]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/rdma/ib_verbs.h
IB/core: Enforce PKey security on QPs
[mirror_ubuntu-bionic-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>
ea6819e1 65#include <uapi/rdma/ib_user_verbs.h>
1da177e4 66
f0626710 67extern struct workqueue_struct *ib_wq;
14d3a3b2 68extern struct workqueue_struct *ib_comp_wq;
f0626710 69
1da177e4
LT
70union ib_gid {
71 u8 raw[16];
72 struct {
97f52eb4
SH
73 __be64 subnet_prefix;
74 __be64 interface_id;
1da177e4
LT
75 } global;
76};
77
e26be1bf
MS
78extern union ib_gid zgid;
79
b39ffa1d
MB
80enum ib_gid_type {
81 /* If link layer is Ethernet, this is RoCE V1 */
82 IB_GID_TYPE_IB = 0,
83 IB_GID_TYPE_ROCE = 0,
7766a99f 84 IB_GID_TYPE_ROCE_UDP_ENCAP = 1,
b39ffa1d
MB
85 IB_GID_TYPE_SIZE
86};
87
7ead4bcb 88#define ROCE_V2_UDP_DPORT 4791
03db3a2d 89struct ib_gid_attr {
b39ffa1d 90 enum ib_gid_type gid_type;
03db3a2d
MB
91 struct net_device *ndev;
92};
93
07ebafba
TT
94enum rdma_node_type {
95 /* IB values map to NodeInfo:NodeType. */
96 RDMA_NODE_IB_CA = 1,
97 RDMA_NODE_IB_SWITCH,
98 RDMA_NODE_IB_ROUTER,
180771a3
UM
99 RDMA_NODE_RNIC,
100 RDMA_NODE_USNIC,
5db5765e 101 RDMA_NODE_USNIC_UDP,
1da177e4
LT
102};
103
a0c1b2a3
EC
104enum {
105 /* set the local administered indication */
106 IB_SA_WELL_KNOWN_GUID = BIT_ULL(57) | 2,
107};
108
07ebafba
TT
109enum rdma_transport_type {
110 RDMA_TRANSPORT_IB,
180771a3 111 RDMA_TRANSPORT_IWARP,
248567f7
UM
112 RDMA_TRANSPORT_USNIC,
113 RDMA_TRANSPORT_USNIC_UDP
07ebafba
TT
114};
115
6b90a6d6
MW
116enum rdma_protocol_type {
117 RDMA_PROTOCOL_IB,
118 RDMA_PROTOCOL_IBOE,
119 RDMA_PROTOCOL_IWARP,
120 RDMA_PROTOCOL_USNIC_UDP
121};
122
8385fd84
RD
123__attribute_const__ enum rdma_transport_type
124rdma_node_get_transport(enum rdma_node_type node_type);
07ebafba 125
c865f246
SK
126enum rdma_network_type {
127 RDMA_NETWORK_IB,
128 RDMA_NETWORK_ROCE_V1 = RDMA_NETWORK_IB,
129 RDMA_NETWORK_IPV4,
130 RDMA_NETWORK_IPV6
131};
132
133static inline enum ib_gid_type ib_network_to_gid_type(enum rdma_network_type network_type)
134{
135 if (network_type == RDMA_NETWORK_IPV4 ||
136 network_type == RDMA_NETWORK_IPV6)
137 return IB_GID_TYPE_ROCE_UDP_ENCAP;
138
139 /* IB_GID_TYPE_IB same as RDMA_NETWORK_ROCE_V1 */
140 return IB_GID_TYPE_IB;
141}
142
143static inline enum rdma_network_type ib_gid_to_network_type(enum ib_gid_type gid_type,
144 union ib_gid *gid)
145{
146 if (gid_type == IB_GID_TYPE_IB)
147 return RDMA_NETWORK_IB;
148
149 if (ipv6_addr_v4mapped((struct in6_addr *)gid))
150 return RDMA_NETWORK_IPV4;
151 else
152 return RDMA_NETWORK_IPV6;
153}
154
a3f5adaf
EC
155enum rdma_link_layer {
156 IB_LINK_LAYER_UNSPECIFIED,
157 IB_LINK_LAYER_INFINIBAND,
158 IB_LINK_LAYER_ETHERNET,
159};
160
1da177e4 161enum ib_device_cap_flags {
7ca0bc53
LR
162 IB_DEVICE_RESIZE_MAX_WR = (1 << 0),
163 IB_DEVICE_BAD_PKEY_CNTR = (1 << 1),
164 IB_DEVICE_BAD_QKEY_CNTR = (1 << 2),
165 IB_DEVICE_RAW_MULTI = (1 << 3),
166 IB_DEVICE_AUTO_PATH_MIG = (1 << 4),
167 IB_DEVICE_CHANGE_PHY_PORT = (1 << 5),
168 IB_DEVICE_UD_AV_PORT_ENFORCE = (1 << 6),
169 IB_DEVICE_CURR_QP_STATE_MOD = (1 << 7),
170 IB_DEVICE_SHUTDOWN_PORT = (1 << 8),
171 IB_DEVICE_INIT_TYPE = (1 << 9),
172 IB_DEVICE_PORT_ACTIVE_EVENT = (1 << 10),
173 IB_DEVICE_SYS_IMAGE_GUID = (1 << 11),
174 IB_DEVICE_RC_RNR_NAK_GEN = (1 << 12),
175 IB_DEVICE_SRQ_RESIZE = (1 << 13),
176 IB_DEVICE_N_NOTIFY_CQ = (1 << 14),
b1adc714
CH
177
178 /*
179 * This device supports a per-device lkey or stag that can be
180 * used without performing a memory registration for the local
181 * memory. Note that ULPs should never check this flag, but
182 * instead of use the local_dma_lkey flag in the ib_pd structure,
183 * which will always contain a usable lkey.
184 */
7ca0bc53
LR
185 IB_DEVICE_LOCAL_DMA_LKEY = (1 << 15),
186 IB_DEVICE_RESERVED /* old SEND_W_INV */ = (1 << 16),
187 IB_DEVICE_MEM_WINDOW = (1 << 17),
e0605d91
EC
188 /*
189 * Devices should set IB_DEVICE_UD_IP_SUM if they support
190 * insertion of UDP and TCP checksum on outgoing UD IPoIB
191 * messages and can verify the validity of checksum for
192 * incoming messages. Setting this flag implies that the
193 * IPoIB driver may set NETIF_F_IP_CSUM for datagram mode.
194 */
7ca0bc53
LR
195 IB_DEVICE_UD_IP_CSUM = (1 << 18),
196 IB_DEVICE_UD_TSO = (1 << 19),
197 IB_DEVICE_XRC = (1 << 20),
b1adc714
CH
198
199 /*
200 * This device supports the IB "base memory management extension",
201 * which includes support for fast registrations (IB_WR_REG_MR,
202 * IB_WR_LOCAL_INV and IB_WR_SEND_WITH_INV verbs). This flag should
203 * also be set by any iWarp device which must support FRs to comply
204 * to the iWarp verbs spec. iWarp devices also support the
205 * IB_WR_RDMA_READ_WITH_INV verb for RDMA READs that invalidate the
206 * stag.
