]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - include/rdma/ib_verbs.h
IB/hfi1: VNIC SDMA support
[mirror_ubuntu-hirsute-kernel.git] / include / rdma / ib_verbs.h
CommitLineData
1da177e4
LT
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.
2a1d9b7f 7 * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
f7c6a7b5 8 * Copyright (c) 2005, 2006, 2007 Cisco Systems. All rights reserved.
1da177e4
LT
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.
1da177e4
LT
37 */
38
39#if !defined(IB_VERBS_H)
40#define IB_VERBS_H
41
42#include <linux/types.h>
43#include <linux/device.h>
9b513090
RC
44#include <linux/mm.h>
45#include <linux/dma-mapping.h>
459d6e2a 46#include <linux/kref.h>
bfb3ea12
DB
47#include <linux/list.h>
48#include <linux/rwsem.h>
87ae9afd 49#include <linux/scatterlist.h>
f0626710 50#include <linux/workqueue.h>
9268f72d 51#include <linux/socket.h>
14d3a3b2 52#include <linux/irq_poll.h>
dd5f03be 53#include <uapi/linux/if_ether.h>
c865f246
SK
54#include <net/ipv6.h>
55#include <net/ip.h>
301a721e
MB
56#include <linux/string.h>
57#include <linux/slab.h>
2fc77572 58#include <linux/netdevice.h>
e2773c06 59
50174a7f 60#include <linux/if_link.h>
60063497 61#include <linux/atomic.h>
882214e2 62#include <linux/mmu_notifier.h>
7c0f6ba6 63#include <linux/uaccess.h>
43579b5f 64#include <linux/cgroup_rdma.h>
1da177e4 65
f0626710 66extern struct workqueue_struct *ib_wq;
14d3a3b2 67extern struct workqueue_struct *ib_comp_wq;
f0626710 68
1da177e4
LT
69union ib_gid {
70 u8 raw[16];
71 struct {
97f52eb4
SH
72 __be64 subnet_prefix;
73 __be64 interface_id;
1da177e4
LT
74 } global;
75};
76
e26be1bf
MS
77extern union ib_gid zgid;
78
b39ffa1d
MB
79enum ib_gid_type {
80 /* If link layer is Ethernet, this is RoCE V1 */
81 IB_GID_TYPE_IB = 0,
82 IB_GID_TYPE_ROCE = 0,
7766a99f 83 IB_GID_TYPE_ROCE_UDP_ENCAP = 1,
b39ffa1d
MB
84 IB_GID_TYPE_SIZE
85};
86
7ead4bcb 87#define ROCE_V2_UDP_DPORT 4791
03db3a2d 88struct ib_gid_attr {
b39ffa1d 89 enum ib_gid_type gid_type;
03db3a2d
MB
90 struct net_device *ndev;
91};
92
07ebafba
TT
93enum rdma_node_type {
94 /* IB values map to NodeInfo:NodeType. */
95 RDMA_NODE_IB_CA = 1,
96 RDMA_NODE_IB_SWITCH,
97 RDMA_NODE_IB_ROUTER,
180771a3
UM
98 RDMA_NODE_RNIC,
99 RDMA_NODE_USNIC,
5db5765e 100 RDMA_NODE_USNIC_UDP,
1da177e4
LT
101};
102
a0c1b2a3
EC
103enum {
104 /* set the local administered indication */
105 IB_SA_WELL_KNOWN_GUID = BIT_ULL(57) | 2,
106};
107
07ebafba
TT
108enum rdma_transport_type {
109 RDMA_TRANSPORT_IB,
180771a3 110 RDMA_TRANSPORT_IWARP,
248567f7
UM
111 RDMA_TRANSPORT_USNIC,
112 RDMA_TRANSPORT_USNIC_UDP
07ebafba
TT
113};
114
6b90a6d6
MW
115enum rdma_protocol_type {
116 RDMA_PROTOCOL_IB,
117 RDMA_PROTOCOL_IBOE,
118 RDMA_PROTOCOL_IWARP,
119 RDMA_PROTOCOL_USNIC_UDP
120};
121
8385fd84
RD
122__attribute_const__ enum rdma_transport_type
123rdma_node_get_transport(enum rdma_node_type node_type);
07ebafba 124
c865f246
SK
125enum rdma_network_type {
126 RDMA_NETWORK_IB,
127 RDMA_NETWORK_ROCE_V1 = RDMA_NETWORK_IB,
128 RDMA_NETWORK_IPV4,
129 RDMA_NETWORK_IPV6
130};
131
132static inline enum ib_gid_type ib_network_to_gid_type(enum rdma_network_type network_type)
133{
134 if (network_type == RDMA_NETWORK_IPV4 ||
135 network_type == RDMA_NETWORK_IPV6)
136 return IB_GID_TYPE_ROCE_UDP_ENCAP;
137
138 /* IB_GID_TYPE_IB same as RDMA_NETWORK_ROCE_V1 */
139 return IB_GID_TYPE_IB;
140}
141
142static inline enum rdma_network_type ib_gid_to_network_type(enum ib_gid_type gid_type,
143 union ib_gid *gid)
144{
145 if (gid_type == IB_GID_TYPE_IB)
146 return RDMA_NETWORK_IB;
147
148 if (ipv6_addr_v4mapped((struct in6_addr *)gid))
149 return RDMA_NETWORK_IPV4;
150 else
151 return RDMA_NETWORK_IPV6;
152}
153
a3f5adaf
EC
154enum rdma_link_layer {
155 IB_LINK_LAYER_UNSPECIFIED,
156 IB_LINK_LAYER_INFINIBAND,
157 IB_LINK_LAYER_ETHERNET,
158};
159
1da177e4 160enum ib_device_cap_flags {
7ca0bc53
LR
161 IB_DEVICE_RESIZE_MAX_WR = (1 << 0),
162 IB_DEVICE_BAD_PKEY_CNTR = (1 << 1),
163 IB_DEVICE_BAD_QKEY_CNTR = (1 << 2),
164 IB_DEVICE_RAW_MULTI = (1 << 3),
165 IB_DEVICE_AUTO_PATH_MIG = (1 << 4),
166 IB_DEVICE_CHANGE_PHY_PORT = (1 << 5),
167 IB_DEVICE_UD_AV_PORT_ENFORCE = (1 << 6),
168 IB_DEVICE_CURR_QP_STATE_MOD = (1 << 7),
169 IB_DEVICE_SHUTDOWN_PORT = (1 << 8),
170 IB_DEVICE_INIT_TYPE = (1 << 9),
171 IB_DEVICE_PORT_ACTIVE_EVENT = (1 << 10),
172 IB_DEVICE_SYS_IMAGE_GUID = (1 << 11),
173 IB_DEVICE_RC_RNR_NAK_GEN = (1 << 12),
174 IB_DEVICE_SRQ_RESIZE = (1 << 13),
175 IB_DEVICE_N_NOTIFY_CQ = (1 << 14),
b1adc714
CH
176
177 /*
178 * This device supports a per-device lkey or stag that can be
179 * used without performing a memory registration for the local
180 * memory. Note that ULPs should never check this flag, but
181 * instead of use the local_dma_lkey flag in the ib_pd structure,
182 * which will always contain a usable lkey.
183 */
7ca0bc53
LR
184 IB_DEVICE_LOCAL_DMA_LKEY = (1 << 15),
185 IB_DEVICE_RESERVED /* old SEND_W_INV */ = (1 << 16),
186 IB_DEVICE_MEM_WINDOW = (1 << 17),
e0605d91
EC
187 /*
188 * Devices should set IB_DEVICE_UD_IP_SUM if they support
189 * insertion of UDP and TCP checksum on outgoing UD IPoIB
190 * messages and can verify the validity of checksum for
191 * incoming messages. Setting this flag implies that the
192 * IPoIB driver may set NETIF_F_IP_CSUM for datagram mode.
193 */
7ca0bc53
LR
194 IB_DEVICE_UD_IP_CSUM = (1 << 18),
195 IB_DEVICE_UD_TSO = (1 << 19),
196 IB_DEVICE_XRC = (1 << 20),
b1adc714
CH
197
198 /*
199 * This device supports the IB "base memory management extension",
200 * which includes support for fast registrations (IB_WR_REG_MR,
201 * IB_WR_LOCAL_INV and IB_WR_SEND_WITH_INV verbs). This flag should
202 * also be set by any iWarp device which must support FRs to comply
203 * to the iWarp verbs spec. iWarp devices also support the
204 * IB_WR_RDMA_READ_WITH_INV verb for RDMA READs that invalidate the
205 * stag.
206 */
7ca0bc53
LR
207 IB_DEVICE_MEM_MGT_EXTENSIONS = (1 << 21),
208 IB_DEVICE_BLOCK_MULTICAST_LOOPBACK = (1 << 22),
209 IB_DEVICE_MEM_WINDOW_TYPE_2A = (1 << 23),
210 IB_DEVICE_MEM_WINDOW_TYPE_2B = (1 << 24),
211 IB_DEVICE_RC_IP_CSUM = (1 << 25),
ebaaee25 212 /* Deprecated. Please use IB_RAW_PACKET_CAP_IP_CSUM. */
7ca0bc53 213 IB_DEVICE_RAW_IP_CSUM = (1 << 26),
8a06ce59
LR
214 /*
215 * Devices should set IB_DEVICE_CROSS_CHANNEL if they
216 * support execution of WQEs that involve synchronization
217 * of I/O operations with single completion queue managed
218 * by hardware.
219 */
220 IB_DEVICE_CROSS_CHANNEL = (1 << 27),
7ca0bc53
LR
221 IB_DEVICE_MANAGED_FLOW_STEERING = (1 << 29),
222 IB_DEVICE_SIGNATURE_HANDOVER = (1 << 30),
47355b3c 223 IB_DEVICE_ON_DEMAND_PAGING = (1ULL << 31),
f5aa9159 224 IB_DEVICE_SG_GAPS_REG = (1ULL << 32),
c7e162a4 225 IB_DEVICE_VIRTUAL_FUNCTION = (1ULL << 33),
ebaaee25 226 /* Deprecated. Please use IB_RAW_PACKET_CAP_SCATTER_FCS. */
c7e162a4 227 IB_DEVICE_RAW_SCATTER_FCS = (1ULL << 34),
62e45949 228 IB_DEVICE_RDMA_NETDEV_OPA_VNIC = (1ULL << 35),
1b01d335
SG
229};
230
231enum ib_signature_prot_cap {
232 IB_PROT_T10DIF_TYPE_1 = 1,
233 IB_PROT_T10DIF_TYPE_2 = 1 << 1,
234 IB_PROT_T10DIF_TYPE_3 = 1 << 2,
235};
236
237enum ib_signature_guard_cap {
238 IB_GUARD_T10DIF_CRC = 1,
239 IB_GUARD_T10DIF_CSUM = 1 << 1,
1da177e4
LT
240};
241
242enum ib_atomic_cap {
243 IB_ATOMIC_NONE,
244 IB_ATOMIC_HCA,
245 IB_ATOMIC_GLOB
246};
247
860f10a7 248enum ib_odp_general_cap_bits {
25bf14d6
AK
249 IB_ODP_SUPPORT = 1 << 0,
250 IB_ODP_SUPPORT_IMPLICIT = 1 << 1,
860f10a7
SG
251};
252
253enum ib_odp_transport_cap_bits {
254 IB_ODP_SUPPORT_SEND = 1 << 0,
255 IB_ODP_SUPPORT_RECV = 1 << 1,
256 IB_ODP_SUPPORT_WRITE = 1 << 2,
257 IB_ODP_SUPPORT_READ = 1 << 3,
258 IB_ODP_SUPPORT_ATOMIC = 1 << 4,
259};
260
261struct ib_odp_caps {
262 uint64_t general_caps;
263 struct {
264 uint32_t rc_odp_caps;
265 uint32_t uc_odp_caps;
266 uint32_t ud_odp_caps;
267 } per_transport_caps;
268};
269
ccf20562
YH
270struct ib_rss_caps {
271 /* Corresponding bit will be set if qp type from
272 * 'enum ib_qp_type' is supported, e.g.
273 * supported_qpts |= 1 << IB_QPT_UD
274 */
275 u32 supported_qpts;
276 u32 max_rwq_indirection_tables;
277 u32 max_rwq_indirection_table_size;
278};
279
b9926b92
MB
280enum ib_cq_creation_flags {
281 IB_CQ_FLAGS_TIMESTAMP_COMPLETION = 1 << 0,
8a06ce59 282 IB_CQ_FLAGS_IGNORE_OVERRUN = 1 << 1,
b9926b92
MB
283};
284
bcf4c1ea
MB
285struct ib_cq_init_attr {
286 unsigned int cqe;
287 int comp_vector;
288 u32 flags;
289};
290
1da177e4
LT
291struct ib_device_attr {
292 u64 fw_ver;
97f52eb4 293 __be64 sys_image_guid;
1da177e4
LT
294 u64 max_mr_size;
295 u64 page_size_cap;
296 u32 vendor_id;
297 u32 vendor_part_id;
298 u32 hw_ver;
299 int max_qp;
300 int max_qp_wr;
fb532d6a 301 u64 device_cap_flags;
1da177e4
LT
302 int max_sge;
303 int max_sge_rd;
304 int max_cq;
305 int max_cqe;
306 int max_mr;
307 int max_pd;
308 int max_qp_rd_atom;
309 int max_ee_rd_atom;
310 int max_res_rd_atom;
311 int max_qp_init_rd_atom;
312 int max_ee_init_rd_atom;
313 enum ib_atomic_cap atomic_cap;
5e80ba8f 314 enum ib_atomic_cap masked_atomic_cap;
1da177e4
LT
315 int max_ee;
316 int max_rdd;
317 int max_mw;
318 int max_raw_ipv6_qp;
319 int max_raw_ethy_qp;
320 int max_mcast_grp;
321 int max_mcast_qp_attach;
322 int max_total_mcast_qp_attach;
323 int max_ah;
324 int max_fmr;
325 int max_map_per_fmr;
326 int max_srq;
327 int max_srq_wr;
328 int max_srq_sge;
00f7ec36 329 unsigned int max_fast_reg_page_list_len;
1da177e4
LT
330 u16 max_pkeys;
331 u8 local_ca_ack_delay;
1b01d335
SG
332 int sig_prot_cap;
333 int sig_guard_cap;
860f10a7 334 struct ib_odp_caps odp_caps;
24306dc6
MB
335 uint64_t timestamp_mask;
336 uint64_t hca_core_clock; /* in KHZ */
ccf20562
YH
337 struct ib_rss_caps rss_caps;
338 u32 max_wq_type_rq;
ebaaee25 339 u32 raw_packet_caps; /* Use ib_raw_packet_caps enum */
1da177e4
LT
340};
341
342enum ib_mtu {
343 IB_MTU_256 = 1,
344 IB_MTU_512 = 2,
345 IB_MTU_1024 = 3,
346 IB_MTU_2048 = 4,
347 IB_MTU_4096 = 5
348};
349
350static inline int ib_mtu_enum_to_int(enum ib_mtu mtu)
351{
352 switch (mtu) {
353 case IB_MTU_256: return 256;
354 case IB_MTU_512: return 512;
355 case IB_MTU_1024: return 1024;
356 case IB_MTU_2048: return 2048;
357 case IB_MTU_4096: return 4096;
358 default: return -1;
359 }
360}
361
d3f4aadd
AR
362static inline enum ib_mtu ib_mtu_int_to_enum(int mtu)
363{
364 if (mtu >= 4096)
365 return IB_MTU_4096;
366 else if (mtu >= 2048)
367 return IB_MTU_2048;
368 else if (mtu >= 1024)
369 return IB_MTU_1024;
370 else if (mtu >= 512)
371 return IB_MTU_512;
372 else
373 return IB_MTU_256;
374}
375
1da177e4
LT
376enum ib_port_state {
377 IB_PORT_NOP = 0,
378 IB_PORT_DOWN = 1,
379 IB_PORT_INIT = 2,
380 IB_PORT_ARMED = 3,
381 IB_PORT_ACTIVE = 4,
382 IB_PORT_ACTIVE_DEFER = 5
383};
384
385enum ib_port_cap_flags {
386 IB_PORT_SM = 1 << 1,
387 IB_PORT_NOTICE_SUP = 1 << 2,
388 IB_PORT_TRAP_SUP = 1 << 3,
389 IB_PORT_OPT_IPD_SUP = 1 << 4,
390 IB_PORT_AUTO_MIGR_SUP = 1 << 5,
391 IB_PORT_SL_MAP_SUP = 1 << 6,
392 IB_PORT_MKEY_NVRAM = 1 << 7,
393 IB_PORT_PKEY_NVRAM = 1 << 8,
394 IB_PORT_LED_INFO_SUP = 1 << 9,
395 IB_PORT_SM_DISABLED = 1 << 10,
396 IB_PORT_SYS_IMAGE_GUID_SUP = 1 << 11,
397 IB_PORT_PKEY_SW_EXT_PORT_TRAP_SUP = 1 << 12,
71eeba16 398 IB_PORT_EXTENDED_SPEEDS_SUP = 1 << 14,
1da177e4
LT
399 IB_PORT_CM_SUP = 1 << 16,
400 IB_PORT_SNMP_TUNNEL_SUP = 1 << 17,
401 IB_PORT_REINIT_SUP = 1 << 18,
402 IB_PORT_DEVICE_MGMT_SUP = 1 << 19,
403 IB_PORT_VENDOR_CLASS_SUP = 1 << 20,
404 IB_PORT_DR_NOTICE_SUP = 1 << 21,
405 IB_PORT_CAP_MASK_NOTICE_SUP = 1 << 22,
406 IB_PORT_BOOT_MGMT_SUP = 1 << 23,
407 IB_PORT_LINK_LATENCY_SUP = 1 << 24,
b4a26a27 408 IB_PORT_CLIENT_REG_SUP = 1 << 25,
03db3a2d 409 IB_PORT_IP_BASED_GIDS = 1 << 26,
1da177e4
LT
410};
411
412enum ib_port_width {
413 IB_WIDTH_1X = 1,
414 IB_WIDTH_4X = 2,
415 IB_WIDTH_8X = 4,
416 IB_WIDTH_12X = 8
417};
418
419static inline int ib_width_enum_to_int(enum ib_port_width width)
420{
421 switch (width) {
422 case IB_WIDTH_1X: return 1;
423 case IB_WIDTH_4X: return 4;
424 case IB_WIDTH_8X: return 8;
425 case IB_WIDTH_12X: return 12;
426 default: return -1;
427 }
428}
429
2e96691c
OG
430enum ib_port_speed {
431 IB_SPEED_SDR = 1,
432 IB_SPEED_DDR = 2,
433 IB_SPEED_QDR = 4,
434 IB_SPEED_FDR10 = 8,
435 IB_SPEED_FDR = 16,
436 IB_SPEED_EDR = 32
437};
438
b40f4757
CL
439/**
440 * struct rdma_hw_stats
441 * @timestamp - Used by the core code to track when the last update was
442 * @lifespan - Used by the core code to determine how old the counters
443 * should be before being updated again. Stored in jiffies, defaults
444 * to 10 milliseconds, drivers can override the default be specifying
445 * their own value during their allocation routine.
