, where oc_priv_data_len <=
436  * IBT_REQ_PRIV_DATA_SZ.
437  *
438  * When ibt_open_rc_channel returns with ibt_cm_reason_t of
439  * IBT_CM_REDIRECT_PORT, the client can re-issue ibt_open_rc_channel setting
440  * new fields as follows:
441  *
442  * Set (ibt_chan_args_t)->oc_cm_cep_path  =
443  *    original (ibt_chan_open_args_t)->oc_path->pi_prim_cep_path.
444  * Set (ibt_chan_args_t)->oc_cm_pkt_lt  =
445  *    original (ibt_chan_open_args_t)->oc_prim_pkt_lt.
446  * Update (ibt_chan_args_t)->oc_path based on path information returned
447  * from ibt_get_paths using the gid in the return data below:
448  * 	(ibt_rc_returns_t)->rc_arej_info.ari_redirect_info.ari_gid.
449  * Set flags to IBT_OCHAN_PORT_REDIRECTED.
450  *
451  * Note : oc_cm_path is not used for any other scenario, and must be set for
452  * IBT_OCHAN_PORT_REDIRECTED.
453  *
454  * When ibt_open_rc_channel returns with ibt_cm_reason_t of
455  * IBT_CM_REDIRECT_CM, the client can re-issue ibt_open_rc_channel setting
456  * new fields as follows:
457  *
458  * Update (ibt_chan_args_t)->oc_path based on path information returned
459  * from ibt_get_paths using the return data in
460  * (ibt_rc_returns_t)->rc_arej_info.ari_redirect_info.
461  *
462  * Set (ibt_chan_args_t)->oc_cm_redirect_info =
463  *    Returned (ibt_rc_returns_t)->rc_arej_info.ari_redirect_info.
464  * Set flags to IBT_OCHAN_REDIRECTED.
465  *
466  * Note:
467  *
468  * IBT_OCHAN_PORT_REDIRECTED flag cannot be used to specify a remote CM MAD
469  * address, that is on a different subnet than the RC connection itself.
470  *
471  * Not specified attributes should be set to "NULL" or "0".
472  */
473 typedef struct ibt_chan_open_args_s {
474 	ibt_path_info_t 	*oc_path;	/* Primary & Alternate */
475 	ibt_cm_handler_t 	oc_cm_handler;	/* cm_handler - required */
476 	void			*oc_cm_clnt_private;	/* First argument to */
477 							/* cm_handler */
478 	ibt_rnr_retry_cnt_t	oc_path_rnr_retry_cnt;
479 	uint8_t			oc_path_retry_cnt:3;
480 	uint8_t			oc_rdma_ra_out;
481 	uint8_t			oc_rdma_ra_in;
482 	ibt_priv_data_len_t	oc_priv_data_len;	/* Number of bytes of */
483 							/* REQ Private data */
484 	void			*oc_priv_data;		/* REQ private data */
485 	ibt_channel_hdl_t	oc_dup_channel; 	/* IBT_OCHAN_DUP */
486 	ibt_redirect_info_t	*oc_cm_redirect_info;	/* Redirect params */
487 							/* for port and CM */
488 							/* redirection */
489 	ibt_cep_path_t		*oc_cm_cep_path;	/* Optional Path for */
490 							/* CM MADs on */
491 							/* port redirection */
492 	ib_time_t		oc_cm_pkt_lt;		/* Pkt life time for */
493 							/* CM MADs */
494 	uint32_t		oc_opaque1:4;
495 	uint32_t		oc_opaque2:24;
496 	uint32_t		oc_opaque3;
497 	uint32_t		oc_opaque4;
498 } ibt_chan_open_args_t;
499 
500 
501 /*
502  * Define an optional RC return arguments structure. This contains return
503  * parameters from ibt_open_rc_channel() when called in BLOCKING mode.
504  *
505  * rc_priv_data should be NULL or point to a buffer allocated by the caller,
506  * the size of which should be in rc_priv_data_len, where rc_priv_data_len <=
507  * IBT_REP_PRIV_DATA_SZ.
508  */
509 typedef struct ibt_rc_returns_s {
510 	uint8_t			rc_rdma_ra_in;	/* Arbitrated resp resources */
511 	uint8_t			rc_rdma_ra_out;	/* Arbitrated initiator depth */
512 	ibt_arej_info_t		rc_arej_info;
513 	ibt_cm_reason_t		rc_status;
514 	uint8_t			rc_failover_status;	/* Failover status */
515 	ibt_priv_data_len_t	rc_priv_data_len;
516 	void			*rc_priv_data;
517 } ibt_rc_returns_t;
518 
519 /*
520  * Define a callback function that can be used in Non-Blocking calls to
521  * ibt_recycle_rc().
522  */
523 
524 typedef	void	(*ibt_recycle_handler_t)(ibt_status_t ibt_status, void *arg);
525 
526 /*
527  * Define an optional return arguments structure from ibt_set_alt_path()
528  * This contains return parameters, when called in BLOCKING mode.
529  *
530  * ap_priv_data should be NULL or point to a buffer allocated by the caller,
531  * the size of which should be in ap_priv_data_len, where ap_priv_data_len <=
532  * IBT_APR_PRIV_DATA_SZ.
533  * The private data from APR is returned in ap_priv_data.
534  * The caller specifies amount of APR private data to be returned in
535  * ap_priv_data_len.
536  */
537 typedef struct ibt_ap_returns_s {
538 	ibt_ap_status_t		ap_status;
539 	boolean_t		ap_arej_info_valid;
540 	ibt_arej_info_t		ap_arej_info;	/* Only valid if redirect */
541 	ibt_priv_data_len_t	ap_priv_data_len;
542 	void			*ap_priv_data;
543 } ibt_ap_returns_t;
544 
545 /*
546  * UD remote destination attributes.
547  *
548  * ud_sid, ud_addr, ud_pkt_lt and ud_pkey_ix must be specified.
549  * These values can be as returned in an ibt_path_info_t struct from an
550  * ibt_get_paths() call.
551  *
552  * ud_priv_data should be NULL or point to a buffer allocated by the caller,
553  * the size of which is in ud_priv_data_len, where ud_priv_data_len <=
554  * IBT_SIDR_REQ_PRIV_DATA_SZ.
555  */
556 typedef struct ibt_ud_dest_attr_s {
557 	ib_svc_id_t		ud_sid;		/* Service ID */
558 	ibt_adds_vect_t		*ud_addr;	/* Address Info */
559 	uint16_t		ud_pkey_ix;	/* Pkey Index */
560 	ib_time_t		ud_pkt_lt;
561 	ibt_cm_ud_handler_t	ud_cm_handler;	/* An optional CM UD event */
562 						/* which must be NULL */
563 						/* if not specified. */
564 	void			*ud_cm_private; /* First arg to ud_cm_handler */
565 	ibt_priv_data_len_t	ud_priv_data_len;
566 	void			*ud_priv_data;	/* SIDR REQ private data */
567 } ibt_ud_dest_attr_t;
568 
569 /*
570  * Define an optional UD return arguments structure.
571  *
572  * ud_priv_data should be NULL or point to a buffer allocated by the caller,
573  * the size of which should be in ud_priv_data_len, where ud_priv_data_len <=
574  * IBT_SIDR_REP_PRIV_DATA_SZ.
575  */
576 typedef struct ibt_ud_returns_s {
577 	ibt_sidr_status_t	ud_status;
578 	ibt_redirect_info_t	ud_redirect;
579 	ib_qpn_t		ud_dqpn;	/* Returned destination QPN */
580 	ib_qkey_t		ud_qkey;	/* Q_Key for destination QPN */
581 	ibt_priv_data_len_t	ud_priv_data_len;
582 	void			*ud_priv_data;
583 } ibt_ud_returns_t;
584 
585 /*
586  * Multicast group attributes
587  * Not specified attributes should be set to "NULL" or "0".
588  * Used by ibt_join_mcg()/ibt_query_mcg().
589  *
590  * mc_qkey, mc_pkey, mc_flow, mc_tclass, mc_sl, mc_join_state are required for
591  * create - ibt_join_mcg().
592  */
593 typedef struct ibt_mcg_attr_s {
594 	ib_gid_t		mc_mgid;	/* MGID */
595 	ib_gid_t		mc_pgid;	/* SGID of the end port being */
596 						/* added to the MCG. */
597 	ib_qkey_t		mc_qkey;	/* Q_Key */
598 	ib_pkey_t		mc_pkey;	/* Partition key for this MCG */
599 	ibt_mtu_req_t		mc_mtu_req;	/* MTU */
600 	ibt_srate_req_t		mc_rate_req;	/* Static rate */
601 	ibt_pkt_lt_req_t	mc_pkt_lt_req;	/* Packet Life Time Request */
602 	uint_t			mc_flow:20;	/* FlowLabel. */
603 	uint8_t			mc_hop;		/* HopLimit */
604 	uint8_t			mc_tclass;	/* Traffic Class. */
605 	uint8_t			mc_sl:4;	/* Service Level */
606 	uint8_t			mc_scope:4,	/* Multicast Address Scope */
607 				mc_join_state:4; /* FULL For create */
608 	ib_lid_t		mc_opaque1;
609 } ibt_mcg_attr_t;
610 
611 /*
612  * Multicast group attributes.