207 */
7ca0bc53
LR
208 IB_DEVICE_MEM_MGT_EXTENSIONS = (1 << 21),
209 IB_DEVICE_BLOCK_MULTICAST_LOOPBACK = (1 << 22),
210 IB_DEVICE_MEM_WINDOW_TYPE_2A = (1 << 23),
211 IB_DEVICE_MEM_WINDOW_TYPE_2B = (1 << 24),
212 IB_DEVICE_RC_IP_CSUM = (1 << 25),
ebaaee25 213 /* Deprecated. Please use IB_RAW_PACKET_CAP_IP_CSUM. */
7ca0bc53 214 IB_DEVICE_RAW_IP_CSUM = (1 << 26),
8a06ce59
LR
215 /*
216 * Devices should set IB_DEVICE_CROSS_CHANNEL if they
217 * support execution of WQEs that involve synchronization
218 * of I/O operations with single completion queue managed
219 * by hardware.
220 */
221 IB_DEVICE_CROSS_CHANNEL = (1 << 27),
7ca0bc53
LR
222 IB_DEVICE_MANAGED_FLOW_STEERING = (1 << 29),
223 IB_DEVICE_SIGNATURE_HANDOVER = (1 << 30),
47355b3c 224 IB_DEVICE_ON_DEMAND_PAGING = (1ULL << 31),
f5aa9159 225 IB_DEVICE_SG_GAPS_REG = (1ULL << 32),
c7e162a4 226 IB_DEVICE_VIRTUAL_FUNCTION = (1ULL << 33),
ebaaee25 227 /* Deprecated. Please use IB_RAW_PACKET_CAP_SCATTER_FCS. */
c7e162a4 228 IB_DEVICE_RAW_SCATTER_FCS = (1ULL << 34),
62e45949 229 IB_DEVICE_RDMA_NETDEV_OPA_VNIC = (1ULL << 35),
1b01d335
SG
230};
231
232enum ib_signature_prot_cap {
233 IB_PROT_T10DIF_TYPE_1 = 1,
234 IB_PROT_T10DIF_TYPE_2 = 1 << 1,
235 IB_PROT_T10DIF_TYPE_3 = 1 << 2,
236};
237
238enum ib_signature_guard_cap {
239 IB_GUARD_T10DIF_CRC = 1,
240 IB_GUARD_T10DIF_CSUM = 1 << 1,
1da177e4
LT
241};
242
243enum ib_atomic_cap {
244 IB_ATOMIC_NONE,
245 IB_ATOMIC_HCA,
246 IB_ATOMIC_GLOB
247};
248
860f10a7 249enum ib_odp_general_cap_bits {
25bf14d6
AK
250 IB_ODP_SUPPORT = 1 << 0,
251 IB_ODP_SUPPORT_IMPLICIT = 1 << 1,
860f10a7
SG
252};
253
254enum ib_odp_transport_cap_bits {
255 IB_ODP_SUPPORT_SEND = 1 << 0,
256 IB_ODP_SUPPORT_RECV = 1 << 1,
257 IB_ODP_SUPPORT_WRITE = 1 << 2,
258 IB_ODP_SUPPORT_READ = 1 << 3,
259 IB_ODP_SUPPORT_ATOMIC = 1 << 4,
260};
261
262struct ib_odp_caps {
263 uint64_t general_caps;
264 struct {
265 uint32_t rc_odp_caps;
266 uint32_t uc_odp_caps;
267 uint32_t ud_odp_caps;
268 } per_transport_caps;
269};
270
ccf20562
YH
271struct ib_rss_caps {
272 /* Corresponding bit will be set if qp type from
273 * 'enum ib_qp_type' is supported, e.g.
274 * supported_qpts |= 1 << IB_QPT_UD
275 */
276 u32 supported_qpts;
277 u32 max_rwq_indirection_tables;
278 u32 max_rwq_indirection_table_size;
279};
280
b9926b92
MB
281enum ib_cq_creation_flags {
282 IB_CQ_FLAGS_TIMESTAMP_COMPLETION = 1 << 0,
8a06ce59 283 IB_CQ_FLAGS_IGNORE_OVERRUN = 1 << 1,
b9926b92
MB
284};
285
bcf4c1ea
MB
286struct ib_cq_init_attr {
287 unsigned int cqe;
288 int comp_vector;
289 u32 flags;
290};
291
1da177e4
LT
292struct ib_device_attr {
293 u64 fw_ver;
97f52eb4 294 __be64 sys_image_guid;
1da177e4
LT
295 u64 max_mr_size;
296 u64 page_size_cap;
297 u32 vendor_id;
298 u32 vendor_part_id;
299 u32 hw_ver;
300 int max_qp;
301 int max_qp_wr;
fb532d6a 302 u64 device_cap_flags;
1da177e4
LT
303 int max_sge;
304 int max_sge_rd;
305 int max_cq;
306 int max_cqe;
307 int max_mr;
308 int max_pd;
309 int max_qp_rd_atom;
310 int max_ee_rd_atom;
311 int max_res_rd_atom;
312 int max_qp_init_rd_atom;
313 int max_ee_init_rd_atom;
314 enum ib_atomic_cap atomic_cap;
5e80ba8f 315 enum ib_atomic_cap masked_atomic_cap;
1da177e4
LT
316 int max_ee;
317 int max_rdd;
318 int max_mw;
319 int max_raw_ipv6_qp;
320 int max_raw_ethy_qp;
321 int max_mcast_grp;
322 int max_mcast_qp_attach;
323 int max_total_mcast_qp_attach;
324 int max_ah;
325 int max_fmr;
326 int max_map_per_fmr;
327 int max_srq;
328 int max_srq_wr;
329 int max_srq_sge;
00f7ec36 330 unsigned int max_fast_reg_page_list_len;
1da177e4
LT
331 u16 max_pkeys;
332 u8 local_ca_ack_delay;
1b01d335
SG
333 int sig_prot_cap;
334 int sig_guard_cap;
860f10a7 335 struct ib_odp_caps odp_caps;
24306dc6
MB
336 uint64_t timestamp_mask;
337 uint64_t hca_core_clock; /* in KHZ */
ccf20562
YH
338 struct ib_rss_caps rss_caps;
339 u32 max_wq_type_rq;
ebaaee25 340 u32 raw_packet_caps; /* Use ib_raw_packet_caps enum */
1da177e4
LT
341};
342
343enum ib_mtu {
344 IB_MTU_256 = 1,
345 IB_MTU_512 = 2,
346 IB_MTU_1024 = 3,
347 IB_MTU_2048 = 4,
348 IB_MTU_4096 = 5
349};
350
351static inline int ib_mtu_enum_to_int(enum ib_mtu mtu)
352{
353 switch (mtu) {
354 case IB_MTU_256: return 256;