446 * @name - Array of pointers to static names used for the counters in
447 * directory.
448 * @num_counters - How many hardware counters there are. If name is
449 * shorter than this number, a kernel oops will result. Driver authors
450 * are encouraged to leave BUILD_BUG_ON(ARRAY_SIZE(@name) < num_counters)
451 * in their code to prevent this.
452 * @value - Array of u64 counters that are accessed by the sysfs code and
453 * filled in by the drivers get_stats routine
454 */
455struct rdma_hw_stats {
456 unsigned long timestamp;
457 unsigned long lifespan;
458 const char * const *names;
459 int num_counters;
460 u64 value[];
7f624d02
SW
461};
462
b40f4757
CL
463#define RDMA_HW_STATS_DEFAULT_LIFESPAN 10
464/**
465 * rdma_alloc_hw_stats_struct - Helper function to allocate dynamic struct
466 * for drivers.
467 * @names - Array of static const char *
468 * @num_counters - How many elements in array
469 * @lifespan - How many milliseconds between updates
470 */
471static inline struct rdma_hw_stats *rdma_alloc_hw_stats_struct(
472 const char * const *names, int num_counters,
473 unsigned long lifespan)
474{
475 struct rdma_hw_stats *stats;
476
477 stats = kzalloc(sizeof(*stats) + num_counters * sizeof(u64),
478 GFP_KERNEL);
479 if (!stats)
480 return NULL;
481 stats->names = names;
482 stats->num_counters = num_counters;
483 stats->lifespan = msecs_to_jiffies(lifespan);
484
485 return stats;
486}
487
488
f9b22e35
IW
489/* Define bits for the various functionality this port needs to be supported by
490 * the core.
491 */
492/* Management 0x00000FFF */
493#define RDMA_CORE_CAP_IB_MAD 0x00000001
494#define RDMA_CORE_CAP_IB_SMI 0x00000002
495#define RDMA_CORE_CAP_IB_CM 0x00000004
496#define RDMA_CORE_CAP_IW_CM 0x00000008
497#define RDMA_CORE_CAP_IB_SA 0x00000010
65995fee 498#define RDMA_CORE_CAP_OPA_MAD 0x00000020
f9b22e35
IW
499
500/* Address format 0x000FF000 */
501#define RDMA_CORE_CAP_AF_IB 0x00001000
502#define RDMA_CORE_CAP_ETH_AH 0x00002000
503
504/* Protocol 0xFFF00000 */
505#define RDMA_CORE_CAP_PROT_IB 0x00100000
506#define RDMA_CORE_CAP_PROT_ROCE 0x00200000
507#define RDMA_CORE_CAP_PROT_IWARP 0x00400000
7766a99f 508#define RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP 0x00800000
aa773bd4 509#define RDMA_CORE_CAP_PROT_RAW_PACKET 0x01000000
ce1e055f 510#define RDMA_CORE_CAP_PROT_USNIC 0x02000000
f9b22e35
IW
511
512#define RDMA_CORE_PORT_IBA_IB (RDMA_CORE_CAP_PROT_IB \
513 | RDMA_CORE_CAP_IB_MAD \
514 | RDMA_CORE_CAP_IB_SMI \
515 | RDMA_CORE_CAP_IB_CM \
516 | RDMA_CORE_CAP_IB_SA \
517 | RDMA_CORE_CAP_AF_IB)
518#define RDMA_CORE_PORT_IBA_ROCE (RDMA_CORE_CAP_PROT_ROCE \
519 | RDMA_CORE_CAP_IB_MAD \
520 | RDMA_CORE_CAP_IB_CM \
f9b22e35
IW
521 | RDMA_CORE_CAP_AF_IB \
522 | RDMA_CORE_CAP_ETH_AH)
7766a99f
MB
523#define RDMA_CORE_PORT_IBA_ROCE_UDP_ENCAP \
524 (RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP \
525 | RDMA_CORE_CAP_IB_MAD \
526 | RDMA_CORE_CAP_IB_CM \
527 | RDMA_CORE_CAP_AF_IB \
528 | RDMA_CORE_CAP_ETH_AH)
f9b22e35
IW
529#define RDMA_CORE_PORT_IWARP (RDMA_CORE_CAP_PROT_IWARP \
530 | RDMA_CORE_CAP_IW_CM)
65995fee
IW
531#define RDMA_CORE_PORT_INTEL_OPA (RDMA_CORE_PORT_IBA_IB \
532 | RDMA_CORE_CAP_OPA_MAD)
f9b22e35 533
aa773bd4
OG
534#define RDMA_CORE_PORT_RAW_PACKET (RDMA_CORE_CAP_PROT_RAW_PACKET)
535
ce1e055f
OG
536#define RDMA_CORE_PORT_USNIC (RDMA_CORE_CAP_PROT_USNIC)
537
1da177e4 538struct ib_port_attr {
fad61ad4 539 u64 subnet_prefix;
1da177e4
LT
540 enum ib_port_state state;
541 enum ib_mtu max_mtu;
542 enum ib_mtu active_mtu;
543 int gid_tbl_len;
544 u32 port_cap_flags;
545 u32 max_msg_sz;
546 u32 bad_pkey_cntr;
547 u32 qkey_viol_cntr;
548 u16 pkey_tbl_len;
549 u16 lid;
550 u16 sm_lid;
551 u8 lmc;
552 u8 max_vl_num;
553 u8 sm_sl;
554 u8 subnet_timeout;
555 u8 init_type_reply;
556 u8 active_width;
557 u8 active_speed;
558 u8 phys_state;
a0c1b2a3 559 bool grh_required;
1da177e4
LT
560};
561
562enum ib_device_modify_flags {
c5bcbbb9
RD
563 IB_DEVICE_MODIFY_SYS_IMAGE_GUID = 1 << 0,
564 IB_DEVICE_MODIFY_NODE_DESC = 1 << 1
1da177e4
LT
565};
566
bd99fdea
YS
567#define IB_DEVICE_NODE_DESC_MAX 64
568
1da177e4
LT
569struct ib_device_modify {
570 u64 sys_image_guid;
bd99fdea 571 char node_desc[IB_DEVICE_NODE_DESC_MAX];
1da177e4
LT
572};
573
574enum ib_port_modify_flags {
575 IB_PORT_SHUTDOWN = 1,
576 IB_PORT_INIT_TYPE = (1<<2),
577 IB_PORT_RESET_QKEY_CNTR = (1<<3)
578};
579
580struct ib_port_modify {
581 u32 set_port_cap_mask;
582 u32 clr_port_cap_mask;
583 u8 init_type;
584};
585
586enum ib_event_type {
587 IB_EVENT_CQ_ERR,
588 IB_EVENT_QP_FATAL,
589 IB_EVENT_QP_REQ_ERR,
590 IB_EVENT_QP_ACCESS_ERR,
591 IB_EVENT_COMM_EST,
592 IB_EVENT_SQ_DRAINED,
593 IB_EVENT_PATH_MIG,
594 IB_EVENT_PATH_MIG_ERR,
595 IB_EVENT_DEVICE_FATAL,
596 IB_EVENT_PORT_ACTIVE,
597 IB_EVENT_PORT_ERR,
598 IB_EVENT_LID_CHANGE,
599 IB_EVENT_PKEY_CHANGE,
d41fcc67
RD
600 IB_EVENT_SM_CHANGE,
601 IB_EVENT_SRQ_ERR,
602 IB_EVENT_SRQ_LIMIT_REACHED,
63942c9a 603 IB_EVENT_QP_LAST_WQE_REACHED,
761d90ed
OG
604 IB_EVENT_CLIENT_REREGISTER,
605 IB_EVENT_GID_CHANGE,
f213c052 606 IB_EVENT_WQ_FATAL,
1da177e4
LT
607};
608
db7489e0 609const char *__attribute_const__ ib_event_msg(enum ib_event_type event);
2b1b5b60 610
1da177e4
LT
611struct ib_event {
612 struct ib_device *device;
613 union {
614 struct ib_cq *cq;
615 struct ib_qp *qp;
d41fcc67 616 struct ib_srq *srq;
f213c052 617 struct ib_wq *wq;
1da177e4
LT
618 u8 port_num;
619 } element;
620 enum ib_event_type event;
621};
622
623struct ib_event_handler {
624 struct ib_device *device;
625 void (*handler)(struct ib_event_handler *, struct ib_event *);
626 struct list_head list;
627};
628
629#define INIT_IB_EVENT_HANDLER(_ptr, _device, _handler) \
630 do { \
631 (_ptr)->device = _device; \
632 (_ptr)->handler = _handler; \
633 INIT_LIST_HEAD(&(_ptr)->list); \
634 } while (0)
635
636struct ib_global_route {
637 union ib_gid dgid;
638 u32 flow_label;
639 u8 sgid_index;
640 u8 hop_limit;
641 u8 traffic_class;
642};
643
513789ed 644struct ib_grh {
97f52eb4
SH
645 __be32 version_tclass_flow;
646 __be16 paylen;
513789ed
HR
647 u8 next_hdr;
648 u8 hop_limit;
649 union ib_gid sgid;
650 union ib_gid dgid;
651};
652
c865f246
SK
653union rdma_network_hdr {
654 struct ib_grh ibgrh;
655 struct {
656 /* The IB spec states that if it's IPv4, the header
657 * is located in the last 20 bytes of the header.
658 */
659 u8 reserved[20];
660 struct iphdr roce4grh;
661 };
662};
663
1da177e4
LT
664enum {
665 IB_MULTICAST_QPN = 0xffffff
666};
667
f3a7c66b 668#define IB_LID_PERMISSIVE cpu_to_be16(0xFFFF)
b4e64397 669#define IB_MULTICAST_LID_BASE cpu_to_be16(0xC000)
97f52eb4 670
1da177e4
LT
671enum ib_ah_flags {
672 IB_AH_GRH = 1
673};
674
bf6a9e31
JM
675enum ib_rate {
676 IB_RATE_PORT_CURRENT = 0,
677 IB_RATE_2_5_GBPS = 2,
678 IB_RATE_5_GBPS = 5,
679 IB_RATE_10_GBPS = 3,
680 IB_RATE_20_GBPS = 6,
681 IB_RATE_30_GBPS = 4,
682 IB_RATE_40_GBPS = 7,
683 IB_RATE_60_GBPS = 8,
684 IB_RATE_80_GBPS = 9,
71eeba16
MA
685 IB_RATE_120_GBPS = 10,
686 IB_RATE_14_GBPS = 11,
687 IB_RATE_56_GBPS = 12,
688 IB_RATE_112_GBPS = 13,
689 IB_RATE_168_GBPS = 14,
690 IB_RATE_25_GBPS = 15,
691 IB_RATE_100_GBPS = 16,
692 IB_RATE_200_GBPS = 17,
693 IB_RATE_300_GBPS = 18
bf6a9e31
JM
694};
695
696/**
697 * ib_rate_to_mult - Convert the IB rate enum to a multiple of the
698 * base rate of 2.5 Gbit/sec. For example, IB_RATE_5_GBPS will be
699 * converted to 2, since 5 Gbit/sec is 2 * 2.5 Gbit/sec.
700 * @rate: rate to convert.
701 */
8385fd84 702__attribute_const__ int ib_rate_to_mult(enum ib_rate rate);
bf6a9e31 703
71eeba16
MA
704/**
705 * ib_rate_to_mbps - Convert the IB rate enum to Mbps.
706 * For example, IB_RATE_2_5_GBPS will be converted to 2500.
707 * @rate: rate to convert.
708 */
8385fd84 709__attribute_const__ int ib_rate_to_mbps(enum ib_rate rate);
71eeba16 710
17cd3a2d
SG
711
712/**
9bee178b
SG
713 * enum ib_mr_type - memory region type
714 * @IB_MR_TYPE_MEM_REG: memory region that is used for
715 * normal registration
716 * @IB_MR_TYPE_SIGNATURE: memory region that is used for
717 * signature operations (data-integrity
718 * capable regions)
f5aa9159
SG
719 * @IB_MR_TYPE_SG_GAPS: memory region that is capable to
720 * register any arbitrary sg lists (without
721 * the normal mr constraints - see
722 * ib_map_mr_sg)
17cd3a2d 723 */
9bee178b
SG
724enum ib_mr_type {
725 IB_MR_TYPE_MEM_REG,
726 IB_MR_TYPE_SIGNATURE,
f5aa9159 727 IB_MR_TYPE_SG_GAPS,
17cd3a2d
SG
728};
729
1b01d335 730/**
78eda2bb
SG
731 * Signature types
732 * IB_SIG_TYPE_NONE: Unprotected.
733 * IB_SIG_TYPE_T10_DIF: Type T10-DIF
1b01d335 734 */
78eda2bb
SG
735enum ib_signature_type {
736 IB_SIG_TYPE_NONE,
737 IB_SIG_TYPE_T10_DIF,
1b01d335
SG
738};
739
740/**
741 * Signature T10-DIF block-guard types
742 * IB_T10DIF_CRC: Corresponds to T10-PI mandated CRC checksum rules.
743 * IB_T10DIF_CSUM: Corresponds to IP checksum rules.
744 */
745enum ib_t10_dif_bg_type {
746 IB_T10DIF_CRC,
747 IB_T10DIF_CSUM
748};
749
750/**
751 * struct ib_t10_dif_domain - Parameters specific for T10-DIF
752 * domain.
1b01d335
SG
753 * @bg_type: T10-DIF block guard type (CRC|CSUM)
754 * @pi_interval: protection information interval.
755 * @bg: seed of guard computation.
756 * @app_tag: application tag of guard block
757 * @ref_tag: initial guard block reference tag.
78eda2bb
SG
758 * @ref_remap: Indicate wethear the reftag increments each block
759 * @app_escape: Indicate to skip block check if apptag=0xffff
760 * @ref_escape: Indicate to skip block check if reftag=0xffffffff
761 * @apptag_check_mask: check bitmask of application tag.
1b01d335
SG
762 */
763struct ib_t10_dif_domain {
1b01d335
SG
764 enum ib_t10_dif_bg_type bg_type;
765 u16 pi_interval;
766 u16 bg;
767 u16 app_tag;
768 u32 ref_tag;
78eda2bb
SG
769 bool ref_remap;
770 bool app_escape;
771 bool ref_escape;
772 u16 apptag_check_mask;
1b01d335
SG
773};
774
775/**
776 * struct ib_sig_domain - Parameters for signature domain
777 * @sig_type: specific signauture type
778 * @sig: union of all signature domain attributes that may
779 * be used to set domain layout.
780 */
781struct ib_sig_domain {
782 enum ib_signature_type sig_type;
783 union {
784 struct ib_t10_dif_domain dif;
785 } sig;
786};
787
788/**
789 * struct ib_sig_attrs - Parameters for signature handover operation
790 * @check_mask: bitmask for signature byte check (8 bytes)
791 * @mem: memory domain layout desciptor.
792 * @wire: wire domain layout desciptor.
793 */
794struct ib_sig_attrs {
795 u8 check_mask;
796 struct ib_sig_domain mem;
797 struct ib_sig_domain wire;
798};
799
800enum ib_sig_err_type {
801 IB_SIG_BAD_GUARD,
802 IB_SIG_BAD_REFTAG,
803 IB_SIG_BAD_APPTAG,
804};
805
806/**
807 * struct ib_sig_err - signature error descriptor
808 */
809struct ib_sig_err {
810 enum ib_sig_err_type err_type;
811 u32 expected;
812 u32 actual;
813 u64 sig_err_offset;
814 u32 key;
815};
816
817enum ib_mr_status_check {
818 IB_MR_CHECK_SIG_STATUS = 1,
819};
820
821/**
822 * struct ib_mr_status - Memory region status container
823 *
824 * @fail_status: Bitmask of MR checks status. For each
825 * failed check a corresponding status bit is set.
826 * @sig_err: Additional info for IB_MR_CEHCK_SIG_STATUS
827 * failure.
828 */
829struct ib_mr_status {
830 u32 fail_status;
831 struct ib_sig_err sig_err;
832};
833
bf6a9e31
JM
834/**
835 * mult_to_ib_rate - Convert a multiple of 2.5 Gbit/sec to an IB rate
836 * enum.
837 * @mult: multiple to convert.
838 */
8385fd84 839__attribute_const__ enum ib_rate mult_to_ib_rate(int mult);
bf6a9e31 840
1da177e4
LT
841struct ib_ah_attr {
842 struct ib_global_route grh;
843 u16 dlid;
844 u8 sl;
845 u8 src_path_bits;
846 u8 static_rate;
847 u8 ah_flags;
848 u8 port_num;
dd5f03be 849 u8 dmac[ETH_ALEN];
1da177e4
LT
850};
851
852enum ib_wc_status {
853 IB_WC_SUCCESS,
854 IB_WC_LOC_LEN_ERR,
855 IB_WC_LOC_QP_OP_ERR,
856 IB_WC_LOC_EEC_OP_ERR,
857 IB_WC_LOC_PROT_ERR,
858 IB_WC_WR_FLUSH_ERR,
859 IB_WC_MW_BIND_ERR,
860 IB_WC_BAD_RESP_ERR,
861 IB_WC_LOC_ACCESS_ERR,
862 IB_WC_REM_INV_REQ_ERR,
863 IB_WC_REM_ACCESS_ERR,
864 IB_WC_REM_OP_ERR,
865 IB_WC_RETRY_EXC_ERR,
866 IB_WC_RNR_RETRY_EXC_ERR,
867 IB_WC_LOC_RDD_VIOL_ERR,
868 IB_WC_REM_INV_RD_REQ_ERR,
869 IB_WC_REM_ABORT_ERR,
870 IB_WC_INV_EECN_ERR,
871 IB_WC_INV_EEC_STATE_ERR,
872 IB_WC_FATAL_ERR,
873 IB_WC_RESP_TIMEOUT_ERR,
874 IB_WC_GENERAL_ERR
875};
876
db7489e0 877const char *__attribute_const__ ib_wc_status_msg(enum ib_wc_status status);
2b1b5b60 878
1da177e4
LT
879enum ib_wc_opcode {
880 IB_WC_SEND,
881 IB_WC_RDMA_WRITE,
882 IB_WC_RDMA_READ,
883 IB_WC_COMP_SWAP,
884 IB_WC_FETCH_ADD,
c93570f2 885 IB_WC_LSO,
00f7ec36 886 IB_WC_LOCAL_INV,
4c67e2bf 887 IB_WC_REG_MR,
5e80ba8f
VS
888 IB_WC_MASKED_COMP_SWAP,
889 IB_WC_MASKED_FETCH_ADD,
1da177e4
LT
890/*
891 * Set value of IB_WC_RECV so consumers can test if a completion is a
892 * receive by testing (opcode & IB_WC_RECV).