613  * returned by ibt_join_mcg()/ibt_query_mcg().
614  */
615 typedef struct ibt_mcg_info_s {
616 	ibt_adds_vect_t		mc_adds_vect;   /* Address information */
617 	ib_mtu_t		mc_mtu;		/* MTU */
618 	ib_qkey_t		mc_qkey;	/* Q_Key */
619 	uint16_t		mc_pkey_ix;	/* Pkey Index */
620 	uint8_t			mc_scope:4;	/* Multicast Address Scope */
621 	clock_t			mc_opaque2;
622 } ibt_mcg_info_t;
623 
624 /*
625  * Define a callback function that can be used in Non-Blocking calls to
626  * ibt_join_mcg().
627  */
628 typedef void (*ibt_mcg_handler_t)(void *arg, ibt_status_t retval,
629     ibt_mcg_info_t *mcg_info_p);
630 
631 
632 /*
633  * Service Flags - sd_flags
634  *
635  *    IBT_SRV_PEER_TYPE_SID	Peer-to-peer Service IDs.
636  */
637 
638 typedef enum ibt_service_flags_e {
639 	IBT_SRV_NO_FLAGS	= 0x0,
640 	IBT_SRV_PEER_TYPE_SID	= 0x1
641 } ibt_service_flags_t;
642 
643 /*
644  * Define a Service ID Registration structure.
645  */
646 typedef struct ibt_srv_desc_s {
647 	ibt_cm_ud_handler_t	sd_ud_handler;	/* UD Service Handler */
648 	ibt_cm_handler_t	sd_handler;	/* Non-UD Service Handler */
649 	ibt_service_flags_t	sd_flags;	/* Flags */
650 } ibt_srv_desc_t;
651 
652 /*
653  * Flag to indicate ibt_bind_service() to or NOT-to clean-up Stale matching
654  * Local Service Records with SA prior to binding the new request.
655  */
656 #define	IBT_SBIND_NO_FLAGS	0
657 #define	IBT_SBIND_NO_CLEANUP	1
658 
659 /*
660  * Define a Service ID Binding structure (data for service records).
661  */
662 typedef struct ibt_srv_bind_s {
663 	uint64_t	sb_key[2];	/* Service Key */
664 	char		*sb_name;	/* Service Name (up to 63 chars) */
665 	uint32_t	sb_lease;	/* Service Lease period (in seconds) */
666 	ib_pkey_t	sb_pkey;	/* Service P_Key */
667 	ibt_srv_data_t	sb_data;	/* Service Data */
668 	uint_t		sb_flag;	/* indicates to/not-to clean-up stale */
669 					/* matching local service records. */
670 } ibt_srv_bind_t;
671 
672 /*
673  * ibt_cm_delay() flags.
674  *
675  * Refer to InfiniBand Architecture Release Volume 1 Rev 1.0a:
676  * Section 12.6.6 MRA
677  */
678 typedef enum ibt_cmdelay_flags_e {
679 	IBT_CM_DELAY_REQ	= 0,
680 	IBT_CM_DELAY_REP	= 1,
681 	IBT_CM_DELAY_LAP	= 2
682 } ibt_cmdelay_flags_t;
683 
684 /*
685  * The payload for DDI events passed on IB_PROP_UPDATE_EVENT.
686  * This is passed as the bus nexus data to event_handler(9e).
687  */
688 typedef struct ibt_prop_update_payload_s {
689 	union {
690 		struct {
691 			uint32_t	srv_updated:1;
692 			uint32_t	gid_updated:1;
693 		} _ib_prop_update_struct;
694 		uint32_t	prop_updated;
695 	} _ib_prop_update_union;
696 	ibt_status_t		ib_reprobe_status;
697 
698 #define	ib_srv_prop_updated	\
699     _ib_prop_update_union._ib_prop_update_struct.srv_updated
700 #define	ib_gid_prop_updated	\
701     _ib_prop_update_union._ib_prop_update_struct.gid_updated
702 #define	ib_prop_updated		\
703     _ib_prop_update_union.prop_updated
704 } ibt_prop_update_payload_t;
705 
706 
707 /*
708  * FUNCTION PROTOTYPES.
709  */
710 
711 /*
712  * ibt_attach() and ibt_detach():
713  *	These are the calls into IBTF used during client driver attach() and
714  *	detach().
715  *
716  *	The IBTF returns an ibt_clnt_hdl_t to the client. This handle is used
717  *	to identify this client device in all subsequent calls into the IBTF.
718  *
719  *	The ibt_detach() routine is called from a client driver's detach()
720  *	routine to deregister itself from the IBTF.
721  */
722 ibt_status_t ibt_attach(ibt_clnt_modinfo_t *mod_infop, dev_info_t *arg,
723     void *clnt_private, ibt_clnt_hdl_t *ibt_hdl_p);
724 
725 ibt_status_t ibt_detach(ibt_clnt_hdl_t ibt_hdl);
726 
727 /*
728  * HCA FUNCTIONS
729  */
730 
731 /*
732  * ibt_get_hca_list()
733  *	Returns the number of HCAs in a system and their node GUIDS.
734  *
735  *	If hca_list_p is not NULL then the memory for the array of GUIDs is
736  *	allocated by the IBTF and should be freed by the caller using
737  *	ibt_free_hca_list(). If hca_list_p is NULL then no memory is allocated
738  *	by ibt_get_hca_list and only the number of HCAs in a system is returned.
739  *
740  *	It is assumed that the caller can block in kmem_alloc.
741  *
742  * ibt_free_hca_list()
743  *	Free the memory allocated by ibt_get_hca_list().
744  */
745 uint_t ibt_get_hca_list(ib_guid_t **hca_list_p);
746 
747 void ibt_free_hca_list(ib_guid_t *hca_list, uint_t entries);
748 
749 /*
750  * ibt_open_hca()	- Open/Close a HCA. HCA can only be opened/closed
751  * ibt_close_hca()	  once. ibt_open_hca() takes a client's ibt handle
752  *			  and a GUID and returns a unique IBT client HCA
753  *			  handle.
754  *
755  * These routines can not be called from interrupt context.
756  */
757 ibt_status_t ibt_open_hca(ibt_clnt_hdl_t ibt_hdl, ib_guid_t hca_guid,
758     ibt_hca_hdl_t *hca_hdl);
759 
760 ibt_status_t ibt_close_hca(ibt_hca_hdl_t hca_hdl);
761 
762 
763 /*
764  * ibt_query_hca()
765  * ibt_query_hca_byguid()
766  * 	Returns the static attributes of the specified HCA
767  */
768 ibt_status_t ibt_query_hca(ibt_hca_hdl_t hca_hdl, ibt_hca_attr_t *hca_attrs);
769 
770 ibt_status_t ibt_query_hca_byguid(ib_guid_t hca_guid,
771     ibt_hca_attr_t *hca_attrs);
772 
773 
774 /*
775  * ibt_query_hca_ports()
776  * ibt_query_hca_ports_byguid()
777  *	Returns HCA port/ports attributes for the specified HCA port/ports.
778  *	ibt_query_hca_ports() allocates the memory required for the
779  *	ibt_hca_portinfo_t struct as well as the memory required for the SGID
780  *	and P_Key tables contained within that struct.
781  *
782  * ibt_free_portinfo()
783  *	Frees the memory allocated for a specified ibt_hca_portinfo_t struct.
784  */
785 ibt_status_t ibt_query_hca_ports(ibt_hca_hdl_t hca_hdl, uint8_t port,
786     ibt_hca_portinfo_t **port_info_p, uint_t *ports_p, uint_t *size_p);
787 
788 ibt_status_t ibt_query_hca_ports_byguid(ib_guid_t hca_guid, uint8_t port,
789     ibt_hca_portinfo_t **port_info_p, uint_t *ports_p, uint_t *size_p);
790 
791 void ibt_free_portinfo(ibt_hca_portinfo_t *port_info, uint_t size);
792 
793 /*
794  * ibt_set_hca_private()	- Set/get the client private data.
795  * ibt_get_hca_private()
796  */
797 void ibt_set_hca_private(ibt_hca_hdl_t hca_hdl, void *clnt_private);
798 void *ibt_get_hca_private(ibt_hca_hdl_t hca_hdl);
799 
800 /*
801  * ibt_hca_handle_to_guid()
802  *	A helper function to retrieve HCA GUID for the specified handle.
803  *	Returns HCA GUID on which the specified Channel is allocated. Valid
804  *	if it is non-NULL on return.
805  */
806 ib_guid_t ibt_hca_handle_to_guid(ibt_hca_hdl_t hca);
807 
808 /*
809  * ibt_hca_guid_to_handle()
810  *	A helper function to retrieve a hca handle from a HCA GUID.