355 case IB_MTU_512: return 512;
356 case IB_MTU_1024: return 1024;
357 case IB_MTU_2048: return 2048;
358 case IB_MTU_4096: return 4096;
359 default: return -1;
360 }
361}
362
d3f4aadd
AR
363static inline enum ib_mtu ib_mtu_int_to_enum(int mtu)
364{
365 if (mtu >= 4096)
366 return IB_MTU_4096;
367 else if (mtu >= 2048)
368 return IB_MTU_2048;
369 else if (mtu >= 1024)
370 return IB_MTU_1024;
371 else if (mtu >= 512)
372 return IB_MTU_512;
373 else
374 return IB_MTU_256;
375}
376
1da177e4
LT
377enum ib_port_state {
378 IB_PORT_NOP = 0,
379 IB_PORT_DOWN = 1,
380 IB_PORT_INIT = 2,
381 IB_PORT_ARMED = 3,
382 IB_PORT_ACTIVE = 4,
383 IB_PORT_ACTIVE_DEFER = 5
384};
385
386enum ib_port_cap_flags {
387 IB_PORT_SM = 1 << 1,
388 IB_PORT_NOTICE_SUP = 1 << 2,
389 IB_PORT_TRAP_SUP = 1 << 3,
390 IB_PORT_OPT_IPD_SUP = 1 << 4,
391 IB_PORT_AUTO_MIGR_SUP = 1 << 5,
392 IB_PORT_SL_MAP_SUP = 1 << 6,
393 IB_PORT_MKEY_NVRAM = 1 << 7,
394 IB_PORT_PKEY_NVRAM = 1 << 8,
395 IB_PORT_LED_INFO_SUP = 1 << 9,
396 IB_PORT_SM_DISABLED = 1 << 10,
397 IB_PORT_SYS_IMAGE_GUID_SUP = 1 << 11,
398 IB_PORT_PKEY_SW_EXT_PORT_TRAP_SUP = 1 << 12,
71eeba16 399 IB_PORT_EXTENDED_SPEEDS_SUP = 1 << 14,
1da177e4
LT
400 IB_PORT_CM_SUP = 1 << 16,
401 IB_PORT_SNMP_TUNNEL_SUP = 1 << 17,
402 IB_PORT_REINIT_SUP = 1 << 18,
403 IB_PORT_DEVICE_MGMT_SUP = 1 << 19,
404 IB_PORT_VENDOR_CLASS_SUP = 1 << 20,
405 IB_PORT_DR_NOTICE_SUP = 1 << 21,
406 IB_PORT_CAP_MASK_NOTICE_SUP = 1 << 22,
407 IB_PORT_BOOT_MGMT_SUP = 1 << 23,
408 IB_PORT_LINK_LATENCY_SUP = 1 << 24,
b4a26a27 409 IB_PORT_CLIENT_REG_SUP = 1 << 25,
03db3a2d 410 IB_PORT_IP_BASED_GIDS = 1 << 26,
1da177e4
LT
411};
412
413enum ib_port_width {
414 IB_WIDTH_1X = 1,
415 IB_WIDTH_4X = 2,
416 IB_WIDTH_8X = 4,
417 IB_WIDTH_12X = 8
418};
419
420static inline int ib_width_enum_to_int(enum ib_port_width width)
421{
422 switch (width) {
423 case IB_WIDTH_1X: return 1;
424 case IB_WIDTH_4X: return 4;
425 case IB_WIDTH_8X: return 8;
426 case IB_WIDTH_12X: return 12;
427 default: return -1;
428 }
429}
430
2e96691c
OG
431enum ib_port_speed {
432 IB_SPEED_SDR = 1,
433 IB_SPEED_DDR = 2,
434 IB_SPEED_QDR = 4,
435 IB_SPEED_FDR10 = 8,
436 IB_SPEED_FDR = 16,
12113a35
NO
437 IB_SPEED_EDR = 32,
438 IB_SPEED_HDR = 64
2e96691c
OG
439};
440
b40f4757
CL
441/**
442 * struct rdma_hw_stats
443 * @timestamp - Used by the core code to track when the last update was
444 * @lifespan - Used by the core code to determine how old the counters
445 * should be before being updated again. Stored in jiffies, defaults
446 * to 10 milliseconds, drivers can override the default be specifying
447 * their own value during their allocation routine.
448 * @name - Array of pointers to static names used for the counters in
449 * directory.
450 * @num_counters - How many hardware counters there are. If name is
451 * shorter than this number, a kernel oops will result. Driver authors
452 * are encouraged to leave BUILD_BUG_ON(ARRAY_SIZE(@name) < num_counters)
453 * in their code to prevent this.
454 * @value - Array of u64 counters that are accessed by the sysfs code and
455 * filled in by the drivers get_stats routine
456 */
457struct rdma_hw_stats {
458 unsigned long timestamp;
459 unsigned long lifespan;
460 const char * const *names;
461 int num_counters;
462 u64 value[];
7f624d02
SW
463};
464
b40f4757
CL
465#define RDMA_HW_STATS_DEFAULT_LIFESPAN 10
466/**
467 * rdma_alloc_hw_stats_struct - Helper function to allocate dynamic struct
468 * for drivers.
469 * @names - Array of static const char *
470 * @num_counters - How many elements in array
471 * @lifespan - How many milliseconds between updates
472 */
473static inline struct rdma_hw_stats *rdma_alloc_hw_stats_struct(
474 const char * const *names, int num_counters,
475 unsigned long lifespan)
476{
477 struct rdma_hw_stats *stats;
478
479 stats = kzalloc(sizeof(*stats) + num_counters * sizeof(u64),
480 GFP_KERNEL);
481 if (!stats)
482 return NULL;
483 stats->names = names;
484 stats->num_counters = num_counters;
485 stats->lifespan = msecs_to_jiffies(lifespan);
486
487 return stats;
488}
489
490
f9b22e35
IW
491/* Define bits for the various functionality this port needs to be supported by
492 * the core.