893 */
894 IB_WC_RECV = 1 << 7,
895 IB_WC_RECV_RDMA_WITH_IMM
896};
897
898enum ib_wc_flags {
899 IB_WC_GRH = 1,
00f7ec36
SW
900 IB_WC_WITH_IMM = (1<<1),
901 IB_WC_WITH_INVALIDATE = (1<<2),
d927d505 902 IB_WC_IP_CSUM_OK = (1<<3),
dd5f03be
MB
903 IB_WC_WITH_SMAC = (1<<4),
904 IB_WC_WITH_VLAN = (1<<5),
c865f246 905 IB_WC_WITH_NETWORK_HDR_TYPE = (1<<6),
1da177e4
LT
906};
907
908struct ib_wc {
14d3a3b2
CH
909 union {
910 u64 wr_id;
911 struct ib_cqe *wr_cqe;
912 };
1da177e4
LT
913 enum ib_wc_status status;
914 enum ib_wc_opcode opcode;
915 u32 vendor_err;
916 u32 byte_len;
062dbb69 917 struct ib_qp *qp;
00f7ec36
SW
918 union {
919 __be32 imm_data;
920 u32 invalidate_rkey;
921 } ex;
1da177e4
LT
922 u32 src_qp;
923 int wc_flags;
924 u16 pkey_index;
925 u16 slid;
926 u8 sl;
927 u8 dlid_path_bits;
928 u8 port_num; /* valid only for DR SMPs on switches */
dd5f03be
MB
929 u8 smac[ETH_ALEN];
930 u16 vlan_id;
c865f246 931 u8 network_hdr_type;
1da177e4
LT
932};
933
ed23a727
RD
934enum ib_cq_notify_flags {
935 IB_CQ_SOLICITED = 1 << 0,
936 IB_CQ_NEXT_COMP = 1 << 1,
937 IB_CQ_SOLICITED_MASK = IB_CQ_SOLICITED | IB_CQ_NEXT_COMP,
938 IB_CQ_REPORT_MISSED_EVENTS = 1 << 2,
1da177e4
LT
939};
940
96104eda 941enum ib_srq_type {
418d5130
SH
942 IB_SRQT_BASIC,
943 IB_SRQT_XRC
96104eda
SH
944};
945
d41fcc67
RD
946enum ib_srq_attr_mask {
947 IB_SRQ_MAX_WR = 1 << 0,
948 IB_SRQ_LIMIT = 1 << 1,
949};
950
951struct ib_srq_attr {
952 u32 max_wr;
953 u32 max_sge;
954 u32 srq_limit;
955};
956
957struct ib_srq_init_attr {
958 void (*event_handler)(struct ib_event *, void *);
959 void *srq_context;
960 struct ib_srq_attr attr;
96104eda 961 enum ib_srq_type srq_type;
418d5130
SH
962
963 union {
964 struct {
965 struct ib_xrcd *xrcd;
966 struct ib_cq *cq;
967 } xrc;
968 } ext;
d41fcc67
RD
969};
970
1da177e4
LT
971struct ib_qp_cap {
972 u32 max_send_wr;
973 u32 max_recv_wr;
974 u32 max_send_sge;
975 u32 max_recv_sge;
976 u32 max_inline_data;
a060b562
CH
977
978 /*
979 * Maximum number of rdma_rw_ctx structures in flight at a time.
980 * ib_create_qp() will calculate the right amount of neededed WRs
981 * and MRs based on this.
982 */
983 u32 max_rdma_ctxs;
1da177e4
LT
984};
985
986enum ib_sig_type {
987 IB_SIGNAL_ALL_WR,
988 IB_SIGNAL_REQ_WR
989};
990
991enum ib_qp_type {
992 /*
993 * IB_QPT_SMI and IB_QPT_GSI have to be the first two entries
994 * here (and in that order) since the MAD layer uses them as
995 * indices into a 2-entry table.
996 */
997 IB_QPT_SMI,
998 IB_QPT_GSI,
999
1000 IB_QPT_RC,
1001 IB_QPT_UC,
1002 IB_QPT_UD,
1003 IB_QPT_RAW_IPV6,
b42b63cf 1004 IB_QPT_RAW_ETHERTYPE,
c938a616 1005 IB_QPT_RAW_PACKET = 8,
b42b63cf
SH
1006 IB_QPT_XRC_INI = 9,
1007 IB_QPT_XRC_TGT,
0134f16b
JM
1008 IB_QPT_MAX,
1009 /* Reserve a range for qp types internal to the low level driver.
1010 * These qp types will not be visible at the IB core layer, so the
1011 * IB_QPT_MAX usages should not be affected in the core layer
1012 */
1013 IB_QPT_RESERVED1 = 0x1000,
1014 IB_QPT_RESERVED2,
1015 IB_QPT_RESERVED3,
1016 IB_QPT_RESERVED4,
1017 IB_QPT_RESERVED5,
1018 IB_QPT_RESERVED6,
1019 IB_QPT_RESERVED7,
1020 IB_QPT_RESERVED8,
1021 IB_QPT_RESERVED9,
1022 IB_QPT_RESERVED10,
1da177e4
LT
1023};
1024
b846f25a 1025enum ib_qp_create_flags {
47ee1b9f
RL
1026 IB_QP_CREATE_IPOIB_UD_LSO = 1 << 0,
1027 IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK = 1 << 1,
8a06ce59
LR
1028 IB_QP_CREATE_CROSS_CHANNEL = 1 << 2,
1029 IB_QP_CREATE_MANAGED_SEND = 1 << 3,
1030 IB_QP_CREATE_MANAGED_RECV = 1 << 4,
90f1d1b4 1031 IB_QP_CREATE_NETIF_QP = 1 << 5,
1b01d335 1032 IB_QP_CREATE_SIGNATURE_EN = 1 << 6,
09b93088 1033 IB_QP_CREATE_USE_GFP_NOIO = 1 << 7,
b531b909 1034 IB_QP_CREATE_SCATTER_FCS = 1 << 8,
9c2b270e 1035 IB_QP_CREATE_CVLAN_STRIPPING = 1 << 9,
d2b57063
JM
1036 /* reserve bits 26-31 for low level drivers' internal use */
1037 IB_QP_CREATE_RESERVED_START = 1 << 26,
1038 IB_QP_CREATE_RESERVED_END = 1 << 31,
b846f25a
EC
1039};
1040
73c40c61
YH
1041/*
1042 * Note: users may not call ib_close_qp or ib_destroy_qp from the event_handler
1043 * callback to destroy the passed in QP.
1044 */
1045
1da177e4
LT
1046struct ib_qp_init_attr {
1047 void (*event_handler)(struct ib_event *, void *);
1048 void *qp_context;
1049 struct ib_cq *send_cq;
1050 struct ib_cq *recv_cq;
1051 struct ib_srq *srq;
b42b63cf 1052 struct ib_xrcd *xrcd; /* XRC TGT QPs only */
1da177e4
LT
1053 struct ib_qp_cap cap;
1054 enum ib_sig_type sq_sig_type;
1055 enum ib_qp_type qp_type;
b846f25a 1056 enum ib_qp_create_flags create_flags;
a060b562
CH
1057
1058 /*
1059 * Only needed for special QP types, or when using the RW API.
1060 */
1061 u8 port_num;
a9017e23 1062 struct ib_rwq_ind_table *rwq_ind_tbl;
1da177e4
LT
1063};
1064
0e0ec7e0
SH
1065struct ib_qp_open_attr {
1066 void (*event_handler)(struct ib_event *, void *);
1067 void *qp_context;
1068 u32 qp_num;
1069 enum ib_qp_type qp_type;
1070};
1071
1da177e4
LT
1072enum ib_rnr_timeout {
1073 IB_RNR_TIMER_655_36 = 0,
1074 IB_RNR_TIMER_000_01 = 1,
1075 IB_RNR_TIMER_000_02 = 2,
1076 IB_RNR_TIMER_000_03 = 3,
1077 IB_RNR_TIMER_000_04 = 4,
1078 IB_RNR_TIMER_000_06 = 5,
1079 IB_RNR_TIMER_000_08 = 6,
1080 IB_RNR_TIMER_000_12 = 7,
1081 IB_RNR_TIMER_000_16 = 8,
1082 IB_RNR_TIMER_000_24 = 9,
1083 IB_RNR_TIMER_000_32 = 10,
1084 IB_RNR_TIMER_000_48 = 11,
1085 IB_RNR_TIMER_000_64 = 12,
1086 IB_RNR_TIMER_000_96 = 13,
1087 IB_RNR_TIMER_001_28 = 14,
1088 IB_RNR_TIMER_001_92 = 15,
1089 IB_RNR_TIMER_002_56 = 16,
1090 IB_RNR_TIMER_003_84 = 17,
1091 IB_RNR_TIMER_005_12 = 18,
1092 IB_RNR_TIMER_007_68 = 19,
1093 IB_RNR_TIMER_010_24 = 20,
1094 IB_RNR_TIMER_015_36 = 21,
1095 IB_RNR_TIMER_020_48 = 22,
1096 IB_RNR_TIMER_030_72 = 23,
1097 IB_RNR_TIMER_040_96 = 24,
1098 IB_RNR_TIMER_061_44 = 25,
1099 IB_RNR_TIMER_081_92 = 26,
1100 IB_RNR_TIMER_122_88 = 27,
1101 IB_RNR_TIMER_163_84 = 28,
1102 IB_RNR_TIMER_245_76 = 29,
1103 IB_RNR_TIMER_327_68 = 30,
1104 IB_RNR_TIMER_491_52 = 31
1105};
1106
1107enum ib_qp_attr_mask {
1108 IB_QP_STATE = 1,
1109 IB_QP_CUR_STATE = (1<<1),
1110 IB_QP_EN_SQD_ASYNC_NOTIFY = (1<<2),
1111 IB_QP_ACCESS_FLAGS = (1<<3),
1112 IB_QP_PKEY_INDEX = (1<<4),
1113 IB_QP_PORT = (1<<5),
1114 IB_QP_QKEY = (1<<6),
1115 IB_QP_AV = (1<<7),
1116 IB_QP_PATH_MTU = (1<<8),
1117 IB_QP_TIMEOUT = (1<<9),
1118 IB_QP_RETRY_CNT = (1<<10),
1119 IB_QP_RNR_RETRY = (1<<11),
1120 IB_QP_RQ_PSN = (1<<12),
1121 IB_QP_MAX_QP_RD_ATOMIC = (1<<13),
1122 IB_QP_ALT_PATH = (1<<14),
1123 IB_QP_MIN_RNR_TIMER = (1<<15),
1124 IB_QP_SQ_PSN = (1<<16),
1125 IB_QP_MAX_DEST_RD_ATOMIC = (1<<17),
1126 IB_QP_PATH_MIG_STATE = (1<<18),
1127 IB_QP_CAP = (1<<19),
dd5f03be 1128 IB_QP_DEST_QPN = (1<<20),
aa744cc0
MB
1129 IB_QP_RESERVED1 = (1<<21),
1130 IB_QP_RESERVED2 = (1<<22),
1131 IB_QP_RESERVED3 = (1<<23),
1132 IB_QP_RESERVED4 = (1<<24),
528e5a1b 1133 IB_QP_RATE_LIMIT = (1<<25),
1da177e4
LT
1134};
1135
1136enum ib_qp_state {
1137 IB_QPS_RESET,
1138 IB_QPS_INIT,
1139 IB_QPS_RTR,
1140 IB_QPS_RTS,
1141 IB_QPS_SQD,
1142 IB_QPS_SQE,
1143 IB_QPS_ERR
1144};
1145
1146enum ib_mig_state {
1147 IB_MIG_MIGRATED,
1148 IB_MIG_REARM,
1149 IB_MIG_ARMED
1150};
1151
7083e42e
SM
1152enum ib_mw_type {
1153 IB_MW_TYPE_1 = 1,
1154 IB_MW_TYPE_2 = 2
1155};
1156
1da177e4
LT
1157struct ib_qp_attr {
1158 enum ib_qp_state qp_state;
1159 enum ib_qp_state cur_qp_state;
1160 enum ib_mtu path_mtu;
1161 enum ib_mig_state path_mig_state;
1162 u32 qkey;
1163 u32 rq_psn;
1164 u32 sq_psn;
1165 u32 dest_qp_num;
1166 int qp_access_flags;
1167 struct ib_qp_cap cap;
1168 struct ib_ah_attr ah_attr;
1169 struct ib_ah_attr alt_ah_attr;
1170 u16 pkey_index;
1171 u16 alt_pkey_index;
1172 u8 en_sqd_async_notify;
1173 u8 sq_draining;
1174 u8 max_rd_atomic;
1175 u8 max_dest_rd_atomic;
1176 u8 min_rnr_timer;
1177 u8 port_num;
1178 u8 timeout;
1179 u8 retry_cnt;
1180 u8 rnr_retry;
1181 u8 alt_port_num;
1182 u8 alt_timeout;
528e5a1b 1183 u32 rate_limit;
1da177e4
LT
1184};
1185
1186enum ib_wr_opcode {
1187 IB_WR_RDMA_WRITE,
1188 IB_WR_RDMA_WRITE_WITH_IMM,
1189 IB_WR_SEND,
1190 IB_WR_SEND_WITH_IMM,
1191 IB_WR_RDMA_READ,
1192 IB_WR_ATOMIC_CMP_AND_SWP,
c93570f2 1193 IB_WR_ATOMIC_FETCH_AND_ADD,
0f39cf3d
RD
1194 IB_WR_LSO,
1195 IB_WR_SEND_WITH_INV,
00f7ec36
SW
1196 IB_WR_RDMA_READ_WITH_INV,
1197 IB_WR_LOCAL_INV,
4c67e2bf 1198 IB_WR_REG_MR,
5e80ba8f
VS
1199 IB_WR_MASKED_ATOMIC_CMP_AND_SWP,
1200 IB_WR_MASKED_ATOMIC_FETCH_AND_ADD,
1b01d335 1201 IB_WR_REG_SIG_MR,
0134f16b
JM
1202 /* reserve values for low level drivers' internal use.
1203 * These values will not be used at all in the ib core layer.
1204 */
1205 IB_WR_RESERVED1 = 0xf0,
1206 IB_WR_RESERVED2,
1207 IB_WR_RESERVED3,
1208 IB_WR_RESERVED4,
1209 IB_WR_RESERVED5,
1210 IB_WR_RESERVED6,
1211 IB_WR_RESERVED7,
1212 IB_WR_RESERVED8,
1213 IB_WR_RESERVED9,
1214 IB_WR_RESERVED10,
1da177e4
LT
1215};
1216
1217enum ib_send_flags {
1218 IB_SEND_FENCE = 1,
1219 IB_SEND_SIGNALED = (1<<1),
1220 IB_SEND_SOLICITED = (1<<2),
e0605d91 1221 IB_SEND_INLINE = (1<<3),
0134f16b
JM
1222 IB_SEND_IP_CSUM = (1<<4),
1223
1224 /* reserve bits 26-31 for low level drivers' internal use */
1225 IB_SEND_RESERVED_START = (1 << 26),
1226 IB_SEND_RESERVED_END = (1 << 31),
1da177e4
LT
1227};
1228
1229struct ib_sge {
1230 u64 addr;
1231 u32 length;
1232 u32 lkey;
1233};
1234
14d3a3b2
CH
1235struct ib_cqe {
1236 void (*done)(struct ib_cq *cq, struct ib_wc *wc);
1237};
1238
1da177e4
LT
1239struct ib_send_wr {
1240 struct ib_send_wr *next;
14d3a3b2
CH
1241 union {
1242 u64 wr_id;
1243 struct ib_cqe *wr_cqe;
1244 };
1da177e4
LT
1245 struct ib_sge *sg_list;
1246 int num_sge;
1247 enum ib_wr_opcode opcode;
1248 int send_flags;
0f39cf3d
RD
1249 union {
1250 __be32 imm_data;
1251 u32 invalidate_rkey;
1252 } ex;
1da177e4
LT
1253};
1254
e622f2f4
CH
1255struct ib_rdma_wr {
1256 struct ib_send_wr wr;
1257 u64 remote_addr;
1258 u32 rkey;
1259};
1260
1261static inline struct ib_rdma_wr *rdma_wr(struct ib_send_wr *wr)
1262{
1263 return container_of(wr, struct ib_rdma_wr, wr);
1264}
1265
1266struct ib_atomic_wr {
1267 struct ib_send_wr wr;
1268 u64 remote_addr;
1269 u64 compare_add;
1270 u64 swap;
1271 u64 compare_add_mask;
1272 u64 swap_mask;
1273 u32 rkey;
1274};
1275
1276static inline struct ib_atomic_wr *atomic_wr(struct ib_send_wr *wr)
1277{
1278 return container_of(wr, struct ib_atomic_wr, wr);
1279}
1280
1281struct ib_ud_wr {
1282 struct ib_send_wr wr;
1283 struct ib_ah *ah;
1284 void *header;
1285 int hlen;
1286 int mss;
1287 u32 remote_qpn;
1288 u32 remote_qkey;
1289 u16 pkey_index; /* valid for GSI only */
1290 u8 port_num; /* valid for DR SMPs on switch only */
1291};
1292
1293static inline struct ib_ud_wr *ud_wr(struct ib_send_wr *wr)
1294{
1295 return container_of(wr, struct ib_ud_wr, wr);
1296}
1297
4c67e2bf
SG
1298struct ib_reg_wr {
1299 struct ib_send_wr wr;
1300 struct ib_mr *mr;
1301 u32 key;
1302 int access;
1303};
1304
1305static inline struct ib_reg_wr *reg_wr(struct ib_send_wr *wr)
1306{
1307 return container_of(wr, struct ib_reg_wr, wr);
1308}
1309
e622f2f4
CH
1310struct ib_sig_handover_wr {
1311 struct ib_send_wr wr;
1312 struct ib_sig_attrs *sig_attrs;
1313 struct ib_mr *sig_mr;
1314 int access_flags;
1315 struct ib_sge *prot;
1316};
1317
1318static inline struct ib_sig_handover_wr *sig_handover_wr(struct ib_send_wr *wr)
1319{
1320 return container_of(wr, struct ib_sig_handover_wr, wr);
1321}
1322
1da177e4
LT
1323struct ib_recv_wr {
1324 struct ib_recv_wr *next;
14d3a3b2
CH
1325 union {
1326 u64 wr_id;
1327 struct ib_cqe *wr_cqe;
1328 };
1da177e4
LT
1329 struct ib_sge *sg_list;
1330 int num_sge;
1331};
1332
1333enum ib_access_flags {
1334 IB_ACCESS_LOCAL_WRITE = 1,
1335 IB_ACCESS_REMOTE_WRITE = (1<<1),
1336 IB_ACCESS_REMOTE_READ = (1<<2),
1337 IB_ACCESS_REMOTE_ATOMIC = (1<<3),
7083e42e 1338 IB_ACCESS_MW_BIND = (1<<4),
860f10a7
SG
1339 IB_ZERO_BASED = (1<<5),
1340 IB_ACCESS_ON_DEMAND = (1<<6),
1da177e4
LT
1341};
1342
b7d3e0a9
CH
1343/*
1344 * XXX: these are apparently used for ->rereg_user_mr, no idea why they
1345 * are hidden here instead of a uapi header!