811  */
812 ibt_status_t ibt_hca_guid_to_handle(ibt_clnt_hdl_t ibt_hdl, ib_guid_t hca_guid,
813     ibt_hca_hdl_t *hca_hdl);
814 
815 /*
816  * CONNECTION ESTABLISHMENT/TEAR DOWN FUNCTIONS.
817  */
818 
819 /*
820  * ibt_get_paths
821  *	Finds the best path to a specified destination (as determined by the
822  *	IBTL) that satisfies the requirements specified in an ibt_path_attr_t
823  *	struct.
824  */
825 ibt_status_t ibt_get_paths(ibt_clnt_hdl_t ibt_hdl, ibt_path_flags_t flags,
826     ibt_path_attr_t *attr, uint8_t max_paths, ibt_path_info_t *paths,
827     uint8_t *num_paths_p);
828 
829 
830 /*
831  * Callback function that can be used in ibt_aget_paths(), a Non-Blocking
832  * version of ibt_get_paths().
833  */
834 typedef void (*ibt_path_handler_t)(void *arg, ibt_status_t retval,
835     ibt_path_info_t *paths, uint8_t num_paths);
836 
837 /*
838  * Find path(s) to a given destination or service asynchronously.
839  * ibt_aget_paths() is a Non-Blocking version of ibt_get_paths().
840  */
841 ibt_status_t ibt_aget_paths(ibt_clnt_hdl_t ibt_hdl, ibt_path_flags_t flags,
842     ibt_path_attr_t *attr, uint8_t max_paths, ibt_path_handler_t func,
843     void  *arg);
844 
845 /*
846  * ibt_get_alt_path
847  *	Finds the best alternate path to a specified channel (as determined by
848  *	the IBTL) that satisfies the requirements specified in an
849  *	ibt_alt_path_attr_t struct.  The specified channel must have been
850  *	previously opened successfully using ibt_open_rc_channel.
851  */
852 ibt_status_t ibt_get_alt_path(ibt_channel_hdl_t chan, ibt_path_flags_t flags,
853     ibt_alt_path_attr_t *attr, ibt_alt_path_info_t *alt_path);
854 
855 /*
856  * ibt_open_rc_channel
857  * 	ibt_open_rc_channel() opens a previously allocated RC communication
858  *	channel. The IBTL initiates the channel establishment protocol.
859  */
860 ibt_status_t ibt_open_rc_channel(ibt_channel_hdl_t rc_chan,
861     ibt_chan_open_flags_t flags, ibt_execution_mode_t mode,
862     ibt_chan_open_args_t *args, ibt_rc_returns_t *returns);
863 
864 /*
865  * ibt_close_rc_channel
866  *	Close the specified channel. Outstanding work requests are flushed
867  *	so that the client can do the associated clean up. After that, the
868  *	client will usually deregister the previously registered memory,
869  *	then free the channel by calling ibt_free_rc_channel().
870  *
871  *	This function will reuse CM event Handler provided in
872  *	ibt_open_rc_channel().
873  */
874 ibt_status_t ibt_close_rc_channel(ibt_channel_hdl_t rc_chan,
875     ibt_execution_mode_t mode, void *priv_data,
876     ibt_priv_data_len_t priv_data_len, uint8_t *ret_status,
877     void *ret_priv_data, ibt_priv_data_len_t *ret_priv_data_len_p);
878 
879 /*
880  * ibt_prime_close_rc_channel
881  *
882  *	Allocates resources required for a close rc channel operation.
883  *	Calling ibt_prime_close_rc_channel() allows a channel to be
884  *	subsequently closed in interrupt context.
885  *
886  *	A call is first made to ibt_prime_close_rc_channel in non-interrupt
887  *	context, followed by ibt_close_rc_channel in non-blocking mode from
888  *	interrupt context
889  *
890  *	ibt_prime_close_rc_channel() can only be called on a previously opened
891  *	channel.
892  */
893 ibt_status_t ibt_prime_close_rc_channel(ibt_channel_hdl_t rc_chan);
894 
895 /*
896  * ibt_recycle_rc
897  *
898  *      Recycle a RC channel which has transitioned to Error state. The
899  *      ibt_recycle_rc() function transitions the channel from Error
900  *      state (IBT_STATE_ERROR) to the state ready for use by
901  *      ibt_open_rc_channel. Basically, this function is very similar to
902  *      ibt_alloc_rc_channel, but reuses an existing RC channel.
903  *
904  * Clients are allowed to make resource clean up/free calls in the CM handler
905  *
906  * Client(s) must not invoke blocking version (ie., func specified as NULL) of
907  * ibt_recycle_rc from cm callback for IBT_CM_EVENT_CONN_CLOSED
908  *
909  * Clients are strongly advised not to issue blocking calls from  func, as this
910  * would block the CM threads, and could delay or block other client connections
911  * and ibtl related API invocations.
912  */
913 ibt_status_t ibt_recycle_rc(ibt_channel_hdl_t rc_chan, ibt_cep_flags_t control,
914     uint8_t hca_port_num, ibt_recycle_handler_t func, void *arg);
915 
916 /*
917  * ibt_recycle_ud
918  *
919  *      Recycle a UD channel which has transitioned to Error state. The
920  *      ibt_recycle_ud() function transitions the channel from Error
921  *      state (IBT_STATE_ERROR) to a usable state (IBT_STATE_RTS).
922  *      Basically, this function is very similar to ibt_alloc_ud_channel,
923  *	but reuses an existing UD channel.
924  */
925 ibt_status_t ibt_recycle_ud(ibt_channel_hdl_t ud_chan, uint8_t hca_port_num,
926     uint16_t pkey_ix, ib_qkey_t qkey);
927 
928 /*
929  * MODIFY CHANNEL ATTRIBUTE FUNCTIONs.
930  */
931 
932 /*
933  * ibt_pause_sendq
934  * ibt_unpause_sendq
935  *	Place the send queue of the specified channel into the send queue
936  *	drained state.
937  *	Applicable for both RC and UD channels.
938  */
939 ibt_status_t ibt_pause_sendq(ibt_channel_hdl_t chan,
940     ibt_cep_modify_flags_t modify_flags);
941 
942 ibt_status_t ibt_unpause_sendq(ibt_channel_hdl_t chan);
943 
944 /*
945  * ibt_resize_queues()
946  *	Resize the SendQ/RecvQ sizes of a channel.
947  *
948  *	Applicable for both RC and UD channels.
949  */
950 ibt_status_t ibt_resize_queues(ibt_channel_hdl_t chan, ibt_qflags_t flags,
951     ibt_queue_sizes_t *request_sz, ibt_queue_sizes_t *actual_sz);
952 
953 /*
954  * ibt_query_queues()
955  *
956  *	Query the SendQ/RecvQ sizes of a channel.
957  *	Applicable for both RC and UD channels.
958  */
959 ibt_status_t ibt_query_queues(ibt_channel_hdl_t chan,
960     ibt_queue_sizes_t *actual_sz);
961 
962 /*
963  * ibt_modify_rdma
964  *	Enable/disable RDMA operations.
965  *
966  *	Applicable for RC channels only.
967  */
968 ibt_status_t ibt_modify_rdma(ibt_channel_hdl_t rc_chan,
969     ibt_cep_modify_flags_t modify_flags, ibt_cep_flags_t flags);
970 
971 
972 /*
973  * ibt_set_rdma_resource
974  *	Change the number of resources to be used for incoming and outgoing
975  *	RDMA reads & Atomics.
976  */
977 ibt_status_t ibt_set_rdma_resource(ibt_channel_hdl_t rc_chan,
978     ibt_cep_modify_flags_t modify_flags, uint8_t rdma_ra_out,
979     uint8_t rdma_ra_in);
980 
981 /*
982  * ibt_change_port
983  *	Change the primary physical port of an RC channel. (This is done only
984  *	if HCA supports this capability).  Can only be called on a paused
985  *	channel.
986  *	Applicable for RC channels only.
987  */
988 ibt_status_t ibt_change_port(ibt_channel_hdl_t rc_chan, uint8_t port_num);
989 
990 
991 /*
992  * SERVICE REGISTRATION FUNCTIONS
993  */
994 
995 /*
996  * ibt_register_service()
997  * ibt_deregister_service()
998  *	Register/deregister a Service (range of Service IDs) with the IBTF.
999  *
1000  * ibt_bind_service()
1001  * ibt_unbind_service()
1002  * ibt_unbind_all_services()
1003  *	Bind a Service to a given port (GID), and optionally create
1004  *	service record(s) with the SA for ibt_get_paths() to find.
1005  */
1006 ibt_status_t ibt_register_service(ibt_clnt_hdl_t ibt_hdl,
1007     ibt_srv_desc_t *service, ib_svc_id_t sid, int num_sids,
1008     ibt_srv_hdl_t *srv_hdl_p, ib_svc_id_t *ret_sid_p);
1009 
1010 ibt_status_t ibt_deregister_service(ibt_clnt_hdl_t ibt_hdl,
1011     ibt_srv_hdl_t srv_hdl);
1012 
1013 ibt_status_t ibt_bind_service(ibt_srv_hdl_t srv_hdl, ib_gid_t gid,
1014     ibt_srv_bind_t *srv_bind, void *cm_private, ibt_sbind_hdl_t *sb_hdl_p);
1015 
1016 ibt_status_t ibt_unbind_service(ibt_srv_hdl_t srv_hdl, ibt_sbind_hdl_t sb_hdl);
1017 ibt_status_t ibt_unbind_all_services(ibt_srv_hdl_t srv_hdl);
1018 
1019 /*
1020  * ibt_cm_delay
1021  *	A client CM handler/srv_handler function can call this function to
1022  *	extend its response time to a CM event.