493 */
494/* Management 0x00000FFF */
495#define RDMA_CORE_CAP_IB_MAD 0x00000001
496#define RDMA_CORE_CAP_IB_SMI 0x00000002
497#define RDMA_CORE_CAP_IB_CM 0x00000004
498#define RDMA_CORE_CAP_IW_CM 0x00000008
499#define RDMA_CORE_CAP_IB_SA 0x00000010
65995fee 500#define RDMA_CORE_CAP_OPA_MAD 0x00000020
f9b22e35
IW
501
502/* Address format 0x000FF000 */
503#define RDMA_CORE_CAP_AF_IB 0x00001000
504#define RDMA_CORE_CAP_ETH_AH 0x00002000
94d595c5 505#define RDMA_CORE_CAP_OPA_AH 0x00004000
f9b22e35
IW
506
507/* Protocol 0xFFF00000 */
508#define RDMA_CORE_CAP_PROT_IB 0x00100000
509#define RDMA_CORE_CAP_PROT_ROCE 0x00200000
510#define RDMA_CORE_CAP_PROT_IWARP 0x00400000
7766a99f 511#define RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP 0x00800000
aa773bd4 512#define RDMA_CORE_CAP_PROT_RAW_PACKET 0x01000000
ce1e055f 513#define RDMA_CORE_CAP_PROT_USNIC 0x02000000
f9b22e35
IW
514
515#define RDMA_CORE_PORT_IBA_IB (RDMA_CORE_CAP_PROT_IB \
516 | RDMA_CORE_CAP_IB_MAD \
517 | RDMA_CORE_CAP_IB_SMI \
518 | RDMA_CORE_CAP_IB_CM \
519 | RDMA_CORE_CAP_IB_SA \
520 | RDMA_CORE_CAP_AF_IB)
521#define RDMA_CORE_PORT_IBA_ROCE (RDMA_CORE_CAP_PROT_ROCE \
522 | RDMA_CORE_CAP_IB_MAD \
523 | RDMA_CORE_CAP_IB_CM \
f9b22e35
IW
524 | RDMA_CORE_CAP_AF_IB \
525 | RDMA_CORE_CAP_ETH_AH)
7766a99f
MB
526#define RDMA_CORE_PORT_IBA_ROCE_UDP_ENCAP \
527 (RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP \
528 | RDMA_CORE_CAP_IB_MAD \
529 | RDMA_CORE_CAP_IB_CM \
530 | RDMA_CORE_CAP_AF_IB \
531 | RDMA_CORE_CAP_ETH_AH)
f9b22e35
IW
532#define RDMA_CORE_PORT_IWARP (RDMA_CORE_CAP_PROT_IWARP \
533 | RDMA_CORE_CAP_IW_CM)
65995fee
IW
534#define RDMA_CORE_PORT_INTEL_OPA (RDMA_CORE_PORT_IBA_IB \
535 | RDMA_CORE_CAP_OPA_MAD)
f9b22e35 536
aa773bd4
OG
537#define RDMA_CORE_PORT_RAW_PACKET (RDMA_CORE_CAP_PROT_RAW_PACKET)
538
ce1e055f
OG
539#define RDMA_CORE_PORT_USNIC (RDMA_CORE_CAP_PROT_USNIC)
540
1da177e4 541struct ib_port_attr {
fad61ad4 542 u64 subnet_prefix;
1da177e4
LT
543 enum ib_port_state state;
544 enum ib_mtu max_mtu;
545 enum ib_mtu active_mtu;
546 int gid_tbl_len;
547 u32 port_cap_flags;
548 u32 max_msg_sz;
549 u32 bad_pkey_cntr;
550 u32 qkey_viol_cntr;
551 u16 pkey_tbl_len;
552 u16 lid;
553 u16 sm_lid;
554 u8 lmc;
555 u8 max_vl_num;
556 u8 sm_sl;
557 u8 subnet_timeout;
558 u8 init_type_reply;
559 u8 active_width;
560 u8 active_speed;
561 u8 phys_state;
a0c1b2a3 562 bool grh_required;
1da177e4
LT
563};
564
565enum ib_device_modify_flags {
c5bcbbb9
RD
566 IB_DEVICE_MODIFY_SYS_IMAGE_GUID = 1 << 0,
567 IB_DEVICE_MODIFY_NODE_DESC = 1 << 1
1da177e4
LT
568};
569
bd99fdea
YS
570#define IB_DEVICE_NODE_DESC_MAX 64
571
1da177e4
LT
572struct ib_device_modify {
573 u64 sys_image_guid;
bd99fdea 574 char node_desc[IB_DEVICE_NODE_DESC_MAX];
1da177e4
LT
575};
576
577enum ib_port_modify_flags {
578 IB_PORT_SHUTDOWN = 1,
579 IB_PORT_INIT_TYPE = (1<<2),
580 IB_PORT_RESET_QKEY_CNTR = (1<<3)
581};
582
583struct ib_port_modify {
584 u32 set_port_cap_mask;
585 u32 clr_port_cap_mask;
586 u8 init_type;
587};
588
589enum ib_event_type {
590 IB_EVENT_CQ_ERR,
591 IB_EVENT_QP_FATAL,
592 IB_EVENT_QP_REQ_ERR,
593 IB_EVENT_QP_ACCESS_ERR,
594 IB_EVENT_COMM_EST,
595 IB_EVENT_SQ_DRAINED,
596 IB_EVENT_PATH_MIG,
597 IB_EVENT_PATH_MIG_ERR,
598 IB_EVENT_DEVICE_FATAL,
599 IB_EVENT_PORT_ACTIVE,
600 IB_EVENT_PORT_ERR,
601 IB_EVENT_LID_CHANGE,
602 IB_EVENT_PKEY_CHANGE,
d41fcc67
RD
603 IB_EVENT_SM_CHANGE,
604 IB_EVENT_SRQ_ERR,
605 IB_EVENT_SRQ_LIMIT_REACHED,
63942c9a 606 IB_EVENT_QP_LAST_WQE_REACHED,
761d90ed
OG
607 IB_EVENT_CLIENT_REREGISTER,
608 IB_EVENT_GID_CHANGE,
f213c052 609 IB_EVENT_WQ_FATAL,
1da177e4
LT
610};
611
db7489e0 612const char *__attribute_const__ ib_event_msg(enum ib_event_type event);
2b1b5b60 613
1da177e4
LT
614struct ib_event {
615 struct ib_device *device;
616 union {
617 struct ib_cq *cq;
618 struct ib_qp *qp;
d41fcc67 619 struct ib_srq *srq;
f213c052 620 struct ib_wq *wq;
1da177e4
LT
621 u8 port_num;
622 } element;
623 enum ib_event_type event;
624};
625
626struct ib_event_handler {
627 struct ib_device *device;
628 void (*handler)(struct ib_event_handler *, struct ib_event *);
629 struct list_head list;
630};
631
632#define INIT_IB_EVENT_HANDLER(_ptr, _device, _handler) \
633 do { \
634 (_ptr)->device = _device; \
635 (_ptr)->handler = _handler; \
636 INIT_LIST_HEAD(&(_ptr)->list); \
637 } while (0)
638
639struct ib_global_route {
640 union ib_gid dgid;
641 u32 flow_label;
642 u8 sgid_index;
643 u8 hop_limit;
644 u8 traffic_class;
645};
646
513789ed 647struct ib_grh {
97f52eb4
SH
648 __be32 version_tclass_flow;
649 __be16 paylen;
513789ed
HR
650 u8 next_hdr;
651 u8 hop_limit;
652 union ib_gid sgid;
653 union ib_gid dgid;
654};
655
c865f246
SK
656union rdma_network_hdr {
657 struct ib_grh ibgrh;
658 struct {
659 /* The IB spec states that if it's IPv4, the header
660 * is located in the last 20 bytes of the header.