1346 */
1da177e4
LT
1347enum ib_mr_rereg_flags {
1348 IB_MR_REREG_TRANS = 1,
1349 IB_MR_REREG_PD = (1<<1),
7e6edb9b
MB
1350 IB_MR_REREG_ACCESS = (1<<2),
1351 IB_MR_REREG_SUPPORTED = ((IB_MR_REREG_ACCESS << 1) - 1)
1da177e4
LT
1352};
1353
1da177e4
LT
1354struct ib_fmr_attr {
1355 int max_pages;
1356 int max_maps;
d36f34aa 1357 u8 page_shift;
1da177e4
LT
1358};
1359
882214e2
HE
1360struct ib_umem;
1361
38321256
MB
1362enum rdma_remove_reason {
1363 /* Userspace requested uobject deletion. Call could fail */
1364 RDMA_REMOVE_DESTROY,
1365 /* Context deletion. This call should delete the actual object itself */
1366 RDMA_REMOVE_CLOSE,
1367 /* Driver is being hot-unplugged. This call should delete the actual object itself */
1368 RDMA_REMOVE_DRIVER_REMOVE,
1369 /* Context is being cleaned-up, but commit was just completed */
1370 RDMA_REMOVE_DURING_CLEANUP,
1371};
1372
43579b5f
PP
1373struct ib_rdmacg_object {
1374#ifdef CONFIG_CGROUP_RDMA
1375 struct rdma_cgroup *cg; /* owner rdma cgroup */
1376#endif
1377};
1378
e2773c06
RD
1379struct ib_ucontext {
1380 struct ib_device *device;
771addf6 1381 struct ib_uverbs_file *ufile;
f7c6a7b5 1382 int closing;
8ada2c1c 1383
38321256
MB
1384 /* locking the uobjects_list */
1385 struct mutex uobjects_lock;
1386 struct list_head uobjects;
1387 /* protects cleanup process from other actions */
1388 struct rw_semaphore cleanup_rwsem;
1389 enum rdma_remove_reason cleanup_reason;
1390
8ada2c1c 1391 struct pid *tgid;
882214e2
HE
1392#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
1393 struct rb_root umem_tree;
1394 /*
1395 * Protects .umem_rbroot and tree, as well as odp_mrs_count and
1396 * mmu notifiers registration.
1397 */
1398 struct rw_semaphore umem_rwsem;
1399 void (*invalidate_range)(struct ib_umem *umem,
1400 unsigned long start, unsigned long end);
1401
1402 struct mmu_notifier mn;
1403 atomic_t notifier_count;
1404 /* A list of umems that don't have private mmu notifier counters yet. */
1405 struct list_head no_private_counters;
1406 int odp_mrs_count;
1407#endif
43579b5f
PP
1408
1409 struct ib_rdmacg_object cg_obj;
e2773c06
RD
1410};
1411
1412struct ib_uobject {
1413 u64 user_handle; /* handle given to us by userspace */
1414 struct ib_ucontext *context; /* associated user context */
9ead190b 1415 void *object; /* containing object */
e2773c06 1416 struct list_head list; /* link to context's list */
43579b5f 1417 struct ib_rdmacg_object cg_obj; /* rdmacg object */
b3d636b0 1418 int id; /* index into kernel idr */
9ead190b 1419 struct kref ref;
38321256 1420 atomic_t usecnt; /* protects exclusive access */
d144da8c 1421 struct rcu_head rcu; /* kfree_rcu() overhead */
38321256
MB
1422
1423 const struct uverbs_obj_type *type;
e2773c06
RD
1424};
1425
cf8966b3
MB
1426struct ib_uobject_file {
1427 struct ib_uobject uobj;
1428 /* ufile contains the lock between context release and file close */
1429 struct ib_uverbs_file *ufile;
e2773c06
RD
1430};
1431
e2773c06 1432struct ib_udata {
309243ec 1433 const void __user *inbuf;
e2773c06
RD
1434 void __user *outbuf;
1435 size_t inlen;
1436 size_t outlen;
1437};
1438
1da177e4 1439struct ib_pd {
96249d70 1440 u32 local_dma_lkey;
ed082d36 1441 u32 flags;
e2773c06
RD
1442 struct ib_device *device;
1443 struct ib_uobject *uobject;
1444 atomic_t usecnt; /* count all resources */
50d46335 1445
ed082d36
CH
1446 u32 unsafe_global_rkey;
1447
50d46335
CH
1448 /*
1449 * Implementation details of the RDMA core, don't use in drivers:
1450 */
1451 struct ib_mr *__internal_mr;
1da177e4
LT
1452};
1453
59991f94
SH
1454struct ib_xrcd {
1455 struct ib_device *device;
d3d72d90 1456 atomic_t usecnt; /* count all exposed resources */
53d0bd1e 1457 struct inode *inode;
d3d72d90
SH
1458
1459 struct mutex tgt_qp_mutex;
1460 struct list_head tgt_qp_list;
59991f94
SH
1461};
1462
1da177e4
LT
1463struct ib_ah {
1464 struct ib_device *device;
1465 struct ib_pd *pd;
e2773c06 1466 struct ib_uobject *uobject;
1da177e4
LT
1467};
1468
1469typedef void (*ib_comp_handler)(struct ib_cq *cq, void *cq_context);
1470
14d3a3b2
CH
1471enum ib_poll_context {
1472 IB_POLL_DIRECT, /* caller context, no hw completions */
1473 IB_POLL_SOFTIRQ, /* poll from softirq context */
1474 IB_POLL_WORKQUEUE, /* poll from workqueue */
1475};
1476
1da177e4 1477struct ib_cq {
e2773c06
RD
1478 struct ib_device *device;
1479 struct ib_uobject *uobject;
1480 ib_comp_handler comp_handler;
1481 void (*event_handler)(struct ib_event *, void *);
4deccd6d 1482 void *cq_context;
e2773c06
RD
1483 int cqe;
1484 atomic_t usecnt; /* count number of work queues */
14d3a3b2
CH
1485 enum ib_poll_context poll_ctx;
1486 struct ib_wc *wc;
1487 union {
1488 struct irq_poll iop;
1489 struct work_struct work;
1490 };
1da177e4
LT
1491};
1492
1493struct ib_srq {
d41fcc67
RD
1494 struct ib_device *device;
1495 struct ib_pd *pd;
1496 struct ib_uobject *uobject;
1497 void (*event_handler)(struct ib_event *, void *);
1498 void *srq_context;
96104eda 1499 enum ib_srq_type srq_type;
1da177e4 1500 atomic_t usecnt;
418d5130
SH
1501
1502 union {
1503 struct {
1504 struct ib_xrcd *xrcd;
1505 struct ib_cq *cq;
1506 u32 srq_num;
1507 } xrc;
1508 } ext;
1da177e4
LT
1509};
1510
ebaaee25
NO
1511enum ib_raw_packet_caps {
1512 /* Strip cvlan from incoming packet and report it in the matching work
1513 * completion is supported.
1514 */
1515 IB_RAW_PACKET_CAP_CVLAN_STRIPPING = (1 << 0),
1516 /* Scatter FCS field of an incoming packet to host memory is supported.
1517 */
1518 IB_RAW_PACKET_CAP_SCATTER_FCS = (1 << 1),
1519 /* Checksum offloads are supported (for both send and receive). */
1520 IB_RAW_PACKET_CAP_IP_CSUM = (1 << 2),
1521};
1522
5fd251c8
YH
1523enum ib_wq_type {
1524 IB_WQT_RQ
1525};
1526
1527enum ib_wq_state {
1528 IB_WQS_RESET,
1529 IB_WQS_RDY,
1530 IB_WQS_ERR
1531};
1532
1533struct ib_wq {
1534 struct ib_device *device;
1535 struct ib_uobject *uobject;
1536 void *wq_context;
1537 void (*event_handler)(struct ib_event *, void *);
1538 struct ib_pd *pd;
1539 struct ib_cq *cq;
1540 u32 wq_num;
1541 enum ib_wq_state state;
1542 enum ib_wq_type wq_type;
1543 atomic_t usecnt;
1544};
1545
10bac72b
NO
1546enum ib_wq_flags {
1547 IB_WQ_FLAGS_CVLAN_STRIPPING = 1 << 0,
27b0df11 1548 IB_WQ_FLAGS_SCATTER_FCS = 1 << 1,
10bac72b
NO
1549};
1550
5fd251c8
YH
1551struct ib_wq_init_attr {
1552 void *wq_context;
1553 enum ib_wq_type wq_type;
1554 u32 max_wr;
1555 u32 max_sge;
1556 struct ib_cq *cq;
1557 void (*event_handler)(struct ib_event *, void *);
10bac72b 1558 u32 create_flags; /* Use enum ib_wq_flags */
5fd251c8
YH
1559};
1560
1561enum ib_wq_attr_mask {
10bac72b
NO
1562 IB_WQ_STATE = 1 << 0,
1563 IB_WQ_CUR_STATE = 1 << 1,
1564 IB_WQ_FLAGS = 1 << 2,
5fd251c8
YH
1565};
1566
1567struct ib_wq_attr {
1568 enum ib_wq_state wq_state;
1569 enum ib_wq_state curr_wq_state;
10bac72b
NO
1570 u32 flags; /* Use enum ib_wq_flags */
1571 u32 flags_mask; /* Use enum ib_wq_flags */
5fd251c8
YH
1572};
1573
6d39786b
YH
1574struct ib_rwq_ind_table {
1575 struct ib_device *device;
1576 struct ib_uobject *uobject;
1577 atomic_t usecnt;
1578 u32 ind_tbl_num;
1579 u32 log_ind_tbl_size;
1580 struct ib_wq **ind_tbl;
1581};
1582
1583struct ib_rwq_ind_table_init_attr {
1584 u32 log_ind_tbl_size;
1585 /* Each entry is a pointer to Receive Work Queue */
1586 struct ib_wq **ind_tbl;
1587};
1588
632bc3f6
BVA
1589/*
1590 * @max_write_sge: Maximum SGE elements per RDMA WRITE request.
1591 * @max_read_sge: Maximum SGE elements per RDMA READ request.
1592 */
1da177e4
LT
1593struct ib_qp {
1594 struct ib_device *device;
1595 struct ib_pd *pd;
1596 struct ib_cq *send_cq;
1597 struct ib_cq *recv_cq;
fffb0383
CH
1598 spinlock_t mr_lock;
1599 int mrs_used;
a060b562 1600 struct list_head rdma_mrs;
0e353e34 1601 struct list_head sig_mrs;
1da177e4 1602 struct ib_srq *srq;
b42b63cf 1603 struct ib_xrcd *xrcd; /* XRC TGT QPs only */
d3d72d90 1604 struct list_head xrcd_list;
fffb0383 1605
319a441d
HHZ
1606 /* count times opened, mcast attaches, flow attaches */
1607 atomic_t usecnt;
0e0ec7e0
SH
1608 struct list_head open_list;
1609 struct ib_qp *real_qp;
e2773c06 1610 struct ib_uobject *uobject;
1da177e4
LT
1611 void (*event_handler)(struct ib_event *, void *);
1612 void *qp_context;
1613 u32 qp_num;
632bc3f6
BVA
1614 u32 max_write_sge;
1615 u32 max_read_sge;
1da177e4 1616 enum ib_qp_type qp_type;
a9017e23 1617 struct ib_rwq_ind_table *rwq_ind_tbl;
1da177e4
LT
1618};
1619
1620struct ib_mr {
e2773c06
RD
1621 struct ib_device *device;
1622 struct ib_pd *pd;
e2773c06
RD
1623 u32 lkey;
1624 u32 rkey;
4c67e2bf
SG
1625 u64 iova;
1626 u32 length;
1627 unsigned int page_size;
d4a85c30 1628 bool need_inval;
fffb0383
CH
1629 union {
1630 struct ib_uobject *uobject; /* user */
1631 struct list_head qp_entry; /* FR */
1632 };
1da177e4
LT
1633};
1634
1635struct ib_mw {
1636 struct ib_device *device;
1637 struct ib_pd *pd;
e2773c06 1638 struct ib_uobject *uobject;
1da177e4 1639 u32 rkey;
7083e42e 1640 enum ib_mw_type type;
1da177e4
LT
1641};
1642
1643struct ib_fmr {
1644 struct ib_device *device;
1645 struct ib_pd *pd;
1646 struct list_head list;
1647 u32 lkey;
1648 u32 rkey;
1649};
1650
319a441d
HHZ
1651/* Supported steering options */
1652enum ib_flow_attr_type {
1653 /* steering according to rule specifications */
1654 IB_FLOW_ATTR_NORMAL = 0x0,
1655 /* default unicast and multicast rule -
1656 * receive all Eth traffic which isn't steered to any QP
1657 */
1658 IB_FLOW_ATTR_ALL_DEFAULT = 0x1,
1659 /* default multicast rule -
1660 * receive all Eth multicast traffic which isn't steered to any QP
1661 */
1662 IB_FLOW_ATTR_MC_DEFAULT = 0x2,
1663 /* sniffer rule - receive all port traffic */
1664 IB_FLOW_ATTR_SNIFFER = 0x3
1665};
1666
1667/* Supported steering header types */
1668enum ib_flow_spec_type {
1669 /* L2 headers*/
76bd23b3
MR
1670 IB_FLOW_SPEC_ETH = 0x20,
1671 IB_FLOW_SPEC_IB = 0x22,
319a441d 1672 /* L3 header*/
76bd23b3
MR
1673 IB_FLOW_SPEC_IPV4 = 0x30,
1674 IB_FLOW_SPEC_IPV6 = 0x31,
319a441d 1675 /* L4 headers*/
76bd23b3
MR
1676 IB_FLOW_SPEC_TCP = 0x40,
1677 IB_FLOW_SPEC_UDP = 0x41,
0dbf3332 1678 IB_FLOW_SPEC_VXLAN_TUNNEL = 0x50,
fbf46860 1679 IB_FLOW_SPEC_INNER = 0x100,
460d0198
MR
1680 /* Actions */
1681 IB_FLOW_SPEC_ACTION_TAG = 0x1000,
319a441d 1682};
240ae00e 1683#define IB_FLOW_SPEC_LAYER_MASK 0xF0
fbf46860 1684#define IB_FLOW_SPEC_SUPPORT_LAYERS 8
22878dbc 1685
319a441d
HHZ
1686/* Flow steering rule priority is set according to it's domain.
1687 * Lower domain value means higher priority.