1023  *	Applicable for RC channels only.
1024  */
1025 ibt_status_t ibt_cm_delay(ibt_cmdelay_flags_t flags, void *cm_session_id,
1026     clock_t service_time, void *priv_data, ibt_priv_data_len_t priv_data_len);
1027 
1028 /*
1029  * ibt_cm_proceed
1030  *
1031  * An IBT client calls ibt_cm_proceed() to proceed with a connection that
1032  * previously deferred by the client returning IBT_CM_DEFER on a CM handler
1033  * callback. CM events that can be deferred and continued with ibt_cm_proceed()
1034  * are REQ_RCV, REP_RCV, LAP_RCV, and DREQ_RCV.
1035  *
1036  * NOTE :
1037  *
1038  * Typically CM completes processing of a client's CM handler return, with
1039  * IBT_CM_DEFER status, before  processing of the corresponding ibt_cm_proceed()
1040  * is started. However a race exists where by CM may not have completed the
1041  * client's handler return processing when ibt_cm_proceed() is called by a
1042  * client. In this case ibt_cm_proceed() will block until processing of the
1043  * client's CM handler return is complete.
1044  *
1045  * A client that returns IBT_CM_DEFER from the cm handler must
1046  * subsequently make a call to ibt_cm_proceed(). It is illegal to call
1047  * ibt_cm_proceed() on a channel that has not had the connection
1048  * establishment deferred.
1049  *
1050  * Client cannot call ibt_cm_proceed from the cm handler.
1051  */
1052 ibt_status_t ibt_cm_proceed(ibt_cm_event_type_t event, void *session_id,
1053     ibt_cm_status_t status, ibt_cm_proceed_reply_t *cm_event_data,
1054     void *priv_data, ibt_priv_data_len_t priv_data_len);
1055 
1056 /*
1057  * ibt_cm_ud_proceed
1058  *
1059  * An IBT client calls ibt_cm_ud_proceed() to proceed with an
1060  * IBT_CM_UD_EVENT_SIDR_REQ  UD event that was previously deferred by the
1061  * client returning IBT_CM_DEFER on a CM UD handler callback.
1062  * NOTE :
1063  *
1064  * Typically CM completes processing of a client's CM handler return, with
1065  * IBT_CM_DEFER status, before  processing of the corresponding
1066  * ibt_cm_ud_proceed() is started. However a race exists where by CM may not
1067  * have completed the client's handler return processing when
1068  * ibt_cm_ud_proceed() is called by a client. In this case ibt_cm_ud_proceed()
1069  * will block until processing of the client's CM handler return is complete.
1070  *
1071  * A client that returns IBT_CM_DEFER from the cm handler must
1072  * subsequently make a call to ibt_cm_ud_proceed(). It is illegal to call
1073  * ibt_cm_ud_proceed() on a channel that has not had the connection
1074  * establishment deferred.
1075  *
1076  * Client cannot call ibt_cm_ud_proceed from the cm handler.
1077  */
1078 ibt_status_t ibt_cm_ud_proceed(void *session_id, ibt_channel_hdl_t ud_channel,
1079     ibt_cm_status_t status, ibt_redirect_info_t *redirect_infop,
1080     void *priv_data, ibt_priv_data_len_t priv_data_len);
1081 
1082 
1083 /*
1084  * COMPLETION QUEUES.
1085  *
1086  * ibt_alloc_cq_sched()
1087  *	Reserve CQ scheduling class resources
1088  *
1089  * ibt_free_cq_sched()
1090  *	Free CQ scheduling class resources
1091  */
1092 ibt_status_t ibt_alloc_cq_sched(ibt_hca_hdl_t hca_hdl,
1093     ibt_cq_sched_attr_t *attr, ibt_sched_hdl_t *sched_hdl_p);
1094 
1095 ibt_status_t ibt_free_cq_sched(ibt_hca_hdl_t hca_hdl,
1096     ibt_sched_hdl_t sched_hdl, uint_t load);
1097 
1098 /*
1099  * ibt_alloc_cq()
1100  *	Allocate a completion queue.
1101  */
1102 ibt_status_t ibt_alloc_cq(ibt_hca_hdl_t hca_hdl, ibt_cq_attr_t *cq_attr,
1103     ibt_cq_hdl_t *ibt_cq_p, uint_t *real_size);
1104 
1105 /*
1106  * ibt_free_cq()
1107  *	Free allocated CQ resources.
1108  */
1109 ibt_status_t ibt_free_cq(ibt_cq_hdl_t ibt_cq);
1110 
1111 
1112 /*
1113  * ibt_enable_cq_notify()
1114  *	Enable notification requests on the specified CQ.
1115  *	Applicable for both RC and UD channels.
1116  *
1117  *	Completion notifications are disabled by setting the completion
1118  *	handler to NULL by calling ibt_set_cq_handler().
1119  */
1120 ibt_status_t ibt_enable_cq_notify(ibt_cq_hdl_t ibt_cq,
1121     ibt_cq_notify_flags_t notify_type);
1122 
1123 /*
1124  * ibt_set_cq_handler()
1125  *	Register a work request completion handler with the IBTF.
1126  *	Applicable for both RC and UD channels.
1127  *
1128  *	Completion notifications are disabled by setting the completion
1129  *	handler to NULL. When setting the handler to NULL, no additional
1130  *	calls to the CQ handler will be initiated.
1131  *
1132  *	This function does not otherwise change the state of previous
1133  *	calls to ibt_enable_cq_notify().
1134  */
1135 void ibt_set_cq_handler(ibt_cq_hdl_t ibt_cq,
1136     ibt_cq_handler_t completion_handler, void *arg);
1137 
1138 /*
1139  * ibt_poll_cq()
1140  *	Poll the specified CQ for the completion of work requests (WRs).
1141  *	If the CQ contains completed WRs, up to num_wc of them are returned.
1142  *	Applicable for both RC and UD channels.
1143  */
1144 ibt_status_t ibt_poll_cq(ibt_cq_hdl_t ibt_cq, ibt_wc_t *work_completions,
1145     uint_t num_wc, uint_t *num_polled);
1146 
1147 /*
1148  * ibt_query_cq()
1149  *	Return the total number of entries in the CQ.
1150  */
1151 ibt_status_t ibt_query_cq(ibt_cq_hdl_t ibt_cq, uint_t *entries,
1152     uint_t *count_p, uint_t *usec_p, ibt_cq_handler_id_t *hid_p);
1153 
1154 /*
1155  * ibt_resize_cq()
1156  *	Change the size of a CQ.
1157  */
1158 ibt_status_t ibt_resize_cq(ibt_cq_hdl_t ibt_cq, uint_t new_sz, uint_t *real_sz);
1159 
1160 /*
1161  * ibt_modify_cq()
1162  *	Change the interrupt moderation values of a CQ.
1163  *	"count" is number of completions before interrupting.
1164  *	"usec" is the number of microseconds before interrupting.
1165  */
1166 ibt_status_t ibt_modify_cq(ibt_cq_hdl_t ibt_cq, uint_t count, uint_t usec,
1167     ibt_cq_handler_id_t hid);
1168 
1169 /*
1170  * ibt_set_cq_private()
1171  * ibt_get_cq_private()
1172  *	Set/get the client private data.
1173  */
1174 void ibt_set_cq_private(ibt_cq_hdl_t ibt_cq, void *clnt_private);
1175 void *ibt_get_cq_private(ibt_cq_hdl_t ibt_cq);
1176 
1177 
1178 /*
1179  * Memory Management Functions.
1180  *	Applicable for both RC and UD channels.
1181  *
1182  * ibt_register_mr()
1183  * 	Prepares a virtually addressed memory region for use by a HCA. A
1184  *	description of the registered memory suitable for use in Work Requests
1185  *	(WRs) is returned in the ibt_mr_desc_t parameter.
1186  *
1187  * ibt_register_buf()
1188  * 	Prepares a memory region described by a buf(9S) struct for use by a
1189  *	HCA. A description of the registered memory suitable for use in
1190  *	Work Requests (WRs) is returned in the ibt_mr_desc_t parameter.
1191  *
1192  * ibt_query_mr()
1193  *	Retrieves information about a specified memory region.
1194  *
1195  * ibt_deregister_mr()
1196  *	Remove a memory region from a HCA translation table, and free all
1197  *	resources associated with the memory region.
1198  *
1199  * ibt_reregister_mr()
1200  * ibt_reregister_buf()
1201  *	Modify the attributes of an existing memory region.