661 */
662 u8 reserved[20];
663 struct iphdr roce4grh;
664 };
665};
666
1da177e4
LT
667enum {
668 IB_MULTICAST_QPN = 0xffffff
669};
670
f3a7c66b 671#define IB_LID_PERMISSIVE cpu_to_be16(0xFFFF)
b4e64397 672#define IB_MULTICAST_LID_BASE cpu_to_be16(0xC000)
97f52eb4 673
1da177e4
LT
674enum ib_ah_flags {
675 IB_AH_GRH = 1
676};
677
bf6a9e31
JM
678enum ib_rate {
679 IB_RATE_PORT_CURRENT = 0,
680 IB_RATE_2_5_GBPS = 2,
681 IB_RATE_5_GBPS = 5,
682 IB_RATE_10_GBPS = 3,
683 IB_RATE_20_GBPS = 6,
684 IB_RATE_30_GBPS = 4,
685 IB_RATE_40_GBPS = 7,
686 IB_RATE_60_GBPS = 8,
687 IB_RATE_80_GBPS = 9,
71eeba16
MA
688 IB_RATE_120_GBPS = 10,
689 IB_RATE_14_GBPS = 11,
690 IB_RATE_56_GBPS = 12,
691 IB_RATE_112_GBPS = 13,
692 IB_RATE_168_GBPS = 14,
693 IB_RATE_25_GBPS = 15,
694 IB_RATE_100_GBPS = 16,
695 IB_RATE_200_GBPS = 17,
696 IB_RATE_300_GBPS = 18
bf6a9e31
JM
697};
698
699/**
700 * ib_rate_to_mult - Convert the IB rate enum to a multiple of the
701 * base rate of 2.5 Gbit/sec. For example, IB_RATE_5_GBPS will be
702 * converted to 2, since 5 Gbit/sec is 2 * 2.5 Gbit/sec.
703 * @rate: rate to convert.
704 */
8385fd84 705__attribute_const__ int ib_rate_to_mult(enum ib_rate rate);
bf6a9e31 706
71eeba16
MA
707/**
708 * ib_rate_to_mbps - Convert the IB rate enum to Mbps.
709 * For example, IB_RATE_2_5_GBPS will be converted to 2500.
710 * @rate: rate to convert.
711 */
8385fd84 712__attribute_const__ int ib_rate_to_mbps(enum ib_rate rate);
71eeba16 713
17cd3a2d
SG
714
715/**
9bee178b
SG
716 * enum ib_mr_type - memory region type
717 * @IB_MR_TYPE_MEM_REG: memory region that is used for
718 * normal registration
719 * @IB_MR_TYPE_SIGNATURE: memory region that is used for
720 * signature operations (data-integrity
721 * capable regions)
f5aa9159
SG
722 * @IB_MR_TYPE_SG_GAPS: memory region that is capable to
723 * register any arbitrary sg lists (without
724 * the normal mr constraints - see
725 * ib_map_mr_sg)
17cd3a2d 726 */
9bee178b
SG
727enum ib_mr_type {
728 IB_MR_TYPE_MEM_REG,
729 IB_MR_TYPE_SIGNATURE,
f5aa9159 730 IB_MR_TYPE_SG_GAPS,
17cd3a2d
SG
731};
732
1b01d335 733/**
78eda2bb
SG
734 * Signature types
735 * IB_SIG_TYPE_NONE: Unprotected.
736 * IB_SIG_TYPE_T10_DIF: Type T10-DIF
1b01d335 737 */
78eda2bb
SG
738enum ib_signature_type {
739 IB_SIG_TYPE_NONE,
740 IB_SIG_TYPE_T10_DIF,
1b01d335
SG
741};
742
743/**
744 * Signature T10-DIF block-guard types
745 * IB_T10DIF_CRC: Corresponds to T10-PI mandated CRC checksum rules.
746 * IB_T10DIF_CSUM: Corresponds to IP checksum rules.
747 */
748enum ib_t10_dif_bg_type {
749 IB_T10DIF_CRC,
750 IB_T10DIF_CSUM
751};
752
753/**
754 * struct ib_t10_dif_domain - Parameters specific for T10-DIF
755 * domain.
1b01d335
SG
756 * @bg_type: T10-DIF block guard type (CRC|CSUM)
757 * @pi_interval: protection information interval.
758 * @bg: seed of guard computation.
759 * @app_tag: application tag of guard block
760 * @ref_tag: initial guard block reference tag.
78eda2bb
SG
761 * @ref_remap: Indicate wethear the reftag increments each block
762 * @app_escape: Indicate to skip block check if apptag=0xffff
763 * @ref_escape: Indicate to skip block check if reftag=0xffffffff
764 * @apptag_check_mask: check bitmask of application tag.
1b01d335
SG
765 */
766struct ib_t10_dif_domain {
1b01d335
SG
767 enum ib_t10_dif_bg_type bg_type;
768 u16 pi_interval;
769 u16 bg;
770 u16 app_tag;
771 u32 ref_tag;
78eda2bb
SG
772 bool ref_remap;
773 bool app_escape;
774 bool ref_escape;
775 u16 apptag_check_mask;
1b01d335
SG
776};
777
778/**
779 * struct ib_sig_domain - Parameters for signature domain
780 * @sig_type: specific signauture type
781 * @sig: union of all signature domain attributes that may
782 * be used to set domain layout.
783 */
784struct ib_sig_domain {
785 enum ib_signature_type sig_type;
786 union {
787 struct ib_t10_dif_domain dif;
788 } sig;
789};
790
791/**
792 * struct ib_sig_attrs - Parameters for signature handover operation
793 * @check_mask: bitmask for signature byte check (8 bytes)
794 * @mem: memory domain layout desciptor.
795 * @wire: wire domain layout desciptor.
796 */
797struct ib_sig_attrs {
798 u8 check_mask;
799 struct ib_sig_domain mem;
800 struct ib_sig_domain wire;
801};
802
803enum ib_sig_err_type {
804 IB_SIG_BAD_GUARD,
805 IB_SIG_BAD_REFTAG,
806 IB_SIG_BAD_APPTAG,
807};
808
809/**
810 * struct ib_sig_err - signature error descriptor
811 */
812struct ib_sig_err {
813 enum ib_sig_err_type err_type;
814 u32 expected;
815 u32 actual;
816 u64 sig_err_offset;
817 u32 key;
818};
819
820enum ib_mr_status_check {
821 IB_MR_CHECK_SIG_STATUS = 1,
822};
823
824/**
825 * struct ib_mr_status - Memory region status container
826 *
827 * @fail_status: Bitmask of MR checks status. For each
828 * failed check a corresponding status bit is set.
829 * @sig_err: Additional info for IB_MR_CEHCK_SIG_STATUS
830 * failure.
831 */
832struct ib_mr_status {
833 u32 fail_status;
834 struct ib_sig_err sig_err;
835};
836
bf6a9e31
JM
837/**
838 * mult_to_ib_rate - Convert a multiple of 2.5 Gbit/sec to an IB rate
839 * enum.
840 * @mult: multiple to convert.