1688 */
1689enum ib_flow_domain {
1690 IB_FLOW_DOMAIN_USER,
1691 IB_FLOW_DOMAIN_ETHTOOL,
1692 IB_FLOW_DOMAIN_RFS,
1693 IB_FLOW_DOMAIN_NIC,
1694 IB_FLOW_DOMAIN_NUM /* Must be last */
1695};
1696
a3100a78
MV
1697enum ib_flow_flags {
1698 IB_FLOW_ATTR_FLAGS_DONT_TRAP = 1UL << 1, /* Continue match, no steal */
1699 IB_FLOW_ATTR_FLAGS_RESERVED = 1UL << 2 /* Must be last */
1700};
1701
319a441d
HHZ
1702struct ib_flow_eth_filter {
1703 u8 dst_mac[6];
1704 u8 src_mac[6];
1705 __be16 ether_type;
1706 __be16 vlan_tag;
15dfbd6b
MG
1707 /* Must be last */
1708 u8 real_sz[0];
319a441d
HHZ
1709};
1710
1711struct ib_flow_spec_eth {
fbf46860 1712 u32 type;
319a441d
HHZ
1713 u16 size;
1714 struct ib_flow_eth_filter val;
1715 struct ib_flow_eth_filter mask;
1716};
1717
240ae00e
MB
1718struct ib_flow_ib_filter {
1719 __be16 dlid;
1720 __u8 sl;
15dfbd6b
MG
1721 /* Must be last */
1722 u8 real_sz[0];
240ae00e
MB
1723};
1724
1725struct ib_flow_spec_ib {
fbf46860 1726 u32 type;
240ae00e
MB
1727 u16 size;
1728 struct ib_flow_ib_filter val;
1729 struct ib_flow_ib_filter mask;
1730};
1731
989a3a8f
MG
1732/* IPv4 header flags */
1733enum ib_ipv4_flags {
1734 IB_IPV4_DONT_FRAG = 0x2, /* Don't enable packet fragmentation */
1735 IB_IPV4_MORE_FRAG = 0X4 /* For All fragmented packets except the
1736 last have this flag set */
1737};
1738
319a441d
HHZ
1739struct ib_flow_ipv4_filter {
1740 __be32 src_ip;
1741 __be32 dst_ip;
989a3a8f
MG
1742 u8 proto;
1743 u8 tos;
1744 u8 ttl;
1745 u8 flags;
15dfbd6b
MG
1746 /* Must be last */
1747 u8 real_sz[0];
319a441d
HHZ
1748};
1749
1750struct ib_flow_spec_ipv4 {
fbf46860 1751 u32 type;
319a441d
HHZ
1752 u16 size;
1753 struct ib_flow_ipv4_filter val;
1754 struct ib_flow_ipv4_filter mask;
1755};
1756
4c2aae71
MG
1757struct ib_flow_ipv6_filter {
1758 u8 src_ip[16];
1759 u8 dst_ip[16];
a72c6a2b
MG
1760 __be32 flow_label;
1761 u8 next_hdr;
1762 u8 traffic_class;
1763 u8 hop_limit;
15dfbd6b
MG
1764 /* Must be last */
1765 u8 real_sz[0];
4c2aae71
MG
1766};
1767
1768struct ib_flow_spec_ipv6 {
fbf46860 1769 u32 type;
4c2aae71
MG
1770 u16 size;
1771 struct ib_flow_ipv6_filter val;
1772 struct ib_flow_ipv6_filter mask;
1773};
1774
319a441d
HHZ
1775struct ib_flow_tcp_udp_filter {
1776 __be16 dst_port;
1777 __be16 src_port;
15dfbd6b
MG
1778 /* Must be last */
1779 u8 real_sz[0];
319a441d
HHZ
1780};
1781
1782struct ib_flow_spec_tcp_udp {
fbf46860 1783 u32 type;
319a441d
HHZ
1784 u16 size;
1785 struct ib_flow_tcp_udp_filter val;
1786 struct ib_flow_tcp_udp_filter mask;
1787};
1788
0dbf3332
MR
1789struct ib_flow_tunnel_filter {
1790 __be32 tunnel_id;
1791 u8 real_sz[0];
1792};
1793
1794/* ib_flow_spec_tunnel describes the Vxlan tunnel
1795 * the tunnel_id from val has the vni value
1796 */
1797struct ib_flow_spec_tunnel {
fbf46860 1798 u32 type;
0dbf3332
MR
1799 u16 size;
1800 struct ib_flow_tunnel_filter val;
1801 struct ib_flow_tunnel_filter mask;
1802};
1803
460d0198
MR
1804struct ib_flow_spec_action_tag {
1805 enum ib_flow_spec_type type;
1806 u16 size;
1807 u32 tag_id;
1808};
1809
319a441d
HHZ
1810union ib_flow_spec {
1811 struct {
fbf46860 1812 u32 type;
319a441d
HHZ
1813 u16 size;
1814 };
1815 struct ib_flow_spec_eth eth;
240ae00e 1816 struct ib_flow_spec_ib ib;
319a441d
HHZ
1817 struct ib_flow_spec_ipv4 ipv4;
1818 struct ib_flow_spec_tcp_udp tcp_udp;
4c2aae71 1819 struct ib_flow_spec_ipv6 ipv6;
0dbf3332 1820 struct ib_flow_spec_tunnel tunnel;
460d0198 1821 struct ib_flow_spec_action_tag flow_tag;
319a441d
HHZ
1822};
1823
1824struct ib_flow_attr {
1825 enum ib_flow_attr_type type;
1826 u16 size;
1827 u16 priority;
1828 u32 flags;
1829 u8 num_of_specs;
1830 u8 port;
1831 /* Following are the optional layers according to user request
1832 * struct ib_flow_spec_xxx
1833 * struct ib_flow_spec_yyy
1834 */
1835};
1836
1837struct ib_flow {
1838 struct ib_qp *qp;
1839 struct ib_uobject *uobject;
1840};
1841
4cd7c947 1842struct ib_mad_hdr;
1da177e4
LT
1843struct ib_grh;
1844
1845enum ib_process_mad_flags {
1846 IB_MAD_IGNORE_MKEY = 1,
1847 IB_MAD_IGNORE_BKEY = 2,
1848 IB_MAD_IGNORE_ALL = IB_MAD_IGNORE_MKEY | IB_MAD_IGNORE_BKEY
1849};
1850
1851enum ib_mad_result {
1852 IB_MAD_RESULT_FAILURE = 0, /* (!SUCCESS is the important flag) */
1853 IB_MAD_RESULT_SUCCESS = 1 << 0, /* MAD was successfully processed */
1854 IB_MAD_RESULT_REPLY = 1 << 1, /* Reply packet needs to be sent */
1855 IB_MAD_RESULT_CONSUMED = 1 << 2 /* Packet consumed: stop processing */
1856};
1857
1858#define IB_DEVICE_NAME_MAX 64
1859
21d6454a
JW
1860struct ib_port_cache {
1861 struct ib_pkey_cache *pkey;
1862 struct ib_gid_table *gid;
1863 u8 lmc;
1864 enum ib_port_state port_state;
1865};
1866
1da177e4
LT
1867struct ib_cache {
1868 rwlock_t lock;
1869 struct ib_event_handler event_handler;
21d6454a 1870 struct ib_port_cache *ports;
1da177e4
LT
1871};
1872
07ebafba
TT
1873struct iw_cm_verbs;
1874
7738613e
IW
1875struct ib_port_immutable {
1876 int pkey_tbl_len;
1877 int gid_tbl_len;
f9b22e35 1878 u32 core_cap_flags;
337877a4 1879 u32 max_mad_size;
7738613e
IW
1880};
1881
2fc77572
VN
1882/* rdma netdev type - specifies protocol type */
1883enum rdma_netdev_t {
1884 RDMA_NETDEV_OPA_VNIC
1885};
1886
1887/**
1888 * struct rdma_netdev - rdma netdev
1889 * For cases where netstack interfacing is required.
1890 */
1891struct rdma_netdev {
1892 void *clnt_priv;
1893 struct ib_device *hca;
1894 u8 port_num;
1895
1896 /* control functions */
1897 void (*set_id)(struct net_device *netdev, int id);
1898};
1899
1da177e4 1900struct ib_device {
0957c29f
BVA
1901 /* Do not access @dma_device directly from ULP nor from HW drivers. */
1902 struct device *dma_device;
1903
1da177e4
LT
1904 char name[IB_DEVICE_NAME_MAX];
1905
1906 struct list_head event_handler_list;
1907 spinlock_t event_handler_lock;
1908
17a55f79 1909 spinlock_t client_data_lock;
1da177e4 1910 struct list_head core_list;
7c1eb45a
HE
1911 /* Access to the client_data_list is protected by the client_data_lock
1912 * spinlock and the lists_rwsem read-write semaphore */
1da177e4 1913 struct list_head client_data_list;
1da177e4
LT
1914
1915 struct ib_cache cache;
7738613e
IW
1916 /**
1917 * port_immutable is indexed by port number
1918 */
1919 struct ib_port_immutable *port_immutable;
1da177e4 1920
f4fd0b22
MT
1921 int num_comp_vectors;
1922
07ebafba
TT
1923 struct iw_cm_verbs *iwcm;
1924
b40f4757
CL
1925 /**
1926 * alloc_hw_stats - Allocate a struct rdma_hw_stats and fill in the
1927 * driver initialized data. The struct is kfree()'ed by the sysfs
1928 * core when the device is removed. A lifespan of -1 in the return
1929 * struct tells the core to set a default lifespan.
1930 */
1931 struct rdma_hw_stats *(*alloc_hw_stats)(struct ib_device *device,
1932 u8 port_num);
1933 /**
1934 * get_hw_stats - Fill in the counter value(s) in the stats struct.
1935 * @index - The index in the value array we wish to have updated, or
1936 * num_counters if we want all stats updated
1937 * Return codes -
1938 * < 0 - Error, no counters updated
1939 * index - Updated the single counter pointed to by index
1940 * num_counters - Updated all counters (will reset the timestamp
1941 * and prevent further calls for lifespan milliseconds)
1942 * Drivers are allowed to update all counters in leiu of just the
1943 * one given in index at their option
1944 */
1945 int (*get_hw_stats)(struct ib_device *device,
1946 struct rdma_hw_stats *stats,
1947 u8 port, int index);
1da177e4 1948 int (*query_device)(struct ib_device *device,
2528e33e
MB
1949 struct ib_device_attr *device_attr,
1950 struct ib_udata *udata);
1da177e4
LT
1951 int (*query_port)(struct ib_device *device,
1952 u8 port_num,
1953 struct ib_port_attr *port_attr);
a3f5adaf
EC
1954 enum rdma_link_layer (*get_link_layer)(struct ib_device *device,
1955 u8 port_num);
03db3a2d
MB
1956 /* When calling get_netdev, the HW vendor's driver should return the
1957 * net device of device @device at port @port_num or NULL if such
1958 * a net device doesn't exist. The vendor driver should call dev_hold
1959 * on this net device. The HW vendor's device driver must guarantee
1960 * that this function returns NULL before the net device reaches
1961 * NETDEV_UNREGISTER_FINAL state.
1962 */
1963 struct net_device *(*get_netdev)(struct ib_device *device,
1964 u8 port_num);
1da177e4
LT
1965 int (*query_gid)(struct ib_device *device,
1966 u8 port_num, int index,
1967 union ib_gid *gid);
03db3a2d
MB
1968 /* When calling add_gid, the HW vendor's driver should
1969 * add the gid of device @device at gid index @index of
1970 * port @port_num to be @gid. Meta-info of that gid (for example,
1971 * the network device related to this gid is available
1972 * at @attr. @context allows the HW vendor driver to store extra
1973 * information together with a GID entry. The HW vendor may allocate
1974 * memory to contain this information and store it in @context when a
1975 * new GID entry is written to. Params are consistent until the next
1976 * call of add_gid or delete_gid. The function should return 0 on
1977 * success or error otherwise. The function could be called
1978 * concurrently for different ports. This function is only called
1979 * when roce_gid_table is used.
1980 */
1981 int (*add_gid)(struct ib_device *device,
1982 u8 port_num,
1983 unsigned int index,
1984 const union ib_gid *gid,
1985 const struct ib_gid_attr *attr,
1986 void **context);
1987 /* When calling del_gid, the HW vendor's driver should delete the
1988 * gid of device @device at gid index @index of port @port_num.
1989 * Upon the deletion of a GID entry, the HW vendor must free any
1990 * allocated memory. The caller will clear @context afterwards.
1991 * This function is only called when roce_gid_table is used.
1992 */
1993 int (*del_gid)(struct ib_device *device,
1994 u8 port_num,
1995 unsigned int index,
1996 void **context);
1da177e4
LT
1997 int (*query_pkey)(struct ib_device *device,
1998 u8 port_num, u16 index, u16 *pkey);
1999 int (*modify_device)(struct ib_device *device,
2000 int device_modify_mask,
2001 struct ib_device_modify *device_modify);
2002 int (*modify_port)(struct ib_device *device,
2003 u8 port_num, int port_modify_mask,
2004 struct ib_port_modify *port_modify);
e2773c06
RD
2005 struct ib_ucontext * (*alloc_ucontext)(struct ib_device *device,
2006 struct ib_udata *udata);
2007 int (*dealloc_ucontext)(struct ib_ucontext *context);
2008 int (*mmap)(struct ib_ucontext *context,
2009 struct vm_area_struct *vma);
2010 struct ib_pd * (*alloc_pd)(struct ib_device *device,
2011 struct ib_ucontext *context,
2012 struct ib_udata *udata);
1da177e4
LT
2013 int (*dealloc_pd)(struct ib_pd *pd);
2014 struct ib_ah * (*create_ah)(struct ib_pd *pd,
477864c8
MS
2015 struct ib_ah_attr *ah_attr,
2016 struct ib_udata *udata);
1da177e4
LT
2017 int (*modify_ah)(struct ib_ah *ah,
2018 struct ib_ah_attr *ah_attr);
2019 int (*query_ah)(struct ib_ah *ah,
2020 struct ib_ah_attr *ah_attr);
2021 int (*destroy_ah)(struct ib_ah *ah);
d41fcc67
RD
2022 struct ib_srq * (*create_srq)(struct ib_pd *pd,
2023 struct ib_srq_init_attr *srq_init_attr,
2024 struct ib_udata *udata);
2025 int (*modify_srq)(struct ib_srq *srq,
2026 struct ib_srq_attr *srq_attr,
9bc57e2d
RC
2027 enum ib_srq_attr_mask srq_attr_mask,
2028 struct ib_udata *udata);
d41fcc67
RD
2029 int (*query_srq)(struct ib_srq *srq,
2030 struct ib_srq_attr *srq_attr);
2031 int (*destroy_srq)(struct ib_srq *srq);
2032 int (*post_srq_recv)(struct ib_srq *srq,
2033 struct ib_recv_wr *recv_wr,
2034 struct ib_recv_wr **bad_recv_wr);
1da177e4 2035 struct ib_qp * (*create_qp)(struct ib_pd *pd,
e2773c06
RD
2036 struct ib_qp_init_attr *qp_init_attr,
2037 struct ib_udata *udata);
1da177e4
LT
2038 int (*modify_qp)(struct ib_qp *qp,
2039 struct ib_qp_attr *qp_attr,
9bc57e2d
RC
2040 int qp_attr_mask,
2041 struct ib_udata *udata);
1da177e4
LT
2042 int (*query_qp)(struct ib_qp *qp,
2043 struct ib_qp_attr *qp_attr,
2044 int qp_attr_mask,
2045 struct ib_qp_init_attr *qp_init_attr);
2046 int (*destroy_qp)(struct ib_qp *qp);
2047 int (*post_send)(struct ib_qp *qp,
2048 struct ib_send_wr *send_wr,
2049 struct ib_send_wr **bad_send_wr);
2050 int (*post_recv)(struct ib_qp *qp,
2051 struct ib_recv_wr *recv_wr,
2052 struct ib_recv_wr **bad_recv_wr);
bcf4c1ea
MB
2053 struct ib_cq * (*create_cq)(struct ib_device *device,
2054 const struct ib_cq_init_attr *attr,
e2773c06
RD
2055 struct ib_ucontext *context,
2056 struct ib_udata *udata);
2dd57162
EC
2057 int (*modify_cq)(struct ib_cq *cq, u16 cq_count,
2058 u16 cq_period);
1da177e4 2059 int (*destroy_cq)(struct ib_cq *cq);
33b9b3ee
RD
2060 int (*resize_cq)(struct ib_cq *cq, int cqe,
2061 struct ib_udata *udata);
1da177e4
LT
2062 int (*poll_cq)(struct ib_cq *cq, int num_entries,
2063 struct ib_wc *wc);
2064 int (*peek_cq)(struct ib_cq *cq, int wc_cnt);
2065 int (*req_notify_cq)(struct ib_cq *cq,
ed23a727 2066 enum ib_cq_notify_flags flags);
1da177e4
LT
2067 int (*req_ncomp_notif)(struct ib_cq *cq,
2068 int wc_cnt);
2069 struct ib_mr * (*get_dma_mr)(struct ib_pd *pd,
2070 int mr_access_flags);
e2773c06 2071 struct ib_mr * (*reg_user_mr)(struct ib_pd *pd,
f7c6a7b5
RD
2072 u64 start, u64 length,
2073 u64 virt_addr,
e2773c06
RD
2074 int mr_access_flags,
2075 struct ib_udata *udata);
7e6edb9b
MB
2076 int (*rereg_user_mr)(struct ib_mr *mr,
2077 int flags,
2078 u64 start, u64 length,
2079 u64 virt_addr,
2080 int mr_access_flags,
2081 struct ib_pd *pd,
2082 struct ib_udata *udata);
1da177e4 2083 int (*dereg_mr)(struct ib_mr *mr);
9bee178b
SG
2084 struct ib_mr * (*alloc_mr)(struct ib_pd *pd,
2085 enum ib_mr_type mr_type,
2086 u32 max_num_sg);
4c67e2bf
SG
2087 int (*map_mr_sg)(struct ib_mr *mr,
2088 struct scatterlist *sg,
ff2ba993 2089 int sg_nents,
9aa8b321 2090 unsigned int *sg_offset);
7083e42e 2091 struct ib_mw * (*alloc_mw)(struct ib_pd *pd,
b2a239df
MB
2092 enum ib_mw_type type,
2093 struct ib_udata *udata);
1da177e4
LT
2094 int (*dealloc_mw)(struct ib_mw *mw);
2095 struct ib_fmr * (*alloc_fmr)(struct ib_pd *pd,
2096 int mr_access_flags,
2097 struct ib_fmr_attr *fmr_attr);
2098 int (*map_phys_fmr)(struct ib_fmr *fmr,
2099 u64 *page_list, int list_len,
2100 u64 iova);
2101 int (*unmap_fmr)(struct list_head *fmr_list);
2102 int (*dealloc_fmr)(struct ib_fmr *fmr);
2103 int (*attach_mcast)(struct ib_qp *qp,
2104 union ib_gid *gid,
2105 u16 lid);
2106 int (*detach_mcast)(struct ib_qp *qp,
2107 union ib_gid *gid,
2108 u16 lid);
2109 int (*process_mad)(struct ib_device *device,
2110 int process_mad_flags,
2111 u8 port_num,
a97e2d86
IW
2112 const struct ib_wc *in_wc,
2113 const struct ib_grh *in_grh,
4cd7c947
IW
2114 const struct ib_mad_hdr *in_mad,
2115 size_t in_mad_size,
2116 struct ib_mad_hdr *out_mad,
2117 size_t *out_mad_size,
2118 u16 *out_mad_pkey_index);
59991f94
SH
2119 struct ib_xrcd * (*alloc_xrcd)(struct ib_device *device,
2120 struct ib_ucontext *ucontext,
2121 struct ib_udata *udata);
2122 int (*dealloc_xrcd)(struct ib_xrcd *xrcd);
319a441d
HHZ
2123 struct ib_flow * (*create_flow)(struct ib_qp *qp,
2124 struct ib_flow_attr
2125 *flow_attr,
2126 int domain);
2127 int (*destroy_flow)(struct ib_flow *flow_id);
1b01d335
SG
2128 int (*check_mr_status)(struct ib_mr *mr, u32 check_mask,
2129 struct ib_mr_status *mr_status);
036b1063 2130 void (*disassociate_ucontext)(struct ib_ucontext *ibcontext);
765d6774
SW
2131 void (*drain_rq)(struct ib_qp *qp);
2132 void (*drain_sq)(struct ib_qp *qp);
50174a7f
EC
2133 int (*set_vf_link_state)(struct ib_device *device, int vf, u8 port,
2134 int state);
2135 int (*get_vf_config)(struct ib_device *device, int vf, u8 port,
2136 struct ifla_vf_info *ivf);
2137 int (*get_vf_stats)(struct ib_device *device, int vf, u8 port,
2138 struct ifla_vf_stats *stats);
2139 int (*set_vf_guid)(struct ib_device *device, int vf, u8 port, u64 guid,
2140 int type);
5fd251c8
YH
2141 struct ib_wq * (*create_wq)(struct ib_pd *pd,
2142 struct ib_wq_init_attr *init_attr,
2143 struct ib_udata *udata);
2144 int (*destroy_wq)(struct ib_wq *wq);
2145 int (*modify_wq)(struct ib_wq *wq,
2146 struct ib_wq_attr *attr,
2147 u32 wq_attr_mask,
2148 struct ib_udata *udata);
6d39786b
YH
2149 struct ib_rwq_ind_table * (*create_rwq_ind_table)(struct ib_device *device,
2150 struct ib_rwq_ind_table_init_attr *init_attr,
2151 struct ib_udata *udata);
2152 int (*destroy_rwq_ind_table)(struct ib_rwq_ind_table *wq_ind_table);
2fc77572
VN
2153 /**
2154 * rdma netdev operations
2155 *
2156 * Driver implementing alloc_rdma_netdev must return -EOPNOTSUPP if it
2157 * doesn't support the specified rdma netdev type.