1202  *
1203  * ibt_register_shared_mr()
1204  *	Given an existing memory region, a new memory region associated with
1205  *	the same physical locations is created.
1206  *
1207  * ibt_sync_mr()
1208  *	Sync a memory region for either RDMA reads or RDMA writes
1209  *
1210  * ibt_alloc_mw()
1211  *	Allocate a memory window.
1212  *
1213  * ibt_query_mw()
1214  *	Retrieves information about a specified memory window.
1215  *
1216  * ibt_free_mw()
1217  *	De-allocate the Memory Window.
1218  */
1219 ibt_status_t ibt_register_mr(ibt_hca_hdl_t hca_hdl, ibt_pd_hdl_t pd,
1220     ibt_mr_attr_t *mem_attr, ibt_mr_hdl_t *mr_hdl_p, ibt_mr_desc_t *mem_desc);
1221 
1222 ibt_status_t ibt_register_buf(ibt_hca_hdl_t hca_hdl, ibt_pd_hdl_t pd,
1223     ibt_smr_attr_t *mem_bpattr, struct buf *bp, ibt_mr_hdl_t *mr_hdl_p,
1224     ibt_mr_desc_t *mem_desc);
1225 
1226 ibt_status_t ibt_query_mr(ibt_hca_hdl_t hca_hdl, ibt_mr_hdl_t mr_hdl,
1227     ibt_mr_query_attr_t *attr);
1228 
1229 ibt_status_t ibt_deregister_mr(ibt_hca_hdl_t hca_hdl, ibt_mr_hdl_t mr_hdl);
1230 
1231 ibt_status_t ibt_reregister_mr(ibt_hca_hdl_t hca_hdl, ibt_mr_hdl_t mr_hdl,
1232     ibt_pd_hdl_t pd, ibt_mr_attr_t *mem_attr, ibt_mr_hdl_t *mr_hdl_p,
1233     ibt_mr_desc_t *mem_desc);
1234 
1235 ibt_status_t ibt_reregister_buf(ibt_hca_hdl_t hca_hdl, ibt_mr_hdl_t mr_hdl,
1236     ibt_pd_hdl_t pd, ibt_smr_attr_t *mem_bpattr, struct buf *bp,
1237     ibt_mr_hdl_t *mr_hdl_p, ibt_mr_desc_t *mem_desc);
1238 
1239 ibt_status_t ibt_register_shared_mr(ibt_hca_hdl_t hca_hdl, ibt_mr_hdl_t mr_hdl,
1240     ibt_pd_hdl_t pd, ibt_smr_attr_t *mem_sattr, ibt_mr_hdl_t *mr_hdl_p,
1241     ibt_mr_desc_t *mem_desc);
1242 
1243 ibt_status_t ibt_sync_mr(ibt_hca_hdl_t hca_hdl, ibt_mr_sync_t *mr_segments,
1244     size_t num_segments);
1245 
1246 ibt_status_t ibt_alloc_mw(ibt_hca_hdl_t hca_hdl, ibt_pd_hdl_t pd,
1247     ibt_mw_flags_t flags, ibt_mw_hdl_t *mw_hdl_p, ibt_rkey_t *rkey);
1248 
1249 ibt_status_t ibt_query_mw(ibt_hca_hdl_t hca_hdl, ibt_mw_hdl_t mw_hdl,
1250     ibt_mw_query_attr_t *mw_attr_p);
1251 
1252 ibt_status_t ibt_free_mw(ibt_hca_hdl_t hca_hdl, ibt_mw_hdl_t mw_hdl);
1253 
1254 /*
1255  * ibt_alloc_lkey()
1256  * 	Allocates physical buffer list resources for use in memory
1257  *	registrations.
1258  *
1259  *	Applicable for both RC and UD channels.
1260  */
1261 ibt_status_t ibt_alloc_lkey(ibt_hca_hdl_t hca_hdl, ibt_pd_hdl_t pd,
1262     ibt_lkey_flags_t flags, uint_t phys_buf_list_sz, ibt_mr_hdl_t *mr_p,
1263     ibt_pmr_desc_t *mem_desc_p);
1264 
1265 
1266 /*
1267  * Physical Memory Management Functions.
1268  *	Applicable for both RC and UD channels.
1269  *
1270  * ibt_register_phys_mr()
1271  *	Prepares a physically addressed memory region for use by a HCA.
1272  *
1273  * ibt_reregister_phys_mr()
1274  *	Modify the attributes of an existing memory region.
1275  */
1276 ibt_status_t ibt_register_phys_mr(ibt_hca_hdl_t hca_hdl, ibt_pd_hdl_t pd,
1277     ibt_pmr_attr_t *mem_pattr, ibt_mr_hdl_t *mr_hdl_p,
1278     ibt_pmr_desc_t *mem_desc_p);
1279 
1280 ibt_status_t ibt_reregister_phys_mr(ibt_hca_hdl_t hca_hdl, ibt_mr_hdl_t mr_hdl,
1281     ibt_pd_hdl_t pd, ibt_pmr_attr_t *mem_pattr, ibt_mr_hdl_t *mr_hdl_p,
1282     ibt_pmr_desc_t *mem_desc_p);
1283 
1284 
1285 /*
1286  * Address Translation.
1287  */
1288 
1289 /*
1290  * ibt_map_mem_area()
1291  *	Translate a kernel virtual address range into HCA physical addresses.
1292  *	A set of physical addresses, that can be used with "Reserved L_Key",
1293  *	register physical,  and "Fast Registration Work Request" operations
1294  *	is returned.
1295  */
1296 ibt_status_t ibt_map_mem_area(ibt_hca_hdl_t hca_hdl, ibt_va_attr_t *va_attrs,
1297     uint_t paddr_list_len, ibt_phys_buf_t *paddr_list_p, uint_t *num_paddr_p,
1298     size_t *paddr_bufsz_p, ib_memlen_t *paddr_offset_p, ibt_ma_hdl_t *ma_hdl_p);
1299 
1300 /*
1301  * ibt_unmap_mem_area()
1302  *	Un pin physical pages pinned during an ibt_map_mem_area() call.
1303  */
1304 ibt_status_t ibt_unmap_mem_area(ibt_hca_hdl_t hca_hdl, ibt_ma_hdl_t ma_hdl);
1305 
1306 /* ibt_map_mem_iov() */
1307 ibt_status_t ibt_map_mem_iov(ibt_hca_hdl_t hca_hdl,
1308     ibt_iov_attr_t *iov_attr, ibt_all_wr_t *wr, ibt_mi_hdl_t *mi_hdl);
1309 
1310 /* ibt_unmap_mem_iov() */
1311 ibt_status_t ibt_unmap_mem_iov(ibt_hca_hdl_t hca_hdl, ibt_mi_hdl_t mi_hdl);
1312 
1313 /*
1314  * Work Request Functions
1315  *	Applicable for RC and UD channels.
1316  *
1317  * ibt_post_send()
1318  *	Post send work requests to the specified channel.
1319  *
1320  * ibt_post_recv()
1321  * ibt_post_srq()
1322  *	Post receive work requests to the specified channel.
1323  */
1324 ibt_status_t ibt_post_send(ibt_channel_hdl_t chan, ibt_send_wr_t *wr_list,
1325     uint_t num_wr, uint_t *posted);
1326 
1327 ibt_status_t ibt_post_recv(ibt_channel_hdl_t chan, ibt_recv_wr_t *wr_list,
1328     uint_t num_wr, uint_t *posted);
1329 
1330 ibt_status_t ibt_post_srq(ibt_srq_hdl_t srq, ibt_recv_wr_t *wr_list,
1331     uint_t num_wr, uint_t *posted);
1332 
1333 
1334 /*
1335  * Alternate Path Migration Functions.
1336  *	Applicable for RC channels only.
1337  *
1338  *
1339  * ibt_get_alt_path()
1340  *	Finds the best alternate path to a specified channel (as determined by
1341  *	the IBTL) that satisfies the requirements specified in an
1342  *	ibt_alt_path_attr_t struct.  The specified channel must have been
1343  *	previously opened successfully using ibt_open_rc_channel.
1344  *	This function also ensures that the service being accessed by the
1345  *	channel is available at the selected alternate port.
1346  *
1347  *	Note: The apa_dgid must be on the same destination channel adapter,
1348  *	if specified.
1349  *
1350  *
1351  * ibt_set_alt_path()
1352  *	Load the specified alternate path. Causes the CM to send an LAP message
1353  *	to the remote node. If successful, the local channel is updated with
1354  *	the new alternate path and the channel migration state is set to REARM.
1355  *	Can only be called on a previously opened RC channel. The channel must
1356  *	be either in RTS or paused state.
1357  *
1358  *
1359  * ibt_migrate_path()
1360  *	Force the CI to use the alternate path. The alternate path becomes
1361  *	the primary path. A new alternate path should be loaded and enabled.