841 */
8385fd84 842__attribute_const__ enum ib_rate mult_to_ib_rate(int mult);
bf6a9e31 843
44c58487
DC
844enum rdma_ah_attr_type {
845 RDMA_AH_ATTR_TYPE_IB,
846 RDMA_AH_ATTR_TYPE_ROCE,
64b4646e 847 RDMA_AH_ATTR_TYPE_OPA,
44c58487
DC
848};
849
850struct ib_ah_attr {
851 u16 dlid;
852 u8 src_path_bits;
853};
854
855struct roce_ah_attr {
856 u8 dmac[ETH_ALEN];
857};
858
64b4646e
DC
859struct opa_ah_attr {
860 u32 dlid;
861 u8 src_path_bits;
862};
863
90898850 864struct rdma_ah_attr {
1da177e4 865 struct ib_global_route grh;
1da177e4 866 u8 sl;
1da177e4 867 u8 static_rate;
1da177e4 868 u8 port_num;
44c58487
DC
869 u8 ah_flags;
870 enum rdma_ah_attr_type type;
871 union {
872 struct ib_ah_attr ib;
873 struct roce_ah_attr roce;
64b4646e 874 struct opa_ah_attr opa;
44c58487 875 };
1da177e4
LT
876};
877
878enum ib_wc_status {
879 IB_WC_SUCCESS,
880 IB_WC_LOC_LEN_ERR,
881 IB_WC_LOC_QP_OP_ERR,
882 IB_WC_LOC_EEC_OP_ERR,
883 IB_WC_LOC_PROT_ERR,
884 IB_WC_WR_FLUSH_ERR,
885 IB_WC_MW_BIND_ERR,
886 IB_WC_BAD_RESP_ERR,
887 IB_WC_LOC_ACCESS_ERR,
888 IB_WC_REM_INV_REQ_ERR,
889 IB_WC_REM_ACCESS_ERR,
890 IB_WC_REM_OP_ERR,
891 IB_WC_RETRY_EXC_ERR,
892 IB_WC_RNR_RETRY_EXC_ERR,
893 IB_WC_LOC_RDD_VIOL_ERR,
894 IB_WC_REM_INV_RD_REQ_ERR,
895 IB_WC_REM_ABORT_ERR,
896 IB_WC_INV_EECN_ERR,
897 IB_WC_INV_EEC_STATE_ERR,
898 IB_WC_FATAL_ERR,
899 IB_WC_RESP_TIMEOUT_ERR,
900 IB_WC_GENERAL_ERR
901};
902
db7489e0 903const char *__attribute_const__ ib_wc_status_msg(enum ib_wc_status status);
2b1b5b60 904
1da177e4
LT
905enum ib_wc_opcode {
906 IB_WC_SEND,
907 IB_WC_RDMA_WRITE,
908 IB_WC_RDMA_READ,
909 IB_WC_COMP_SWAP,
910 IB_WC_FETCH_ADD,
c93570f2 911 IB_WC_LSO,
00f7ec36 912 IB_WC_LOCAL_INV,
4c67e2bf 913 IB_WC_REG_MR,
5e80ba8f
VS
914 IB_WC_MASKED_COMP_SWAP,
915 IB_WC_MASKED_FETCH_ADD,
1da177e4
LT
916/*
917 * Set value of IB_WC_RECV so consumers can test if a completion is a
918 * receive by testing (opcode & IB_WC_RECV).
919 */
920 IB_WC_RECV = 1 << 7,
921 IB_WC_RECV_RDMA_WITH_IMM
922};
923
924enum ib_wc_flags {
925 IB_WC_GRH = 1,
00f7ec36
SW
926 IB_WC_WITH_IMM = (1<<1),
927 IB_WC_WITH_INVALIDATE = (1<<2),
d927d505 928 IB_WC_IP_CSUM_OK = (1<<3),
dd5f03be
MB
929 IB_WC_WITH_SMAC = (1<<4),
930 IB_WC_WITH_VLAN = (1<<5),
c865f246 931 IB_WC_WITH_NETWORK_HDR_TYPE = (1<<6),
1da177e4
LT
932};
933
934struct ib_wc {
14d3a3b2
CH
935 union {
936 u64 wr_id;
937 struct ib_cqe *wr_cqe;
938 };
1da177e4
LT
939 enum ib_wc_status status;
940 enum ib_wc_opcode opcode;
941 u32 vendor_err;
942 u32 byte_len;
062dbb69 943 struct ib_qp *qp;
00f7ec36
SW
944 union {
945 __be32 imm_data;
946 u32 invalidate_rkey;
947 } ex;
1da177e4
LT
948 u32 src_qp;
949 int wc_flags;
950 u16 pkey_index;
951 u16 slid;
952 u8 sl;
953 u8 dlid_path_bits;
954 u8 port_num; /* valid only for DR SMPs on switches */
dd5f03be
MB
955 u8 smac[ETH_ALEN];
956 u16 vlan_id;
c865f246 957 u8 network_hdr_type;
1da177e4
LT
958};
959
ed23a727
RD
960enum ib_cq_notify_flags {
961 IB_CQ_SOLICITED = 1 << 0,
962 IB_CQ_NEXT_COMP = 1 << 1,
963 IB_CQ_SOLICITED_MASK = IB_CQ_SOLICITED | IB_CQ_NEXT_COMP,
964 IB_CQ_REPORT_MISSED_EVENTS = 1 << 2,
1da177e4
LT
965};
966
96104eda 967enum ib_srq_type {
418d5130
SH
968 IB_SRQT_BASIC,
969 IB_SRQT_XRC
96104eda
SH
970};
971
d41fcc67
RD
972enum ib_srq_attr_mask {
973 IB_SRQ_MAX_WR = 1 << 0,
974 IB_SRQ_LIMIT = 1 << 1,
975};
976
977struct ib_srq_attr {
978 u32 max_wr;
979 u32 max_sge;
980 u32 srq_limit;
981};
982
983struct ib_srq_init_attr {
984 void (*event_handler)(struct ib_event *, void *);
985 void *srq_context;
986 struct ib_srq_attr attr;
96104eda 987 enum ib_srq_type srq_type;
418d5130
SH
988
989 union {
990 struct {
991 struct ib_xrcd *xrcd;
992 struct ib_cq *cq;
993 } xrc;
994 } ext;
d41fcc67
RD
995};
996
1da177e4
LT
997struct ib_qp_cap {
998 u32 max_send_wr;
999 u32 max_recv_wr;
1000 u32 max_send_sge;
1001 u32 max_recv_sge;
1002 u32 max_inline_data;
a060b562
CH
1003
1004 /*
1005 * Maximum number of rdma_rw_ctx structures in flight at a time.
1006 * ib_create_qp() will calculate the right amount of neededed WRs
1007 * and MRs based on this.
1008 */
1009 u32 max_rdma_ctxs;
1da177e4
LT
1010};
1011
1012enum ib_sig_type {
1013 IB_SIGNAL_ALL_WR,
1014 IB_SIGNAL_REQ_WR
1015};
1016
1017enum ib_qp_type {
1018 /*
1019 * IB_QPT_SMI and IB_QPT_GSI have to be the first two entries
1020 * here (and in that order) since the MAD layer uses them as
1021 * indices into a 2-entry table.
1022 */
1023 IB_QPT_SMI,
1024 IB_QPT_GSI,
1025
1026 IB_QPT_RC,
1027 IB_QPT_UC,
1028 IB_QPT_UD,
1029 IB_QPT_RAW_IPV6,
b42b63cf 1030 IB_QPT_RAW_ETHERTYPE,
c938a616 1031 IB_QPT_RAW_PACKET = 8,
b42b63cf
SH
1032 IB_QPT_XRC_INI = 9,
1033 IB_QPT_XRC_TGT,
0134f16b
JM
1034 IB_QPT_MAX,
1035 /* Reserve a range for qp types internal to the low level driver.