2158 */
2159 struct net_device *(*alloc_rdma_netdev)(
2160 struct ib_device *device,
2161 u8 port_num,
2162 enum rdma_netdev_t type,
2163 const char *name,
2164 unsigned char name_assign_type,
2165 void (*setup)(struct net_device *));
2166 void (*free_rdma_netdev)(struct net_device *netdev);
9b513090 2167
e2773c06 2168 struct module *owner;
f4e91eb4 2169 struct device dev;
35be0681 2170 struct kobject *ports_parent;
1da177e4
LT
2171 struct list_head port_list;
2172
2173 enum {
2174 IB_DEV_UNINITIALIZED,
2175 IB_DEV_REGISTERED,
2176 IB_DEV_UNREGISTERED
2177 } reg_state;
2178
274c0891 2179 int uverbs_abi_ver;
17a55f79 2180 u64 uverbs_cmd_mask;
f21519b2 2181 u64 uverbs_ex_cmd_mask;
274c0891 2182
bd99fdea 2183 char node_desc[IB_DEVICE_NODE_DESC_MAX];
cf311cd4 2184 __be64 node_guid;
96f15c03 2185 u32 local_dma_lkey;
4139032b 2186 u16 is_switch:1;
1da177e4
LT
2187 u8 node_type;
2188 u8 phys_port_cnt;
3e153a93 2189 struct ib_device_attr attrs;
b40f4757
CL
2190 struct attribute_group *hw_stats_ag;
2191 struct rdma_hw_stats *hw_stats;
7738613e 2192
43579b5f
PP
2193#ifdef CONFIG_CGROUP_RDMA
2194 struct rdmacg_device cg_device;
2195#endif
2196
7738613e
IW
2197 /**
2198 * The following mandatory functions are used only at device
2199 * registration. Keep functions such as these at the end of this
2200 * structure to avoid cache line misses when accessing struct ib_device
2201 * in fast paths.
2202 */
2203 int (*get_port_immutable)(struct ib_device *, u8, struct ib_port_immutable *);
5fa76c20 2204 void (*get_dev_fw_str)(struct ib_device *, char *str, size_t str_len);
1da177e4
LT
2205};
2206
2207struct ib_client {
2208 char *name;
2209 void (*add) (struct ib_device *);
7c1eb45a 2210 void (*remove)(struct ib_device *, void *client_data);
1da177e4 2211
9268f72d
YK
2212 /* Returns the net_dev belonging to this ib_client and matching the
2213 * given parameters.
2214 * @dev: An RDMA device that the net_dev use for communication.
2215 * @port: A physical port number on the RDMA device.
2216 * @pkey: P_Key that the net_dev uses if applicable.
2217 * @gid: A GID that the net_dev uses to communicate.
2218 * @addr: An IP address the net_dev is configured with.
2219 * @client_data: The device's client data set by ib_set_client_data().
2220 *
2221 * An ib_client that implements a net_dev on top of RDMA devices
2222 * (such as IP over IB) should implement this callback, allowing the
2223 * rdma_cm module to find the right net_dev for a given request.
2224 *
2225 * The caller is responsible for calling dev_put on the returned
2226 * netdev. */
2227 struct net_device *(*get_net_dev_by_params)(
2228 struct ib_device *dev,
2229 u8 port,
2230 u16 pkey,
2231 const union ib_gid *gid,
2232 const struct sockaddr *addr,
2233 void *client_data);
1da177e4
LT
2234 struct list_head list;
2235};
2236
2237struct ib_device *ib_alloc_device(size_t size);
2238void ib_dealloc_device(struct ib_device *device);
2239
5fa76c20
IW
2240void ib_get_device_fw_str(struct ib_device *device, char *str, size_t str_len);
2241
9a6edb60
RC
2242int ib_register_device(struct ib_device *device,
2243 int (*port_callback)(struct ib_device *,
2244 u8, struct kobject *));
1da177e4
LT
2245void ib_unregister_device(struct ib_device *device);
2246
2247int ib_register_client (struct ib_client *client);
2248void ib_unregister_client(struct ib_client *client);
2249
2250void *ib_get_client_data(struct ib_device *device, struct ib_client *client);
2251void ib_set_client_data(struct ib_device *device, struct ib_client *client,
2252 void *data);
2253
e2773c06
RD
2254static inline int ib_copy_from_udata(void *dest, struct ib_udata *udata, size_t len)
2255{
2256 return copy_from_user(dest, udata->inbuf, len) ? -EFAULT : 0;
2257}
2258
2259static inline int ib_copy_to_udata(struct ib_udata *udata, void *src, size_t len)
2260{
43c61165 2261 return copy_to_user(udata->outbuf, src, len) ? -EFAULT : 0;
e2773c06
RD
2262}
2263
301a721e
MB
2264static inline bool ib_is_udata_cleared(struct ib_udata *udata,
2265 size_t offset,
2266 size_t len)
2267{
2268 const void __user *p = udata->inbuf + offset;
92d27ae6 2269 bool ret;
301a721e
MB
2270 u8 *buf;
2271
2272 if (len > USHRT_MAX)
2273 return false;
2274
92d27ae6
ME
2275 buf = memdup_user(p, len);
2276 if (IS_ERR(buf))
301a721e
MB
2277 return false;
2278
301a721e 2279 ret = !memchr_inv(buf, 0, len);
301a721e
MB
2280 kfree(buf);
2281 return ret;
2282}
2283
8a51866f
RD
2284/**
2285 * ib_modify_qp_is_ok - Check that the supplied attribute mask
2286 * contains all required attributes and no attributes not allowed for
2287 * the given QP state transition.
2288 * @cur_state: Current QP state
2289 * @next_state: Next QP state
2290 * @type: QP type
2291 * @mask: Mask of supplied QP attributes
dd5f03be 2292 * @ll : link layer of port
8a51866f
RD
2293 *
2294 * This function is a helper function that a low-level driver's
2295 * modify_qp method can use to validate the consumer's input. It
2296 * checks that cur_state and next_state are valid QP states, that a
2297 * transition from cur_state to next_state is allowed by the IB spec,
2298 * and that the attribute mask supplied is allowed for the transition.
2299 */
2300int ib_modify_qp_is_ok(enum ib_qp_state cur_state, enum ib_qp_state next_state,
dd5f03be
MB
2301 enum ib_qp_type type, enum ib_qp_attr_mask mask,
2302 enum rdma_link_layer ll);
8a51866f 2303
1da177e4
LT
2304int ib_register_event_handler (struct ib_event_handler *event_handler);
2305int ib_unregister_event_handler(struct ib_event_handler *event_handler);
2306void ib_dispatch_event(struct ib_event *event);
2307
1da177e4
LT
2308int ib_query_port(struct ib_device *device,
2309 u8 port_num, struct ib_port_attr *port_attr);
2310
a3f5adaf
EC
2311enum rdma_link_layer rdma_port_get_link_layer(struct ib_device *device,
2312 u8 port_num);
2313
4139032b
HR
2314/**
2315 * rdma_cap_ib_switch - Check if the device is IB switch
2316 * @device: Device to check
2317 *
2318 * Device driver is responsible for setting is_switch bit on
2319 * in ib_device structure at init time.
2320 *
2321 * Return: true if the device is IB switch.
2322 */
2323static inline bool rdma_cap_ib_switch(const struct ib_device *device)
2324{
2325 return device->is_switch;
2326}
2327
0cf18d77
IW
2328/**
2329 * rdma_start_port - Return the first valid port number for the device
2330 * specified
2331 *
2332 * @device: Device to be checked
2333 *
2334 * Return start port number
2335 */
2336static inline u8 rdma_start_port(const struct ib_device *device)
2337{
4139032b 2338 return rdma_cap_ib_switch(device) ? 0 : 1;
0cf18d77
IW
2339}
2340
2341/**
2342 * rdma_end_port - Return the last valid port number for the device
2343 * specified
2344 *
2345 * @device: Device to be checked
2346 *
2347 * Return last port number
2348 */
2349static inline u8 rdma_end_port(const struct ib_device *device)
2350{
4139032b 2351 return rdma_cap_ib_switch(device) ? 0 : device->phys_port_cnt;
0cf18d77
IW
2352}
2353
24dc831b
YS
2354static inline int rdma_is_port_valid(const struct ib_device *device,
2355 unsigned int port)
2356{
2357 return (port >= rdma_start_port(device) &&
2358 port <= rdma_end_port(device));
2359}
2360
5ede9289 2361static inline bool rdma_protocol_ib(const struct ib_device *device, u8 port_num)
de66be94 2362{
f9b22e35 2363 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_IB;
de66be94
MW
2364}
2365
5ede9289 2366static inline bool rdma_protocol_roce(const struct ib_device *device, u8 port_num)
7766a99f
MB
2367{
2368 return device->port_immutable[port_num].core_cap_flags &
2369 (RDMA_CORE_CAP_PROT_ROCE | RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP);
2370}
2371
2372static inline bool rdma_protocol_roce_udp_encap(const struct ib_device *device, u8 port_num)
2373{
2374 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP;
2375}
2376
2377static inline bool rdma_protocol_roce_eth_encap(const struct ib_device *device, u8 port_num)
de66be94 2378{
f9b22e35 2379 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_ROCE;
de66be94
MW
2380}
2381
5ede9289 2382static inline bool rdma_protocol_iwarp(const struct ib_device *device, u8 port_num)
de66be94 2383{
f9b22e35 2384 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_IWARP;
de66be94
MW
2385}
2386
5ede9289 2387static inline bool rdma_ib_or_roce(const struct ib_device *device, u8 port_num)
de66be94 2388{
7766a99f
MB
2389 return rdma_protocol_ib(device, port_num) ||
2390 rdma_protocol_roce(device, port_num);
de66be94
MW
2391}
2392
aa773bd4
OG
2393static inline bool rdma_protocol_raw_packet(const struct ib_device *device, u8 port_num)
2394{
2395 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_RAW_PACKET;
2396}
2397
ce1e055f
OG
2398static inline bool rdma_protocol_usnic(const struct ib_device *device, u8 port_num)
2399{
2400 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_USNIC;
2401}
2402
c757dea8 2403/**
296ec009 2404 * rdma_cap_ib_mad - Check if the port of a device supports Infiniband
c757dea8 2405 * Management Datagrams.
296ec009
MW
2406 * @device: Device to check
2407 * @port_num: Port number to check
c757dea8 2408 *
296ec009
MW
2409 * Management Datagrams (MAD) are a required part of the InfiniBand
2410 * specification and are supported on all InfiniBand devices. A slightly
2411 * extended version are also supported on OPA interfaces.
c757dea8 2412 *
296ec009 2413 * Return: true if the port supports sending/receiving of MAD packets.
c757dea8 2414 */
5ede9289 2415static inline bool rdma_cap_ib_mad(const struct ib_device *device, u8 port_num)
c757dea8 2416{
f9b22e35 2417 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IB_MAD;
c757dea8
MW
2418}
2419
65995fee
IW
2420/**
2421 * rdma_cap_opa_mad - Check if the port of device provides support for OPA
2422 * Management Datagrams.
2423 * @device: Device to check
2424 * @port_num: Port number to check
2425 *
2426 * Intel OmniPath devices extend and/or replace the InfiniBand Management
2427 * datagrams with their own versions. These OPA MADs share many but not all of
2428 * the characteristics of InfiniBand MADs.
2429 *
2430 * OPA MADs differ in the following ways:
2431 *
2432 * 1) MADs are variable size up to 2K
2433 * IBTA defined MADs remain fixed at 256 bytes
2434 * 2) OPA SMPs must carry valid PKeys
2435 * 3) OPA SMP packets are a different format
2436 *
2437 * Return: true if the port supports OPA MAD packet formats.
2438 */
2439static inline bool rdma_cap_opa_mad(struct ib_device *device, u8 port_num)
2440{
2441 return (device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_OPA_MAD)
2442 == RDMA_CORE_CAP_OPA_MAD;
2443}
2444
29541e3a 2445/**
296ec009
MW
2446 * rdma_cap_ib_smi - Check if the port of a device provides an Infiniband
2447 * Subnet Management Agent (SMA) on the Subnet Management Interface (SMI).
2448 * @device: Device to check
2449 * @port_num: Port number to check
29541e3a 2450 *
296ec009
MW
2451 * Each InfiniBand node is required to provide a Subnet Management Agent
2452 * that the subnet manager can access. Prior to the fabric being fully
2453 * configured by the subnet manager, the SMA is accessed via a well known
2454 * interface called the Subnet Management Interface (SMI). This interface
2455 * uses directed route packets to communicate with the SM to get around the
2456 * chicken and egg problem of the SM needing to know what's on the fabric
2457 * in order to configure the fabric, and needing to configure the fabric in
2458 * order to send packets to the devices on the fabric. These directed
2459 * route packets do not need the fabric fully configured in order to reach
2460 * their destination. The SMI is the only method allowed to send
2461 * directed route packets on an InfiniBand fabric.
29541e3a 2462 *
296ec009 2463 * Return: true if the port provides an SMI.
29541e3a 2464 */
5ede9289 2465static inline bool rdma_cap_ib_smi(const struct ib_device *device, u8 port_num)
29541e3a 2466{
f9b22e35 2467 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IB_SMI;
29541e3a
MW
2468}
2469
72219cea
MW
2470/**
2471 * rdma_cap_ib_cm - Check if the port of device has the capability Infiniband
2472 * Communication Manager.
296ec009
MW
2473 * @device: Device to check
2474 * @port_num: Port number to check
72219cea 2475 *
296ec009
MW
2476 * The InfiniBand Communication Manager is one of many pre-defined General
2477 * Service Agents (GSA) that are accessed via the General Service
2478 * Interface (GSI). It's role is to facilitate establishment of connections
2479 * between nodes as well as other management related tasks for established
2480 * connections.
72219cea 2481 *
296ec009
MW
2482 * Return: true if the port supports an IB CM (this does not guarantee that
2483 * a CM is actually running however).
72219cea 2484 */
5ede9289 2485static inline bool rdma_cap_ib_cm(const struct ib_device *device, u8 port_num)
72219cea 2486{
f9b22e35 2487 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IB_CM;
72219cea
MW
2488}
2489
04215330
MW
2490/**
2491 * rdma_cap_iw_cm - Check if the port of device has the capability IWARP
2492 * Communication Manager.
296ec009
MW
2493 * @device: Device to check
2494 * @port_num: Port number to check
04215330 2495 *
296ec009
MW
2496 * Similar to above, but specific to iWARP connections which have a different
2497 * managment protocol than InfiniBand.
04215330 2498 *
296ec009
MW
2499 * Return: true if the port supports an iWARP CM (this does not guarantee that
2500 * a CM is actually running however).
04215330 2501 */
5ede9289 2502static inline bool rdma_cap_iw_cm(const struct ib_device *device, u8 port_num)
04215330 2503{
f9b22e35 2504 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IW_CM;
04215330
MW
2505}
2506
fe53ba2f
MW
2507/**
2508 * rdma_cap_ib_sa - Check if the port of device has the capability Infiniband
2509 * Subnet Administration.
296ec009
MW
2510 * @device: Device to check
2511 * @port_num: Port number to check
fe53ba2f 2512 *
296ec009
MW
2513 * An InfiniBand Subnet Administration (SA) service is a pre-defined General
2514 * Service Agent (GSA) provided by the Subnet Manager (SM). On InfiniBand
2515 * fabrics, devices should resolve routes to other hosts by contacting the
2516 * SA to query the proper route.
fe53ba2f 2517 *
296ec009
MW
2518 * Return: true if the port should act as a client to the fabric Subnet
2519 * Administration interface. This does not imply that the SA service is
2520 * running locally.
fe53ba2f 2521 */
5ede9289 2522static inline bool rdma_cap_ib_sa(const struct ib_device *device, u8 port_num)
fe53ba2f 2523{
f9b22e35 2524 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IB_SA;
fe53ba2f
MW
2525}
2526
a31ad3b0
MW
2527/**
2528 * rdma_cap_ib_mcast - Check if the port of device has the capability Infiniband
2529 * Multicast.
296ec009
MW
2530 * @device: Device to check
2531 * @port_num: Port number to check
a31ad3b0 2532 *
296ec009
MW
2533 * InfiniBand multicast registration is more complex than normal IPv4 or
2534 * IPv6 multicast registration. Each Host Channel Adapter must register
2535 * with the Subnet Manager when it wishes to join a multicast group. It
2536 * should do so only once regardless of how many queue pairs it subscribes
2537 * to this group. And it should leave the group only after all queue pairs
2538 * attached to the group have been detached.
a31ad3b0 2539 *
296ec009
MW
2540 * Return: true if the port must undertake the additional adminstrative
2541 * overhead of registering/unregistering with the SM and tracking of the
2542 * total number of queue pairs attached to the multicast group.
a31ad3b0 2543 */
5ede9289 2544static inline bool rdma_cap_ib_mcast(const struct ib_device *device, u8 port_num)
a31ad3b0
MW
2545{
2546 return rdma_cap_ib_sa(device, port_num);
2547}
2548
30a74ef4
MW
2549/**
2550 * rdma_cap_af_ib - Check if the port of device has the capability
2551 * Native Infiniband Address.
296ec009
MW
2552 * @device: Device to check
2553 * @port_num: Port number to check
30a74ef4 2554 *
296ec009
MW
2555 * InfiniBand addressing uses a port's GUID + Subnet Prefix to make a default
2556 * GID. RoCE uses a different mechanism, but still generates a GID via
2557 * a prescribed mechanism and port specific data.
30a74ef4 2558 *
296ec009
MW
2559 * Return: true if the port uses a GID address to identify devices on the
2560 * network.
30a74ef4 2561 */
5ede9289 2562static inline bool rdma_cap_af_ib(const struct ib_device *device, u8 port_num)
30a74ef4 2563{
f9b22e35 2564 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_AF_IB;
30a74ef4
MW
2565}
2566
227128fc
MW
2567/**
2568 * rdma_cap_eth_ah - Check if the port of device has the capability
296ec009
MW
2569 * Ethernet Address Handle.
2570 * @device: Device to check
2571 * @port_num: Port number to check
227128fc 2572 *
296ec009
MW
2573 * RoCE is InfiniBand over Ethernet, and it uses a well defined technique
2574 * to fabricate GIDs over Ethernet/IP specific addresses native to the
2575 * port. Normally, packet headers are generated by the sending host
2576 * adapter, but when sending connectionless datagrams, we must manually
2577 * inject the proper headers for the fabric we are communicating over.
227128fc 2578 *
296ec009
MW
2579 * Return: true if we are running as a RoCE port and must force the
2580 * addition of a Global Route Header built from our Ethernet Address
2581 * Handle into our header list for connectionless packets.