1362  */
1363 ibt_status_t ibt_get_alt_path(ibt_channel_hdl_t rc_chan, ibt_path_flags_t flags,
1364     ibt_alt_path_attr_t *attr, ibt_alt_path_info_t *alt_pathp);
1365 
1366 ibt_status_t ibt_set_alt_path(ibt_channel_hdl_t rc_chan,
1367     ibt_execution_mode_t mode, ibt_alt_path_info_t *alt_pinfo, void *priv_data,
1368     ibt_priv_data_len_t priv_data_len, ibt_ap_returns_t *ret_args);
1369 
1370 ibt_status_t ibt_migrate_path(ibt_channel_hdl_t rc_chan);
1371 
1372 
1373 /*
1374  * Multicast group Functions.
1375  *	Applicable for UD channels only.
1376  */
1377 
1378 /*
1379  * ibt_attach_mcg()
1380  *	Attaches a UD channel to the specified multicast group. On successful
1381  *	completion, this channel will be provided with a copy of every
1382  *	multicast message addressed to the group specified by the MGID
1383  *	(mcg_info->mc_adds_vect.av_dgid) and received on the HCA port with
1384  *	which the channel is associated.
1385  */
1386 ibt_status_t ibt_attach_mcg(ibt_channel_hdl_t ud_chan,
1387     ibt_mcg_info_t *mcg_info);
1388 
1389 /*
1390  * ibt_detach_mcg()
1391  *	Detach the specified UD channel from the specified multicast group.
1392  */
1393 ibt_status_t ibt_detach_mcg(ibt_channel_hdl_t ud_chan,
1394     ibt_mcg_info_t *mcg_info);
1395 
1396 /*
1397  * ibt_join_mcg()
1398  *	Join a multicast group.  The first full member "join" causes the MCG
1399  *	to be created.
1400  */
1401 ibt_status_t ibt_join_mcg(ib_gid_t rgid, ibt_mcg_attr_t *mcg_attr,
1402     ibt_mcg_info_t *mcg_info_p,  ibt_mcg_handler_t func, void  *arg);
1403 
1404 /*
1405  * ibt_leave_mcg()
1406  *	The port associated with the port GID shall be removed from the
1407  *	multicast group specified by MGID (mc_gid) or from all the multicast
1408  *	groups of which it is a member if the MGID (mc_gid) is not specified
1409  *	(i.e. mc_gid.mgid_prefix must have 8-bits of 11111111 at the start of
1410  *	the GID to identify this as being a multicast GID).
1411  *
1412  *	The last full member to leave causes the destruction of the Multicast
1413  *	Group.
1414  */
1415 ibt_status_t ibt_leave_mcg(ib_gid_t rgid, ib_gid_t mc_gid, ib_gid_t port_gid,
1416     uint8_t mc_join_state);
1417 
1418 /*
1419  * ibt_query_mcg()
1420  *	Request information on multicast groups that match the parameters
1421  *	specified in mcg_attr. Information on each multicast group is returned
1422  *	to the caller in the form of an array of ibt_mcg_info_t.
1423  *	ibt_query_mcg() allocates the memory for this array and returns a
1424  *	pointer to the array (mcgs_p) and the number of entries in the array
1425  *	(entries_p). This memory should be freed by the client using
1426  *	ibt_free_mcg_info().
1427  */
1428 ibt_status_t ibt_query_mcg(ib_gid_t rgid, ibt_mcg_attr_t *mcg_attr,
1429     uint_t mcgs_max_num, ibt_mcg_info_t **mcgs_info_p, uint_t *entries_p);
1430 
1431 /*
1432  * ibt_free_mcg_info()
1433  *	Free the memory allocated by successful ibt_query_mcg()
1434  */
1435 void ibt_free_mcg_info(ibt_mcg_info_t *mcgs_info, uint_t entries);
1436 
1437 
1438 /*
1439  * ibt_register_subnet_notices()
1440  *	Register a handler to be called for subnet notifications.
1441  */
1442 void ibt_register_subnet_notices(ibt_clnt_hdl_t ibt_hdl,
1443     ibt_sm_notice_handler_t sm_notice_handler, void *private);
1444 
1445 
1446 /*
1447  * Protection Domain Functions.
1448  *
1449  * ibt_alloc_pd()
1450  * ibt_free_pd()
1451  * 	Allocate/Release a protection domain
1452  */
1453 ibt_status_t ibt_alloc_pd(ibt_hca_hdl_t hca_hdl, ibt_pd_flags_t flags,
1454     ibt_pd_hdl_t *pd);
1455 ibt_status_t ibt_free_pd(ibt_hca_hdl_t hca_hdl, ibt_pd_hdl_t pd);
1456 
1457 /*
1458  * P_Key to P_Key Index conversion Functions.
1459  *
1460  * ibt_pkey2index_byguid
1461  * ibt_pkey2index	Convert a P_Key into a P_Key index.
1462  *
1463  * ibt_index2pkey_byguid
1464  * ibt_index2pkey	Convert a P_Key Index into a P_Key.
1465  */
1466 ibt_status_t ibt_pkey2index(ibt_hca_hdl_t hca_hdl, uint8_t port_num,
1467     ib_pkey_t pkey, uint16_t *pkey_ix);
1468 
1469 ibt_status_t ibt_index2pkey(ibt_hca_hdl_t hca_hdl, uint8_t port_num,
1470     uint16_t pkey_ix, ib_pkey_t *pkey);
1471 
1472 ibt_status_t ibt_pkey2index_byguid(ib_guid_t hca_guid, uint8_t port_num,
1473     ib_pkey_t pkey, uint16_t *pkey_ix);
1474 
1475 ibt_status_t ibt_index2pkey_byguid(ib_guid_t hca_guid, uint8_t port_num,
1476     uint16_t pkey_ix, ib_pkey_t *pkey);
1477 
1478 /*
1479  *  ibt_ci_data_in()
1480  *
1481  *  Pass CI specific userland data for CI objects to the CI.
1482  */
1483 ibt_status_t ibt_ci_data_in(ibt_hca_hdl_t hca, ibt_ci_data_flags_t flags,
1484     ibt_object_type_t object, void *ibt_object_handle, void *data_p,
1485     size_t data_sz);
1486 
1487 /*
1488  *  ibt_ci_data_out()
1489  *
1490  *  Obtain CI specific userland data for CI objects.
1491  */
1492 ibt_status_t ibt_ci_data_out(ibt_hca_hdl_t hca, ibt_ci_data_flags_t flags,
1493     ibt_object_type_t object, void *ibt_object_handle, void *data_p,
1494     size_t data_sz);
1495 
1496 
1497 /*
1498  * Node Information.
1499  */
1500 
1501 /* Node type : n_node_type */
1502 #define	IBT_NODE_TYPE_CHANNEL_ADAPTER	1	/* HCA or TCA */
1503 #define	IBT_NODE_TYPE_SWITCH		2
1504 #define	IBT_NODE_TYPE_ROUTER		3
1505 
1506 typedef struct ibt_node_info_s {
1507 	ib_guid_t	n_sys_img_guid;	/* System Image GUID */
1508 	ib_guid_t	n_node_guid;	/* Node GUID */
1509 	ib_guid_t	n_port_guid;	/* Port GUID */
1510 	uint16_t	n_dev_id;	/* Device ID */
1511 	uint32_t	n_revision;	/* Device Revision */
1512 	uint32_t	n_vendor_id:24;	/* Device Vendor ID */
1513 	uint8_t		n_num_ports;	/* Number of ports on this node. */
1514 	uint8_t		n_port_num;	/* Port number. */
1515 	uint8_t		n_node_type;	/* Node type */
1516 	char		n_description[64]; /* NULL terminated ASCII string */
1517 } ibt_node_info_t;
1518 
1519 
1520 /*
1521  * ibt_gid_to_node_info()
1522  *	Retrieve node information for the specified GID.
1523  */
1524 ibt_status_t ibt_gid_to_node_info(ib_gid_t gid, ibt_node_info_t *node_info_p);
1525 
1526 /*
1527  * ibt_reprobe_dev
1528  *	Reprobe properties for IOC device node.
1529  */
1530 ibt_status_t	ibt_reprobe_dev(dev_info_t *dip);
1531 
1532 /*
1533  * ibt_get_companion_port_gids()
1534  *
1535  *	Get list of GID's available on a companion port(s) of the specified
1536  *	GID or list of GIDs available on a specified Node GUID/System Image
1537  *	GUID.
1538  */
1539 ibt_status_t ibt_get_companion_port_gids(ib_gid_t gid, ib_guid_t hca_guid,
1540     ib_guid_t sysimg_guid, ib_gid_t **gids_p, uint_t *num_gids_p);
1541 
1542 /*
1543  * SHARED RECEIVE QUEUE
1544  */
1545 
1546 
1547 /*
1548  * ibt_alloc_srq()
1549  *	Allocate a shared receive queue.
1550  */
1551 ibt_status_t ibt_alloc_srq(ibt_hca_hdl_t hca_hdl, ibt_srq_flags_t flags,
1552     ibt_pd_hdl_t pd, ibt_srq_sizes_t *sizes, ibt_srq_hdl_t *ibt_srq_p,
1553     ibt_srq_sizes_t *real_size_p);
1554 
1555 /*
1556  * ibt_free_srq()
1557  *	Free allocated SRQ resources.