1036 * These qp types will not be visible at the IB core layer, so the
1037 * IB_QPT_MAX usages should not be affected in the core layer
1038 */
1039 IB_QPT_RESERVED1 = 0x1000,
1040 IB_QPT_RESERVED2,
1041 IB_QPT_RESERVED3,
1042 IB_QPT_RESERVED4,
1043 IB_QPT_RESERVED5,
1044 IB_QPT_RESERVED6,
1045 IB_QPT_RESERVED7,
1046 IB_QPT_RESERVED8,
1047 IB_QPT_RESERVED9,
1048 IB_QPT_RESERVED10,
1da177e4
LT
1049};
1050
b846f25a 1051enum ib_qp_create_flags {
47ee1b9f
RL
1052 IB_QP_CREATE_IPOIB_UD_LSO = 1 << 0,
1053 IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK = 1 << 1,
8a06ce59
LR
1054 IB_QP_CREATE_CROSS_CHANNEL = 1 << 2,
1055 IB_QP_CREATE_MANAGED_SEND = 1 << 3,
1056 IB_QP_CREATE_MANAGED_RECV = 1 << 4,
90f1d1b4 1057 IB_QP_CREATE_NETIF_QP = 1 << 5,
1b01d335 1058 IB_QP_CREATE_SIGNATURE_EN = 1 << 6,
09b93088 1059 IB_QP_CREATE_USE_GFP_NOIO = 1 << 7,
b531b909 1060 IB_QP_CREATE_SCATTER_FCS = 1 << 8,
9c2b270e 1061 IB_QP_CREATE_CVLAN_STRIPPING = 1 << 9,
d2b57063
JM
1062 /* reserve bits 26-31 for low level drivers' internal use */
1063 IB_QP_CREATE_RESERVED_START = 1 << 26,
1064 IB_QP_CREATE_RESERVED_END = 1 << 31,
b846f25a
EC
1065};
1066
73c40c61
YH
1067/*
1068 * Note: users may not call ib_close_qp or ib_destroy_qp from the event_handler
1069 * callback to destroy the passed in QP.
1070 */
1071
1da177e4
LT
1072struct ib_qp_init_attr {
1073 void (*event_handler)(struct ib_event *, void *);
1074 void *qp_context;
1075 struct ib_cq *send_cq;
1076 struct ib_cq *recv_cq;
1077 struct ib_srq *srq;
b42b63cf 1078 struct ib_xrcd *xrcd; /* XRC TGT QPs only */
1da177e4
LT
1079 struct ib_qp_cap cap;
1080 enum ib_sig_type sq_sig_type;
1081 enum ib_qp_type qp_type;
b846f25a 1082 enum ib_qp_create_flags create_flags;
a060b562
CH
1083
1084 /*
1085 * Only needed for special QP types, or when using the RW API.
1086 */
1087 u8 port_num;
a9017e23 1088 struct ib_rwq_ind_table *rwq_ind_tbl;
1da177e4
LT
1089};
1090
0e0ec7e0
SH
1091struct ib_qp_open_attr {
1092 void (*event_handler)(struct ib_event *, void *);
1093 void *qp_context;
1094 u32 qp_num;
1095 enum ib_qp_type qp_type;
1096};
1097
1da177e4
LT
1098enum ib_rnr_timeout {
1099 IB_RNR_TIMER_655_36 = 0,
1100 IB_RNR_TIMER_000_01 = 1,
1101 IB_RNR_TIMER_000_02 = 2,
1102 IB_RNR_TIMER_000_03 = 3,
1103 IB_RNR_TIMER_000_04 = 4,
1104 IB_RNR_TIMER_000_06 = 5,
1105 IB_RNR_TIMER_000_08 = 6,
1106 IB_RNR_TIMER_000_12 = 7,
1107 IB_RNR_TIMER_000_16 = 8,
1108 IB_RNR_TIMER_000_24 = 9,
1109 IB_RNR_TIMER_000_32 = 10,
1110 IB_RNR_TIMER_000_48 = 11,
1111 IB_RNR_TIMER_000_64 = 12,
1112 IB_RNR_TIMER_000_96 = 13,
1113 IB_RNR_TIMER_001_28 = 14,
1114 IB_RNR_TIMER_001_92 = 15,
1115 IB_RNR_TIMER_002_56 = 16,
1116 IB_RNR_TIMER_003_84 = 17,
1117 IB_RNR_TIMER_005_12 = 18,
1118 IB_RNR_TIMER_007_68 = 19,
1119 IB_RNR_TIMER_010_24 = 20,
1120 IB_RNR_TIMER_015_36 = 21,
1121 IB_RNR_TIMER_020_48 = 22,
1122 IB_RNR_TIMER_030_72 = 23,
1123 IB_RNR_TIMER_040_96 = 24,
1124 IB_RNR_TIMER_061_44 = 25,
1125 IB_RNR_TIMER_081_92 = 26,
1126 IB_RNR_TIMER_122_88 = 27,
1127 IB_RNR_TIMER_163_84 = 28,
1128 IB_RNR_TIMER_245_76 = 29,
1129 IB_RNR_TIMER_327_68 = 30,
1130 IB_RNR_TIMER_491_52 = 31
1131};
1132
1133enum ib_qp_attr_mask {
1134 IB_QP_STATE = 1,
1135 IB_QP_CUR_STATE = (1<<1),
1136 IB_QP_EN_SQD_ASYNC_NOTIFY = (1<<2),
1137 IB_QP_ACCESS_FLAGS = (1<<3),
1138 IB_QP_PKEY_INDEX = (1<<4),
1139 IB_QP_PORT = (1<<5),
1140 IB_QP_QKEY = (1<<6),
1141 IB_QP_AV = (1<<7),
1142 IB_QP_PATH_MTU = (1<<8),
1143 IB_QP_TIMEOUT = (1<<9),
1144 IB_QP_RETRY_CNT = (1<<10),
1145 IB_QP_RNR_RETRY = (1<<11),
1146 IB_QP_RQ_PSN = (1<<12),
1147 IB_QP_MAX_QP_RD_ATOMIC = (1<<13),
1148 IB_QP_ALT_PATH = (1<<14),
1149 IB_QP_MIN_RNR_TIMER = (1<<15),
1150 IB_QP_SQ_PSN = (1<<16),
1151 IB_QP_MAX_DEST_RD_ATOMIC = (1<<17),
1152 IB_QP_PATH_MIG_STATE = (1<<18),
1153 IB_QP_CAP = (1<<19),
dd5f03be 1154 IB_QP_DEST_QPN = (1<<20),
aa744cc0
MB
1155 IB_QP_RESERVED1 = (1<<21),
1156 IB_QP_RESERVED2 = (1<<22),
1157 IB_QP_RESERVED3 = (1<<23),
1158 IB_QP_RESERVED4 = (1<<24),
528e5a1b 1159 IB_QP_RATE_LIMIT = (1<<25),
1da177e4
LT
1160};
1161
1162enum ib_qp_state {
1163 IB_QPS_RESET,
1164 IB_QPS_INIT,
1165 IB_QPS_RTR,
1166 IB_QPS_RTS,
1167 IB_QPS_SQD,
1168 IB_QPS_SQE,
1169 IB_QPS_ERR
1170};
1171