227128fc 2582 */
5ede9289 2583static inline bool rdma_cap_eth_ah(const struct ib_device *device, u8 port_num)
227128fc 2584{
f9b22e35 2585 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_ETH_AH;
227128fc
MW
2586}
2587
337877a4
IW
2588/**
2589 * rdma_max_mad_size - Return the max MAD size required by this RDMA Port.
2590 *
2591 * @device: Device
2592 * @port_num: Port number
2593 *
2594 * This MAD size includes the MAD headers and MAD payload. No other headers
2595 * are included.
2596 *
2597 * Return the max MAD size required by the Port. Will return 0 if the port
2598 * does not support MADs
2599 */
2600static inline size_t rdma_max_mad_size(const struct ib_device *device, u8 port_num)
2601{
2602 return device->port_immutable[port_num].max_mad_size;
2603}
2604
03db3a2d
MB
2605/**
2606 * rdma_cap_roce_gid_table - Check if the port of device uses roce_gid_table
2607 * @device: Device to check
2608 * @port_num: Port number to check
2609 *
2610 * RoCE GID table mechanism manages the various GIDs for a device.
2611 *
2612 * NOTE: if allocating the port's GID table has failed, this call will still
2613 * return true, but any RoCE GID table API will fail.
2614 *
2615 * Return: true if the port uses RoCE GID table mechanism in order to manage
2616 * its GIDs.
2617 */
2618static inline bool rdma_cap_roce_gid_table(const struct ib_device *device,
2619 u8 port_num)
2620{
2621 return rdma_protocol_roce(device, port_num) &&
2622 device->add_gid && device->del_gid;
2623}
2624
002516ed
CH
2625/*
2626 * Check if the device supports READ W/ INVALIDATE.
2627 */
2628static inline bool rdma_cap_read_inv(struct ib_device *dev, u32 port_num)
2629{
2630 /*
2631 * iWarp drivers must support READ W/ INVALIDATE. No other protocol
2632 * has support for it yet.
2633 */
2634 return rdma_protocol_iwarp(dev, port_num);
2635}
2636
1da177e4 2637int ib_query_gid(struct ib_device *device,
55ee3ab2
MB
2638 u8 port_num, int index, union ib_gid *gid,
2639 struct ib_gid_attr *attr);
1da177e4 2640
50174a7f
EC
2641int ib_set_vf_link_state(struct ib_device *device, int vf, u8 port,
2642 int state);
2643int ib_get_vf_config(struct ib_device *device, int vf, u8 port,
2644 struct ifla_vf_info *info);
2645int ib_get_vf_stats(struct ib_device *device, int vf, u8 port,
2646 struct ifla_vf_stats *stats);
2647int ib_set_vf_guid(struct ib_device *device, int vf, u8 port, u64 guid,
2648 int type);
2649
1da177e4
LT
2650int ib_query_pkey(struct ib_device *device,
2651 u8 port_num, u16 index, u16 *pkey);
2652
2653int ib_modify_device(struct ib_device *device,
2654 int device_modify_mask,
2655 struct ib_device_modify *device_modify);
2656
2657int ib_modify_port(struct ib_device *device,
2658 u8 port_num, int port_modify_mask,
2659 struct ib_port_modify *port_modify);
2660
5eb620c8 2661int ib_find_gid(struct ib_device *device, union ib_gid *gid,
b39ffa1d
MB
2662 enum ib_gid_type gid_type, struct net_device *ndev,
2663 u8 *port_num, u16 *index);
5eb620c8
YE
2664
2665int ib_find_pkey(struct ib_device *device,
2666 u8 port_num, u16 pkey, u16 *index);
2667
ed082d36
CH
2668enum ib_pd_flags {
2669 /*
2670 * Create a memory registration for all memory in the system and place
2671 * the rkey for it into pd->unsafe_global_rkey. This can be used by
2672 * ULPs to avoid the overhead of dynamic MRs.
2673 *
2674 * This flag is generally considered unsafe and must only be used in
2675 * extremly trusted environments. Every use of it will log a warning
2676 * in the kernel log.
2677 */
2678 IB_PD_UNSAFE_GLOBAL_RKEY = 0x01,
2679};
1da177e4 2680
ed082d36
CH
2681struct ib_pd *__ib_alloc_pd(struct ib_device *device, unsigned int flags,
2682 const char *caller);
2683#define ib_alloc_pd(device, flags) \
2684 __ib_alloc_pd((device), (flags), __func__)
7dd78647 2685void ib_dealloc_pd(struct ib_pd *pd);
1da177e4
LT
2686
2687/**
2688 * ib_create_ah - Creates an address handle for the given address vector.
2689 * @pd: The protection domain associated with the address handle.
2690 * @ah_attr: The attributes of the address vector.
2691 *
2692 * The address handle is used to reference a local or global destination
2693 * in all UD QP post sends.
2694 */
2695struct ib_ah *ib_create_ah(struct ib_pd *pd, struct ib_ah_attr *ah_attr);
2696
850d8fd7
MS
2697/**
2698 * ib_get_gids_from_rdma_hdr - Get sgid and dgid from GRH or IPv4 header
2699 * work completion.
2700 * @hdr: the L3 header to parse
2701 * @net_type: type of header to parse
2702 * @sgid: place to store source gid
2703 * @dgid: place to store destination gid
2704 */
2705int ib_get_gids_from_rdma_hdr(const union rdma_network_hdr *hdr,
2706 enum rdma_network_type net_type,
2707 union ib_gid *sgid, union ib_gid *dgid);
2708
2709/**
2710 * ib_get_rdma_header_version - Get the header version
2711 * @hdr: the L3 header to parse
2712 */
2713int ib_get_rdma_header_version(const union rdma_network_hdr *hdr);
2714
4e00d694
SH
2715/**
2716 * ib_init_ah_from_wc - Initializes address handle attributes from a
2717 * work completion.
2718 * @device: Device on which the received message arrived.
2719 * @port_num: Port on which the received message arrived.
2720 * @wc: Work completion associated with the received message.
2721 * @grh: References the received global route header. This parameter is
2722 * ignored unless the work completion indicates that the GRH is valid.
2723 * @ah_attr: Returned attributes that can be used when creating an address
2724 * handle for replying to the message.
2725 */
73cdaaee
IW
2726int ib_init_ah_from_wc(struct ib_device *device, u8 port_num,
2727 const struct ib_wc *wc, const struct ib_grh *grh,
2728 struct ib_ah_attr *ah_attr);
4e00d694 2729
513789ed
HR
2730/**
2731 * ib_create_ah_from_wc - Creates an address handle associated with the
2732 * sender of the specified work completion.
2733 * @pd: The protection domain associated with the address handle.
2734 * @wc: Work completion information associated with a received message.
2735 * @grh: References the received global route header. This parameter is
2736 * ignored unless the work completion indicates that the GRH is valid.
2737 * @port_num: The outbound port number to associate with the address.
2738 *
2739 * The address handle is used to reference a local or global destination
2740 * in all UD QP post sends.
2741 */
73cdaaee
IW
2742struct ib_ah *ib_create_ah_from_wc(struct ib_pd *pd, const struct ib_wc *wc,
2743 const struct ib_grh *grh, u8 port_num);
513789ed 2744
1da177e4
LT
2745/**
2746 * ib_modify_ah - Modifies the address vector associated with an address
2747 * handle.
2748 * @ah: The address handle to modify.
2749 * @ah_attr: The new address vector attributes to associate with the
2750 * address handle.
2751 */
2752int ib_modify_ah(struct ib_ah *ah, struct ib_ah_attr *ah_attr);
2753
2754/**
2755 * ib_query_ah - Queries the address vector associated with an address
2756 * handle.
2757 * @ah: The address handle to query.
2758 * @ah_attr: The address vector attributes associated with the address
2759 * handle.
2760 */
2761int ib_query_ah(struct ib_ah *ah, struct ib_ah_attr *ah_attr);
2762
2763/**
2764 * ib_destroy_ah - Destroys an address handle.
2765 * @ah: The address handle to destroy.
2766 */
2767int ib_destroy_ah(struct ib_ah *ah);
2768
d41fcc67
RD
2769/**
2770 * ib_create_srq - Creates a SRQ associated with the specified protection
2771 * domain.
2772 * @pd: The protection domain associated with the SRQ.
abb6e9ba
DB
2773 * @srq_init_attr: A list of initial attributes required to create the
2774 * SRQ. If SRQ creation succeeds, then the attributes are updated to
2775 * the actual capabilities of the created SRQ.
d41fcc67
RD
2776 *
2777 * srq_attr->max_wr and srq_attr->max_sge are read the determine the
2778 * requested size of the SRQ, and set to the actual values allocated
2779 * on return. If ib_create_srq() succeeds, then max_wr and max_sge
2780 * will always be at least as large as the requested values.
2781 */
2782struct ib_srq *ib_create_srq(struct ib_pd *pd,
2783 struct ib_srq_init_attr *srq_init_attr);
2784
2785/**
2786 * ib_modify_srq - Modifies the attributes for the specified SRQ.
2787 * @srq: The SRQ to modify.
2788 * @srq_attr: On input, specifies the SRQ attributes to modify. On output,
2789 * the current values of selected SRQ attributes are returned.
2790 * @srq_attr_mask: A bit-mask used to specify which attributes of the SRQ
2791 * are being modified.
2792 *
2793 * The mask may contain IB_SRQ_MAX_WR to resize the SRQ and/or
2794 * IB_SRQ_LIMIT to set the SRQ's limit and request notification when
2795 * the number of receives queued drops below the limit.
2796 */
2797int ib_modify_srq(struct ib_srq *srq,
2798 struct ib_srq_attr *srq_attr,
2799 enum ib_srq_attr_mask srq_attr_mask);
2800
2801/**
2802 * ib_query_srq - Returns the attribute list and current values for the
2803 * specified SRQ.
2804 * @srq: The SRQ to query.
2805 * @srq_attr: The attributes of the specified SRQ.
2806 */
2807int ib_query_srq(struct ib_srq *srq,
2808 struct ib_srq_attr *srq_attr);
2809
2810/**
2811 * ib_destroy_srq - Destroys the specified SRQ.
2812 * @srq: The SRQ to destroy.
2813 */
2814int ib_destroy_srq(struct ib_srq *srq);
2815
2816/**
2817 * ib_post_srq_recv - Posts a list of work requests to the specified SRQ.
2818 * @srq: The SRQ to post the work request on.
2819 * @recv_wr: A list of work requests to post on the receive queue.
2820 * @bad_recv_wr: On an immediate failure, this parameter will reference
2821 * the work request that failed to be posted on the QP.
2822 */
2823static inline int ib_post_srq_recv(struct ib_srq *srq,
2824 struct ib_recv_wr *recv_wr,
2825 struct ib_recv_wr **bad_recv_wr)
2826{
2827 return srq->device->post_srq_recv(srq, recv_wr, bad_recv_wr);
2828}
2829
1da177e4
LT
2830/**
2831 * ib_create_qp - Creates a QP associated with the specified protection
2832 * domain.
2833 * @pd: The protection domain associated with the QP.
abb6e9ba
DB
2834 * @qp_init_attr: A list of initial attributes required to create the
2835 * QP. If QP creation succeeds, then the attributes are updated to
2836 * the actual capabilities of the created QP.
1da177e4
LT
2837 */
2838struct ib_qp *ib_create_qp(struct ib_pd *pd,
2839 struct ib_qp_init_attr *qp_init_attr);
2840
2841/**
2842 * ib_modify_qp - Modifies the attributes for the specified QP and then
2843 * transitions the QP to the given state.
2844 * @qp: The QP to modify.
2845 * @qp_attr: On input, specifies the QP attributes to modify. On output,
2846 * the current values of selected QP attributes are returned.
2847 * @qp_attr_mask: A bit-mask used to specify which attributes of the QP
2848 * are being modified.
2849 */
2850int ib_modify_qp(struct ib_qp *qp,
2851 struct ib_qp_attr *qp_attr,
2852 int qp_attr_mask);
2853
2854/**
2855 * ib_query_qp - Returns the attribute list and current values for the
2856 * specified QP.
2857 * @qp: The QP to query.
2858 * @qp_attr: The attributes of the specified QP.
2859 * @qp_attr_mask: A bit-mask used to select specific attributes to query.
2860 * @qp_init_attr: Additional attributes of the selected QP.
2861 *
2862 * The qp_attr_mask may be used to limit the query to gathering only the
2863 * selected attributes.
2864 */
2865int ib_query_qp(struct ib_qp *qp,
2866 struct ib_qp_attr *qp_attr,
2867 int qp_attr_mask,
2868 struct ib_qp_init_attr *qp_init_attr);
2869
2870/**
2871 * ib_destroy_qp - Destroys the specified QP.
2872 * @qp: The QP to destroy.
2873 */
2874int ib_destroy_qp(struct ib_qp *qp);
2875
d3d72d90 2876/**
0e0ec7e0
SH
2877 * ib_open_qp - Obtain a reference to an existing sharable QP.
2878 * @xrcd - XRC domain
2879 * @qp_open_attr: Attributes identifying the QP to open.
2880 *
2881 * Returns a reference to a sharable QP.
2882 */
2883struct ib_qp *ib_open_qp(struct ib_xrcd *xrcd,
2884 struct ib_qp_open_attr *qp_open_attr);
2885
2886/**
2887 * ib_close_qp - Release an external reference to a QP.
d3d72d90
SH
2888 * @qp: The QP handle to release
2889 *
0e0ec7e0
SH
2890 * The opened QP handle is released by the caller. The underlying
2891 * shared QP is not destroyed until all internal references are released.
d3d72d90 2892 */
0e0ec7e0 2893int ib_close_qp(struct ib_qp *qp);
d3d72d90 2894
1da177e4
LT
2895/**
2896 * ib_post_send - Posts a list of work requests to the send queue of
2897 * the specified QP.
2898 * @qp: The QP to post the work request on.
2899 * @send_wr: A list of work requests to post on the send queue.
2900 * @bad_send_wr: On an immediate failure, this parameter will reference
2901 * the work request that failed to be posted on the QP.
55464d46
BVA
2902 *
2903 * While IBA Vol. 1 section 11.4.1.1 specifies that if an immediate
2904 * error is returned, the QP state shall not be affected,
2905 * ib_post_send() will return an immediate error after queueing any
2906 * earlier work requests in the list.
1da177e4
LT
2907 */
2908static inline int ib_post_send(struct ib_qp *qp,
2909 struct ib_send_wr *send_wr,
2910 struct ib_send_wr **bad_send_wr)
2911{
2912 return qp->device->post_send(qp, send_wr, bad_send_wr);
2913}
2914
2915/**
2916 * ib_post_recv - Posts a list of work requests to the receive queue of
2917 * the specified QP.
2918 * @qp: The QP to post the work request on.
2919 * @recv_wr: A list of work requests to post on the receive queue.
2920 * @bad_recv_wr: On an immediate failure, this parameter will reference
2921 * the work request that failed to be posted on the QP.
2922 */
2923static inline int ib_post_recv(struct ib_qp *qp,
2924 struct ib_recv_wr *recv_wr,
2925 struct ib_recv_wr **bad_recv_wr)
2926{
2927 return qp->device->post_recv(qp, recv_wr, bad_recv_wr);
2928}
2929
14d3a3b2
CH
2930struct ib_cq *ib_alloc_cq(struct ib_device *dev, void *private,
2931 int nr_cqe, int comp_vector, enum ib_poll_context poll_ctx);
2932void ib_free_cq(struct ib_cq *cq);
2933int ib_process_cq_direct(struct ib_cq *cq, int budget);
2934
1da177e4
LT
2935/**
2936 * ib_create_cq - Creates a CQ on the specified device.
2937 * @device: The device on which to create the CQ.
2938 * @comp_handler: A user-specified callback that is invoked when a
2939 * completion event occurs on the CQ.
2940 * @event_handler: A user-specified callback that is invoked when an
2941 * asynchronous event not associated with a completion occurs on the CQ.
2942 * @cq_context: Context associated with the CQ returned to the user via
2943 * the associated completion and event handlers.
8e37210b 2944 * @cq_attr: The attributes the CQ should be created upon.
1da177e4
LT
2945 *
2946 * Users can examine the cq structure to determine the actual CQ size.
2947 */
2948struct ib_cq *ib_create_cq(struct ib_device *device,
2949 ib_comp_handler comp_handler,
2950 void (*event_handler)(struct ib_event *, void *),
8e37210b
MB
2951 void *cq_context,
2952 const struct ib_cq_init_attr *cq_attr);
1da177e4
LT
2953
2954/**
2955 * ib_resize_cq - Modifies the capacity of the CQ.
2956 * @cq: The CQ to resize.
2957 * @cqe: The minimum size of the CQ.
2958 *
2959 * Users can examine the cq structure to determine the actual CQ size.
2960 */
2961int ib_resize_cq(struct ib_cq *cq, int cqe);
2962
2dd57162
EC
2963/**
2964 * ib_modify_cq - Modifies moderation params of the CQ
2965 * @cq: The CQ to modify.
2966 * @cq_count: number of CQEs that will trigger an event
2967 * @cq_period: max period of time in usec before triggering an event
2968 *
2969 */
2970int ib_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period);
2971
1da177e4
LT
2972/**
2973 * ib_destroy_cq - Destroys the specified CQ.
2974 * @cq: The CQ to destroy.
2975 */
2976int ib_destroy_cq(struct ib_cq *cq);
2977
2978/**
2979 * ib_poll_cq - poll a CQ for completion(s)
2980 * @cq:the CQ being polled
2981 * @num_entries:maximum number of completions to return
2982 * @wc:array of at least @num_entries &struct ib_wc where completions
2983 * will be returned
2984 *
2985 * Poll a CQ for (possibly multiple) completions. If the return value
2986 * is < 0, an error occurred. If the return value is >= 0, it is the
2987 * number of completions returned. If the return value is
2988 * non-negative and < num_entries, then the CQ was emptied.
2989 */
2990static inline int ib_poll_cq(struct ib_cq *cq, int num_entries,
2991 struct ib_wc *wc)
2992{
2993 return cq->device->poll_cq(cq, num_entries, wc);
2994}
2995
2996/**
2997 * ib_peek_cq - Returns the number of unreaped completions currently
2998 * on the specified CQ.
2999 * @cq: The CQ to peek.
3000 * @wc_cnt: A minimum number of unreaped completions to check for.
3001 *
3002 * If the number of unreaped completions is greater than or equal to wc_cnt,
3003 * this function returns wc_cnt, otherwise, it returns the actual number of
3004 * unreaped completions.
3005 */
3006int ib_peek_cq(struct ib_cq *cq, int wc_cnt);
3007
3008/**
3009 * ib_req_notify_cq - Request completion notification on a CQ.