1558  */
1559 ibt_status_t ibt_free_srq(ibt_srq_hdl_t ibt_srq);
1560 
1561 /*
1562  * ibt_query_srq()
1563  *	Query a shared receive queue.
1564  */
1565 ibt_status_t ibt_query_srq(ibt_srq_hdl_t ibt_srq, ibt_pd_hdl_t *pd_p,
1566     ibt_srq_sizes_t *sizes_p, uint_t *limit_p);
1567 
1568 /*
1569  * ibt_modify_srq()
1570  *	Modify a shared receive queue.
1571  */
1572 ibt_status_t ibt_modify_srq(ibt_srq_hdl_t ibt_srq, ibt_srq_modify_flags_t flags,
1573     uint_t size, uint_t limit, uint_t *real_size_p);
1574 
1575 /*
1576  * ibt_set_srq_private()
1577  * ibt_get_srq_private()
1578  *	Set/get the SRQ client private data.
1579  */
1580 void ibt_set_srq_private(ibt_srq_hdl_t ibt_srq, void *clnt_private);
1581 void *ibt_get_srq_private(ibt_srq_hdl_t ibt_srq);
1582 
1583 /*
1584  * ibt_check_failure()
1585  * 	Function to test for special case failures
1586  */
1587 ibt_failure_type_t ibt_check_failure(ibt_status_t status, uint64_t *reserved_p);
1588 
1589 
1590 /*
1591  * ibt_hw_is_present() returns 0 when there is no IB hardware actively
1592  * running.  This is primarily useful for modules like rpcmod which needs a
1593  * quick check to decide whether or not it should try to use InfiniBand.
1594  */
1595 int ibt_hw_is_present();
1596 
1597 /*
1598  * Fast Memory Registration (FMR).
1599  *
1600  * ibt_create_fmr_pool
1601  *	Not fast-path.
1602  *	ibt_create_fmr_pool() verifies that the HCA supports FMR and allocates
1603  *	and initializes an "FMR pool".  This pool contains state specific to
1604  *	this registration, including the watermark setting to determine when
1605  *	to sync, and the total number of FMR regions available within this pool.
1606  *
1607  * ibt_destroy_fmr_pool
1608  *	ibt_destroy_fmr_pool() deallocates all of the FMR regions in a specific
1609  *	pool.  All state and information regarding the pool are destroyed and
1610  *	returned as free space once again.  No more use of FMR regions in this
1611  *	pool are possible without a subsequent call to ibt_create_fmr_pool().
1612  *
1613  * ibt_flush_fmr_pool
1614  *	ibt_flush_fmr_pool forces a flush to occur.  At the client's request,
1615  *	any unmapped FMR regions (See 'ibt_deregister_mr())') are returned to
1616  *	a free state.  This function allows for an asynchronous cleanup of
1617  *	formerly used FMR regions.  Sync operation is also performed internally
1618  *	by HCA driver, when 'watermark' settings for the number of free FMR
1619  *	regions left in the "pool" is reached.
1620  *
1621  * ibt_register_physical_fmr
1622  *	ibt_register_physical_fmr() assigns a "free" entry from the FMR Pool.
1623  *	It first consults the "FMR cache" to see if this is a duplicate memory
1624  *	registration to something already in use.  If not, then a free entry
1625  *	in the "pool" is marked used.
1626  *
1627  * ibt_deregister_fmr
1628  *	The ibt_deregister_fmr un-maps the resources reserved from the FMR
1629  *	pool by ibt_register_physical_fmr().   The ibt_deregister_fmr() will
1630  *	mark the region as free in the FMR Pool.
1631  */
1632 ibt_status_t ibt_create_fmr_pool(ibt_hca_hdl_t hca_hdl, ibt_pd_hdl_t pd,
1633     ibt_fmr_pool_attr_t *fmr_params, ibt_fmr_pool_hdl_t *fmr_pool_p);
1634 
1635 ibt_status_t ibt_destroy_fmr_pool(ibt_hca_hdl_t hca_hdl,
1636     ibt_fmr_pool_hdl_t fmr_pool);
1637 
1638 ibt_status_t ibt_flush_fmr_pool(ibt_hca_hdl_t hca_hdl,
1639     ibt_fmr_pool_hdl_t fmr_pool);
1640 
1641 ibt_status_t ibt_register_physical_fmr(ibt_hca_hdl_t hca_hdl,
1642     ibt_fmr_pool_hdl_t fmr_pool, ibt_pmr_attr_t *mem_pattr,
1643     ibt_mr_hdl_t *mr_hdl_p, ibt_pmr_desc_t *mem_desc_p);
1644 
1645 ibt_status_t ibt_deregister_fmr(ibt_hca_hdl_t hca, ibt_mr_hdl_t mr_hdl);
1646 
1647 /*
1648  * IP SUPPORT
1649  */
1650 
1651 /*
1652  * IP get_paths
1653  * Returns an array (or single) of paths and source IP addresses. In the
1654  * simplest form just the destination IP address is specified, and one path
1655  * is requested, then one ibt_path_info_t struct and one source IP.
1656  *
1657  * More than one path can be requested to a single destination, in which case
1658  * the requested number of ibt_path_info_t's are returned, and the same
1659  * number of SRC IP address, with the first SRC IP address corrosponding
1660  * to the first ibt_path_info_t, etc.
1661  *
1662  * Restrictions on the source end point can be specified, in the form of a
1663  * source IP address (this implicitly defines the HCA, HCA port and Pkey)
1664  * HCA, HCA port, and sgid (implicitly defines HCA and HCA port).
1665  * Combinations are allowed but they  must be consistent.
1666  *
1667  * Path attributes can also be specified, these can also affect local HCA
1668  * selection.
1669  *
1670  * ibt_get_ip_paths()  internally does (among other things):
1671  *
1672  *   o ibt_get_list_of_ibd_ipaddr_and_macaddr( OUT list_ipaddr_macaddr)
1673  *
1674  *   o extract_pkey_and_sgid(IN list_ipaddr_macaddr, OUT list_pkey_and_sgid)
1675  *
1676  *   o map_dst_ip_addr(IN dst_ip_addr, OUT dst_pkey, OUT dgid) - See Note
1677  *
1678  *   o filter_by_pkey(IN list_pkey_and_sgid, IN dst_pkey, OUT list_of_sgid)
1679  *
1680  *   o do_multipath_query(IN list_of_sgid, IN dst_pkey, IN dgid, OUT path_list)
1681  *
1682  *   o pick_a_good_path(IN path_list, OUT the_path)
1683  *
1684  *   o find_matching_src_ip(IN the_path, IN list_ipaddr_macaddr, OUT src_ip)
1685  *
1686  * The ibd instance which got the ARP response is only on one P_Key
1687  * knowing the ibd instance (or which IPonIB MCG) got the ARP response
1688  * determins the P_Key associated with a dgid. If the proposedi "ip2mac()"
1689  * API is used to get an IP to GID translations, then returned 'sockaddr_dl'
1690  * contains the interface name and index.
1691  *
1692  *
1693  * Example:
1694  *   ip_path_attr.ipa_dst_ip = dst_ip_addr;
1695  *   ip_path_attr.ipa_ndst = 1;
1696  *   ip_path_attr.ipa_max_paths = 1;
1697  *
1698  *   status = ibt_get_ip_paths(clnt_hdl, flags, &ip_path_attr, &paths,
1699  *      &num_paths_p, &src_ip);
1700  *
1701  *   sid = ibt_get_ip_sid(protocol_num, dst_port);
1702  *   path_info->sid = sid;
1703  *
1704  *   ip_cm_info.src_addr = src_ip;
1705  *   ip_cm_info.dst_addr = dst_ip_addr;
1706  *   ip_cm_info.src_port = src_port
1707  *
1708  *   ibt_format_ip_private_data(ip_cm_info, priv_data_len, &priv_data);
1709  *   ibt_open_rc_channel(chan, private_data);
1710  */
1711 typedef struct ibt_ip_path_attr_s {
1712 	ibt_ip_addr_t		*ipa_dst_ip;		/* Required */
1713 	ibt_ip_addr_t		ipa_src_ip;		/* Optional */
1714 	ib_guid_t		ipa_hca_guid;		/* Optional */
1715 	uint8_t			ipa_hca_port_num;	/* Optional */
1716 	uint8_t			ipa_max_paths;		/* Required */
1717 	uint8_t			ipa_ndst;		/* Required */
1718 	uint8_t			ipa_sl:4;		/* Optional */
1719 	ibt_mtu_req_t		ipa_mtu;		/* Optional */
1720 	ibt_srate_req_t		ipa_srate;		/* Optional */
1721 	ibt_pkt_lt_req_t	ipa_pkt_lt;		/* Optional */
1722 	uint_t			ipa_flow:20;		/* Optional */
1723 	uint8_t			ipa_hop;		/* Optional */
1724 	uint8_t			ipa_tclass;		/* Optional */
1725 } ibt_ip_path_attr_t;
1726 
1727 /*
1728  * Path SRC IP addresses
1729  */
1730 typedef struct ibt_path_ip_src_s {
1731 	ibt_ip_addr_t	ip_primary;
1732 	ibt_ip_addr_t	ip_alternate;
1733 } ibt_path_ip_src_t;
1734 
1735 
1736 ibt_status_t ibt_get_ip_paths(ibt_clnt_hdl_t ibt_hdl, ibt_path_flags_t flags,
1737     ibt_ip_path_attr_t *attr, ibt_path_info_t *paths_p, uint8_t *num_paths_p,
1738     ibt_path_ip_src_t *src_ip_p);
1739 
1740 ibt_status_t ibt_get_src_ip(ib_gid_t gid, ib_pkey_t pkey,
1741     ibt_ip_addr_t *src_ip);
1742 
1743 /*
1744  * Callback function that can be used in ibt_aget_ip_paths(), a Non-Blocking
1745  * version of ibt_get_ip_paths().
1746  */
1747 typedef void (*ibt_ip_path_handler_t)(void *arg, ibt_status_t retval,
1748     ibt_path_info_t *paths_p, uint8_t num_paths, ibt_path_ip_src_t *src_ip_p);
1749 
1750 /*
1751  * Find path(s) to a given destination or service asynchronously.
1752  * ibt_aget_ip_paths() is a Non-Blocking version of ibt_get_ip_paths().
1753  */
1754 ibt_status_t ibt_aget_ip_paths(ibt_clnt_hdl_t ibt_hdl, ibt_path_flags_t flags,
1755     ibt_ip_path_attr_t *attr, ibt_ip_path_handler_t func, void  *arg);
1756 
1757 /*
1758  * IP RDMA protocol functions
1759  */
1760 
1761 /*
1762  * IBTF manages the port number space for non well known ports. If a ULP
1763  * is not using TCP/UDP and a well known port, then ibt_get_ip_sid() returns
1764  * an sid based on the IP protocol number '0' (reserved) and an IBTF assigned
1765  * port number.  ibt_release_ip_sid() should be used to release the hold
1766  * of SID created by ibt_get_ip_sid().
1767  */
1768 ib_svc_id_t ibt_get_ip_sid(uint8_t protocol_num, in_port_t dst_port);
1769 ibt_status_t ibt_release_ip_sid(ib_svc_id_t sid);
1770 
1771 uint8_t ibt_get_ip_protocol_num(ib_svc_id_t sid);
1772 in_port_t ibt_get_ip_dst_port(ib_svc_id_t sid);
1773 
1774 /*
1775  * Functions to format/extract the RDMA IP CM private data
1776  */
1777 typedef struct ibt_ip_cm_info_s {
1778 	ibt_ip_addr_t	src_addr;
1779 	ibt_ip_addr_t	dst_addr;
1780 	in_port_t	src_port;
1781 } ibt_ip_cm_info_t;
1782 
1783 /*
1784  * If a ULP is using IP addressing as defined by the RDMA IP CM Service IBTA
1785  * Annex 11, then it must always allocate a private data buffer for use in
1786  * the ibt_open_rc_channel(9F) call. The minimum size of the buffer is
1787  * IBT_IP_HDR_PRIV_DATA_SZ, if the ULP has no ULP specific private data.
1788  * This allows ibt_format_ip_private_data() to place the RDMA IP CM service
1789  * hello message in the private data of the REQ. If the ULP has some ULP
1790  * specific private data then it should allocate a buffer big enough to
1791  * contain that data plus an additional IBT_IP_HDR_PRIV_DATA_SZ bytes.
1792  * The ULP should place its  ULP specific private data at offset
1793  * IBT_IP_HDR_PRIV_DATA_SZ in the allocated buffer before calling
1794  * ibt_format_ip_private_data().
1795  */
1796 ibt_status_t ibt_format_ip_private_data(ibt_ip_cm_info_t *ip_cm_info,
1797     ibt_priv_data_len_t priv_data_len, void *priv_data_p);
1798 ibt_status_t ibt_get_ip_data(ibt_priv_data_len_t priv_data_len,
1799     void *priv_data, ibt_ip_cm_info_t *ip_info_p);
1800 
1801 /*
1802  * The ibt_alt_ip_path_attr_t structure is used to specify additional optional
1803  * attributes when requesting an alternate path for an existing channel.
1804  *
1805  * Attributes that are don't care should be set to NULL or '0'.
1806  */
1807 typedef struct ibt_alt_ip_path_attr_s {
1808 	ibt_ip_addr_t		apa_dst_ip;
1809 	ibt_ip_addr_t		apa_src_ip;
1810 	ibt_srate_req_t		apa_srate;
1811 	ibt_pkt_lt_req_t	apa_pkt_lt;	/* Packet Life Time Request */
1812 	uint_t			apa_flow:20;
1813 	uint8_t			apa_sl:4;
1814 	uint8_t			apa_hop;
1815 	uint8_t			apa_tclass;
1816 } ibt_alt_ip_path_attr_t;
1817 
1818 ibt_status_t ibt_get_ip_alt_path(ibt_channel_hdl_t rc_chan,
1819     ibt_path_flags_t flags, ibt_alt_ip_path_attr_t *attr,
1820     ibt_alt_path_info_t *alt_path);
1821 
1822 /*
1823  * CONTRACT PRIVATE ONLY INTERFACES
1824  *
1825  * DO NOT USE THE FOLLOWING FUNCTIONS WITHOUT SIGNING THE CONTRACT
1826  * WITH IBTF GROUP.
1827  */
1828 
1829 /* Define an Address Record structure (data for ATS service records). */
1830 typedef struct ibt_ar_s {
1831 	ib_gid_t	ar_gid;		/* GID of local HCA port */
1832 	ib_pkey_t	ar_pkey;	/* P_Key valid on port of ar_gid */
1833 	uint8_t		ar_data[16];	/* Data affiliated with GID/P_Key */
1834 } ibt_ar_t;
1835 
1836 /*
1837  * ibt_register_ar()
1838  * ibt_deregister_ar()
1839  *	Register/deregister an Address Record with the SA.
1840  * ibt_query_ar()
1841  *	Query the SA for Address Records matching either GID/P_Key or Data.
1842  */
1843 ibt_status_t ibt_register_ar(ibt_clnt_hdl_t ibt_hdl, ibt_ar_t *arp);
1844 
1845 ibt_status_t ibt_deregister_ar(ibt_clnt_hdl_t ibt_hdl, ibt_ar_t *arp);
1846 
1847 ibt_status_t ibt_query_ar(ib_gid_t *sgid, ibt_ar_t *queryp, ibt_ar_t *resultp);
1848 
1849 
1850 /*
1851  * ibt_modify_system_image()
1852  * ibt_modify_system_image_byguid()
1853  *	Modify specified HCA's system image GUID.
1854  */
1855 ibt_status_t ibt_modify_system_image(ibt_hca_hdl_t hca_hdl, ib_guid_t sys_guid);
1856 
1857 ibt_status_t ibt_modify_system_image_byguid(ib_guid_t hca_guid,
1858     ib_guid_t sys_guid);
1859 
1860 
1861 /*
1862  * ibt_modify_port()
1863  * ibt_modify_port_byguid()
1864  *	Modify the specified port, or all ports attribute(s).
1865  */
1866 ibt_status_t ibt_modify_port(ibt_hca_hdl_t hca_hdl, uint8_t port,
1867     ibt_port_modify_flags_t flags, uint8_t init_type);
1868 
1869 ibt_status_t ibt_modify_port_byguid(ib_guid_t hca_guid, uint8_t port,
1870     ibt_port_modify_flags_t flags, uint8_t init_type);
1871 
1872 
1873 /*
1874  * ibt_get_port_state()
1875  * ibt_get_port_state_byguid()
1876  *	Return the most commonly requested attributes of an HCA port.
1877  *	If the link state is not IBT_PORT_ACTIVE, the other returned values
1878  *	are undefined.
1879  */
1880 ibt_status_t ibt_get_port_state(ibt_hca_hdl_t hca_hdl, uint8_t port,
1881     ib_gid_t *sgid_p, ib_lid_t *base_lid_p);
1882 
1883 ibt_status_t ibt_get_port_state_byguid(ib_guid_t hca_guid, uint8_t port,
1884     ib_gid_t *sgid_p, ib_lid_t *base_lid_p);
1885 
1886 /*
1887  * ibt_alloc_io_mem()
1888  * ibt_free_io_mem()
1889  *	Allocate and deallocate dma-able memory.
1890  */
1891 ibt_status_t ibt_alloc_io_mem(ibt_hca_hdl_t, size_t, ibt_mr_flags_t,
1892     caddr_t *, ibt_mem_alloc_hdl_t *);
1893 
1894 ibt_status_t ibt_free_io_mem(ibt_hca_hdl_t, ibt_mem_alloc_hdl_t);
1895 
1896 #ifdef __cplusplus
1897 }
1898 #endif
1899 
1900 #endif /* _SYS_IB_IBTL_IBTI_COMMON_H */
1901