3010 * @cq: The CQ to generate an event for.
ed23a727
RD
3011 * @flags:
3012 * Must contain exactly one of %IB_CQ_SOLICITED or %IB_CQ_NEXT_COMP
3013 * to request an event on the next solicited event or next work
3014 * completion at any type, respectively. %IB_CQ_REPORT_MISSED_EVENTS
3015 * may also be |ed in to request a hint about missed events, as
3016 * described below.
3017 *
3018 * Return Value:
3019 * < 0 means an error occurred while requesting notification
3020 * == 0 means notification was requested successfully, and if
3021 * IB_CQ_REPORT_MISSED_EVENTS was passed in, then no events
3022 * were missed and it is safe to wait for another event. In
3023 * this case is it guaranteed that any work completions added
3024 * to the CQ since the last CQ poll will trigger a completion
3025 * notification event.
3026 * > 0 is only returned if IB_CQ_REPORT_MISSED_EVENTS was passed
3027 * in. It means that the consumer must poll the CQ again to
3028 * make sure it is empty to avoid missing an event because of a
3029 * race between requesting notification and an entry being
3030 * added to the CQ. This return value means it is possible
3031 * (but not guaranteed) that a work completion has been added
3032 * to the CQ since the last poll without triggering a
3033 * completion notification event.
1da177e4
LT
3034 */
3035static inline int ib_req_notify_cq(struct ib_cq *cq,
ed23a727 3036 enum ib_cq_notify_flags flags)
1da177e4 3037{
ed23a727 3038 return cq->device->req_notify_cq(cq, flags);
1da177e4
LT
3039}
3040
3041/**
3042 * ib_req_ncomp_notif - Request completion notification when there are
3043 * at least the specified number of unreaped completions on the CQ.
3044 * @cq: The CQ to generate an event for.
3045 * @wc_cnt: The number of unreaped completions that should be on the
3046 * CQ before an event is generated.
3047 */
3048static inline int ib_req_ncomp_notif(struct ib_cq *cq, int wc_cnt)
3049{
3050 return cq->device->req_ncomp_notif ?
3051 cq->device->req_ncomp_notif(cq, wc_cnt) :
3052 -ENOSYS;
3053}
3054
9b513090
RC
3055/**
3056 * ib_dma_mapping_error - check a DMA addr for error
3057 * @dev: The device for which the dma_addr was created
3058 * @dma_addr: The DMA address to check
3059 */
3060static inline int ib_dma_mapping_error(struct ib_device *dev, u64 dma_addr)
3061{
0957c29f 3062 return dma_mapping_error(dev->dma_device, dma_addr);
9b513090
RC
3063}
3064
3065/**
3066 * ib_dma_map_single - Map a kernel virtual address to DMA address
3067 * @dev: The device for which the dma_addr is to be created
3068 * @cpu_addr: The kernel virtual address
3069 * @size: The size of the region in bytes
3070 * @direction: The direction of the DMA
3071 */
3072static inline u64 ib_dma_map_single(struct ib_device *dev,
3073 void *cpu_addr, size_t size,
3074 enum dma_data_direction direction)
3075{
0957c29f 3076 return dma_map_single(dev->dma_device, cpu_addr, size, direction);
9b513090
RC
3077}
3078
3079/**
3080 * ib_dma_unmap_single - Destroy a mapping created by ib_dma_map_single()
3081 * @dev: The device for which the DMA address was created
3082 * @addr: The DMA address
3083 * @size: The size of the region in bytes
3084 * @direction: The direction of the DMA
3085 */
3086static inline void ib_dma_unmap_single(struct ib_device *dev,
3087 u64 addr, size_t size,
3088 enum dma_data_direction direction)
3089{
0957c29f 3090 dma_unmap_single(dev->dma_device, addr, size, direction);
cb9fbc5c
AK
3091}
3092
9b513090
RC
3093/**
3094 * ib_dma_map_page - Map a physical page to DMA address
3095 * @dev: The device for which the dma_addr is to be created
3096 * @page: The page to be mapped
3097 * @offset: The offset within the page
3098 * @size: The size of the region in bytes
3099 * @direction: The direction of the DMA
3100 */
3101static inline u64 ib_dma_map_page(struct ib_device *dev,
3102 struct page *page,
3103 unsigned long offset,
3104 size_t size,
3105 enum dma_data_direction direction)
3106{
0957c29f 3107 return dma_map_page(dev->dma_device, page, offset, size, direction);
9b513090
RC
3108}
3109
3110/**
3111 * ib_dma_unmap_page - Destroy a mapping created by ib_dma_map_page()
3112 * @dev: The device for which the DMA address was created
3113 * @addr: The DMA address
3114 * @size: The size of the region in bytes
3115 * @direction: The direction of the DMA
3116 */
3117static inline void ib_dma_unmap_page(struct ib_device *dev,
3118 u64 addr, size_t size,
3119 enum dma_data_direction direction)
3120{
0957c29f 3121 dma_unmap_page(dev->dma_device, addr, size, direction);
9b513090
RC
3122}
3123
3124/**
3125 * ib_dma_map_sg - Map a scatter/gather list to DMA addresses
3126 * @dev: The device for which the DMA addresses are to be created
3127 * @sg: The array of scatter/gather entries
3128 * @nents: The number of scatter/gather entries
3129 * @direction: The direction of the DMA
3130 */
3131static inline int ib_dma_map_sg(struct ib_device *dev,
3132 struct scatterlist *sg, int nents,
3133 enum dma_data_direction direction)
3134{
0957c29f 3135 return dma_map_sg(dev->dma_device, sg, nents, direction);
9b513090
RC
3136}
3137
3138/**
3139 * ib_dma_unmap_sg - Unmap a scatter/gather list of DMA addresses
3140 * @dev: The device for which the DMA addresses were created
3141 * @sg: The array of scatter/gather entries
3142 * @nents: The number of scatter/gather entries
3143 * @direction: The direction of the DMA
3144 */
3145static inline void ib_dma_unmap_sg(struct ib_device *dev,
3146 struct scatterlist *sg, int nents,
3147 enum dma_data_direction direction)
3148{
0957c29f 3149 dma_unmap_sg(dev->dma_device, sg, nents, direction);
9b513090
RC
3150}
3151
cb9fbc5c
AK
3152static inline int ib_dma_map_sg_attrs(struct ib_device *dev,
3153 struct scatterlist *sg, int nents,
3154 enum dma_data_direction direction,
00085f1e 3155 unsigned long dma_attrs)
cb9fbc5c 3156{
0957c29f
BVA
3157 return dma_map_sg_attrs(dev->dma_device, sg, nents, direction,
3158 dma_attrs);
cb9fbc5c
AK
3159}
3160
3161static inline void ib_dma_unmap_sg_attrs(struct ib_device *dev,
3162 struct scatterlist *sg, int nents,
3163 enum dma_data_direction direction,
00085f1e 3164 unsigned long dma_attrs)
cb9fbc5c 3165{
0957c29f 3166 dma_unmap_sg_attrs(dev->dma_device, sg, nents, direction, dma_attrs);
cb9fbc5c 3167}
9b513090
RC
3168/**
3169 * ib_sg_dma_address - Return the DMA address from a scatter/gather entry
3170 * @dev: The device for which the DMA addresses were created
3171 * @sg: The scatter/gather entry
ea58a595
MM
3172 *
3173 * Note: this function is obsolete. To do: change all occurrences of
3174 * ib_sg_dma_address() into sg_dma_address().
9b513090
RC
3175 */
3176static inline u64 ib_sg_dma_address(struct ib_device *dev,
3177 struct scatterlist *sg)
3178{
d1998ef3 3179 return sg_dma_address(sg);
9b513090
RC
3180}
3181
3182/**
3183 * ib_sg_dma_len - Return the DMA length from a scatter/gather entry
3184 * @dev: The device for which the DMA addresses were created
3185 * @sg: The scatter/gather entry
ea58a595
MM
3186 *
3187 * Note: this function is obsolete. To do: change all occurrences of
3188 * ib_sg_dma_len() into sg_dma_len().
9b513090
RC
3189 */
3190static inline unsigned int ib_sg_dma_len(struct ib_device *dev,
3191 struct scatterlist *sg)
3192{
d1998ef3 3193 return sg_dma_len(sg);
9b513090
RC
3194}
3195
3196/**
3197 * ib_dma_sync_single_for_cpu - Prepare DMA region to be accessed by CPU
3198 * @dev: The device for which the DMA address was created
3199 * @addr: The DMA address
3200 * @size: The size of the region in bytes
3201 * @dir: The direction of the DMA
3202 */
3203static inline void ib_dma_sync_single_for_cpu(struct ib_device *dev,
3204 u64 addr,
3205 size_t size,
3206 enum dma_data_direction dir)
3207{
0957c29f 3208 dma_sync_single_for_cpu(dev->dma_device, addr, size, dir);
9b513090
RC
3209}
3210
3211/**
3212 * ib_dma_sync_single_for_device - Prepare DMA region to be accessed by device
3213 * @dev: The device for which the DMA address was created
3214 * @addr: The DMA address
3215 * @size: The size of the region in bytes
3216 * @dir: The direction of the DMA
3217 */
3218static inline void ib_dma_sync_single_for_device(struct ib_device *dev,
3219 u64 addr,
3220 size_t size,
3221 enum dma_data_direction dir)
3222{
0957c29f 3223 dma_sync_single_for_device(dev->dma_device, addr, size, dir);
9b513090
RC
3224}
3225
3226/**
3227 * ib_dma_alloc_coherent - Allocate memory and map it for DMA
3228 * @dev: The device for which the DMA address is requested
3229 * @size: The size of the region to allocate in bytes
3230 * @dma_handle: A pointer for returning the DMA address of the region
3231 * @flag: memory allocator flags
3232 */
3233static inline void *ib_dma_alloc_coherent(struct ib_device *dev,
3234 size_t size,
d43dbacf 3235 dma_addr_t *dma_handle,
9b513090
RC
3236 gfp_t flag)
3237{
0957c29f 3238 return dma_alloc_coherent(dev->dma_device, size, dma_handle, flag);
9b513090
RC
3239}
3240
3241/**
3242 * ib_dma_free_coherent - Free memory allocated by ib_dma_alloc_coherent()
3243 * @dev: The device for which the DMA addresses were allocated
3244 * @size: The size of the region
3245 * @cpu_addr: the address returned by ib_dma_alloc_coherent()
3246 * @dma_handle: the DMA address returned by ib_dma_alloc_coherent()
3247 */
3248static inline void ib_dma_free_coherent(struct ib_device *dev,
3249 size_t size, void *cpu_addr,
d43dbacf 3250 dma_addr_t dma_handle)
9b513090 3251{
0957c29f 3252 dma_free_coherent(dev->dma_device, size, cpu_addr, dma_handle);
9b513090
RC
3253}
3254
1da177e4
LT
3255/**
3256 * ib_dereg_mr - Deregisters a memory region and removes it from the
3257 * HCA translation table.
3258 * @mr: The memory region to deregister.
7083e42e
SM
3259 *
3260 * This function can fail, if the memory region has memory windows bound to it.
1da177e4
LT
3261 */
3262int ib_dereg_mr(struct ib_mr *mr);
3263
9bee178b
SG
3264struct ib_mr *ib_alloc_mr(struct ib_pd *pd,
3265 enum ib_mr_type mr_type,
3266 u32 max_num_sg);
00f7ec36 3267
00f7ec36
SW
3268/**
3269 * ib_update_fast_reg_key - updates the key portion of the fast_reg MR
3270 * R_Key and L_Key.
3271 * @mr - struct ib_mr pointer to be updated.
3272 * @newkey - new key to be used.
3273 */
3274static inline void ib_update_fast_reg_key(struct ib_mr *mr, u8 newkey)
3275{
3276 mr->lkey = (mr->lkey & 0xffffff00) | newkey;
3277 mr->rkey = (mr->rkey & 0xffffff00) | newkey;
3278}
3279
7083e42e
SM
3280/**
3281 * ib_inc_rkey - increments the key portion of the given rkey. Can be used
3282 * for calculating a new rkey for type 2 memory windows.
3283 * @rkey - the rkey to increment.
3284 */
3285static inline u32 ib_inc_rkey(u32 rkey)
3286{
3287 const u32 mask = 0x000000ff;
3288 return ((rkey + 1) & mask) | (rkey & ~mask);
3289}
3290
1da177e4
LT
3291/**
3292 * ib_alloc_fmr - Allocates a unmapped fast memory region.
3293 * @pd: The protection domain associated with the unmapped region.
3294 * @mr_access_flags: Specifies the memory access rights.
3295 * @fmr_attr: Attributes of the unmapped region.
3296 *
3297 * A fast memory region must be mapped before it can be used as part of
3298 * a work request.
3299 */
3300struct ib_fmr *ib_alloc_fmr(struct ib_pd *pd,
3301 int mr_access_flags,
3302 struct ib_fmr_attr *fmr_attr);
3303
3304/**
3305 * ib_map_phys_fmr - Maps a list of physical pages to a fast memory region.
3306 * @fmr: The fast memory region to associate with the pages.
3307 * @page_list: An array of physical pages to map to the fast memory region.
3308 * @list_len: The number of pages in page_list.
3309 * @iova: The I/O virtual address to use with the mapped region.
3310 */
3311static inline int ib_map_phys_fmr(struct ib_fmr *fmr,
3312 u64 *page_list, int list_len,
3313 u64 iova)
3314{
3315 return fmr->device->map_phys_fmr(fmr, page_list, list_len, iova);
3316}
3317
3318/**
3319 * ib_unmap_fmr - Removes the mapping from a list of fast memory regions.
3320 * @fmr_list: A linked list of fast memory regions to unmap.
3321 */
3322int ib_unmap_fmr(struct list_head *fmr_list);
3323
3324/**
3325 * ib_dealloc_fmr - Deallocates a fast memory region.
3326 * @fmr: The fast memory region to deallocate.
3327 */
3328int ib_dealloc_fmr(struct ib_fmr *fmr);
3329
3330/**
3331 * ib_attach_mcast - Attaches the specified QP to a multicast group.
3332 * @qp: QP to attach to the multicast group. The QP must be type
3333 * IB_QPT_UD.
3334 * @gid: Multicast group GID.
3335 * @lid: Multicast group LID in host byte order.
3336 *
3337 * In order to send and receive multicast packets, subnet
3338 * administration must have created the multicast group and configured
3339 * the fabric appropriately. The port associated with the specified
3340 * QP must also be a member of the multicast group.
3341 */
3342int ib_attach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid);
3343
3344/**
3345 * ib_detach_mcast - Detaches the specified QP from a multicast group.
3346 * @qp: QP to detach from the multicast group.
3347 * @gid: Multicast group GID.
3348 * @lid: Multicast group LID in host byte order.
3349 */
3350int ib_detach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid);
3351
59991f94
SH
3352/**
3353 * ib_alloc_xrcd - Allocates an XRC domain.
3354 * @device: The device on which to allocate the XRC domain.
3355 */
3356struct ib_xrcd *ib_alloc_xrcd(struct ib_device *device);
3357
3358/**
3359 * ib_dealloc_xrcd - Deallocates an XRC domain.
3360 * @xrcd: The XRC domain to deallocate.
3361 */
3362int ib_dealloc_xrcd(struct ib_xrcd *xrcd);
3363
319a441d
HHZ
3364struct ib_flow *ib_create_flow(struct ib_qp *qp,
3365 struct ib_flow_attr *flow_attr, int domain);
3366int ib_destroy_flow(struct ib_flow *flow_id);
3367
1c636f80
EC
3368static inline int ib_check_mr_access(int flags)
3369{
3370 /*
3371 * Local write permission is required if remote write or
3372 * remote atomic permission is also requested.
3373 */
3374 if (flags & (IB_ACCESS_REMOTE_ATOMIC | IB_ACCESS_REMOTE_WRITE) &&
3375 !(flags & IB_ACCESS_LOCAL_WRITE))
3376 return -EINVAL;
3377
3378 return 0;
3379}
3380
1b01d335
SG
3381/**
3382 * ib_check_mr_status: lightweight check of MR status.
3383 * This routine may provide status checks on a selected
3384 * ib_mr. first use is for signature status check.
3385 *
3386 * @mr: A memory region.
3387 * @check_mask: Bitmask of which checks to perform from
3388 * ib_mr_status_check enumeration.
3389 * @mr_status: The container of relevant status checks.
3390 * failed checks will be indicated in the status bitmask
3391 * and the relevant info shall be in the error item.
3392 */
3393int ib_check_mr_status(struct ib_mr *mr, u32 check_mask,
3394 struct ib_mr_status *mr_status);
3395
9268f72d
YK
3396struct net_device *ib_get_net_dev_by_params(struct ib_device *dev, u8 port,
3397 u16 pkey, const union ib_gid *gid,
3398 const struct sockaddr *addr);
5fd251c8
YH
3399struct ib_wq *ib_create_wq(struct ib_pd *pd,
3400 struct ib_wq_init_attr *init_attr);
3401int ib_destroy_wq(struct ib_wq *wq);
3402int ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *attr,
3403 u32 wq_attr_mask);
6d39786b
YH
3404struct ib_rwq_ind_table *ib_create_rwq_ind_table(struct ib_device *device,
3405 struct ib_rwq_ind_table_init_attr*
3406 wq_ind_table_init_attr);
3407int ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *wq_ind_table);
9268f72d 3408
ff2ba993 3409int ib_map_mr_sg(struct ib_mr *mr, struct scatterlist *sg, int sg_nents,
9aa8b321 3410 unsigned int *sg_offset, unsigned int page_size);
4c67e2bf
SG
3411
3412static inline int
ff2ba993 3413ib_map_mr_sg_zbva(struct ib_mr *mr, struct scatterlist *sg, int sg_nents,
9aa8b321 3414 unsigned int *sg_offset, unsigned int page_size)
4c67e2bf
SG
3415{
3416 int n;
3417
ff2ba993 3418 n = ib_map_mr_sg(mr, sg, sg_nents, sg_offset, page_size);
4c67e2bf
SG
3419 mr->iova = 0;
3420
3421 return n;
3422}
3423
ff2ba993 3424int ib_sg_to_pages(struct ib_mr *mr, struct scatterlist *sgl, int sg_nents,
9aa8b321 3425 unsigned int *sg_offset, int (*set_page)(struct ib_mr *, u64));
4c67e2bf 3426
765d6774
SW
3427void ib_drain_rq(struct ib_qp *qp);
3428void ib_drain_sq(struct ib_qp *qp);
3429void ib_drain_qp(struct ib_qp *qp);
850d8fd7 3430
c90ea9d8
MS
3431int ib_resolve_eth_dmac(struct ib_device *device,
3432 struct ib_ah_attr *ah_attr);
1da177e4 3433#endif /* IB_VERBS_H */