1172enum ib_mig_state {
1173 IB_MIG_MIGRATED,
1174 IB_MIG_REARM,
1175 IB_MIG_ARMED
1176};
1177
7083e42e
SM
1178enum ib_mw_type {
1179 IB_MW_TYPE_1 = 1,
1180 IB_MW_TYPE_2 = 2
1181};
1182
1da177e4
LT
1183struct ib_qp_attr {
1184 enum ib_qp_state qp_state;
1185 enum ib_qp_state cur_qp_state;
1186 enum ib_mtu path_mtu;
1187 enum ib_mig_state path_mig_state;
1188 u32 qkey;
1189 u32 rq_psn;
1190 u32 sq_psn;
1191 u32 dest_qp_num;
1192 int qp_access_flags;
1193 struct ib_qp_cap cap;
90898850
DC
1194 struct rdma_ah_attr ah_attr;
1195 struct rdma_ah_attr alt_ah_attr;
1da177e4
LT
1196 u16 pkey_index;
1197 u16 alt_pkey_index;
1198 u8 en_sqd_async_notify;
1199 u8 sq_draining;
1200 u8 max_rd_atomic;
1201 u8 max_dest_rd_atomic;
1202 u8 min_rnr_timer;
1203 u8 port_num;
1204 u8 timeout;
1205 u8 retry_cnt;
1206 u8 rnr_retry;
1207 u8 alt_port_num;
1208 u8 alt_timeout;
528e5a1b 1209 u32 rate_limit;
1da177e4
LT
1210};
1211
1212enum ib_wr_opcode {
1213 IB_WR_RDMA_WRITE,
1214 IB_WR_RDMA_WRITE_WITH_IMM,
1215 IB_WR_SEND,
1216 IB_WR_SEND_WITH_IMM,
1217 IB_WR_RDMA_READ,
1218 IB_WR_ATOMIC_CMP_AND_SWP,
c93570f2 1219 IB_WR_ATOMIC_FETCH_AND_ADD,
0f39cf3d
RD
1220 IB_WR_LSO,
1221 IB_WR_SEND_WITH_INV,
00f7ec36
SW
1222 IB_WR_RDMA_READ_WITH_INV,
1223 IB_WR_LOCAL_INV,
4c67e2bf 1224 IB_WR_REG_MR,
5e80ba8f
VS
1225 IB_WR_MASKED_ATOMIC_CMP_AND_SWP,
1226 IB_WR_MASKED_ATOMIC_FETCH_AND_ADD,
1b01d335 1227 IB_WR_REG_SIG_MR,
0134f16b
JM
1228 /* reserve values for low level drivers' internal use.
1229 * These values will not be used at all in the ib core layer.
1230 */
1231 IB_WR_RESERVED1 = 0xf0,
1232 IB_WR_RESERVED2,
1233 IB_WR_RESERVED3,
1234 IB_WR_RESERVED4,
1235 IB_WR_RESERVED5,
1236 IB_WR_RESERVED6,
1237 IB_WR_RESERVED7,
1238 IB_WR_RESERVED8,
1239 IB_WR_RESERVED9,
1240 IB_WR_RESERVED10,
1da177e4
LT
1241};
1242
1243enum ib_send_flags {
1244 IB_SEND_FENCE = 1,
1245 IB_SEND_SIGNALED = (1<<1),
1246 IB_SEND_SOLICITED = (1<<2),
e0605d91 1247 IB_SEND_INLINE = (1<<3),
0134f16b
JM
1248 IB_SEND_IP_CSUM = (1<<4),
1249
1250 /* reserve bits 26-31 for low level drivers' internal use */
1251 IB_SEND_RESERVED_START = (1 << 26),
1252 IB_SEND_RESERVED_END = (1 << 31),
1da177e4
LT
1253};
1254
1255struct ib_sge {
1256 u64 addr;
1257 u32 length;
1258 u32 lkey;
1259};
1260
14d3a3b2
CH
1261struct ib_cqe {
1262 void (*done)(struct ib_cq *cq, struct ib_wc *wc);
1263};
1264
1da177e4
LT
1265struct ib_send_wr {
1266 struct ib_send_wr *next;
14d3a3b2
CH
1267 union {
1268 u64 wr_id;
1269 struct ib_cqe *wr_cqe;
1270 };
1da177e4
LT
1271 struct ib_sge *sg_list;
1272 int num_sge;
1273 enum ib_wr_opcode opcode;
1274 int send_flags;
0f39cf3d
RD
1275 union {
1276 __be32 imm_data;
1277 u32 invalidate_rkey;
1278 } ex;
1da177e4
LT
1279};
1280
e622f2f4
CH
1281struct ib_rdma_wr {
1282 struct ib_send_wr wr;
1283 u64 remote_addr;
1284 u32 rkey;
1285};
1286
1287static inline struct ib_rdma_wr *rdma_wr(struct ib_send_wr *wr)
1288{
1289 return container_of(wr, struct ib_rdma_wr, wr);
1290}
1291
1292struct ib_atomic_wr {
1293 struct ib_send_wr wr;
1294 u64 remote_addr;
1295 u64 compare_add;
1296 u64 swap;
1297 u64 compare_add_mask;
1298 u64 swap_mask;
1299 u32 rkey;
1300};
1301
1302static inline struct ib_atomic_wr *atomic_wr(struct ib_send_wr *wr)
1303{
1304 return container_of(wr, struct ib_atomic_wr, wr);
1305}
1306
1307struct ib_ud_wr {
1308 struct ib_send_wr wr;
1309 struct ib_ah *ah;
1310 void *header;
1311 int hlen;
1312 int mss;
1313 u32 remote_qpn;
1314 u32 remote_qkey;
1315 u16 pkey_index; /* valid for GSI only */
1316 u8 port_num; /* valid for DR SMPs on switch only */
1317};
1318
1319static inline struct ib_ud_wr *ud_wr(struct ib_send_wr *wr)
1320{
1321 return container_of(wr, struct ib_ud_wr, wr);
1322}
1323
4c67e2bf
SG
1324struct ib_reg_wr {
1325 struct ib_send_wr wr;
1326 struct ib_mr *mr;
1327 u32 key;
1328 int access;
1329};
1330
1331static inline struct ib_reg_wr *reg_wr(struct ib_send_wr *wr)
1332{
1333 return container_of(wr, struct ib_reg_wr, wr);
1334}
1335
e622f2f4
CH
1336struct ib_sig_handover_wr {
1337 struct ib_send_wr wr;
1338 struct ib_sig_attrs *sig_attrs;
1339 struct ib_mr *sig_mr;
1340 int access_flags;
1341 struct ib_sge *prot;
1342};
1343
1344static inline struct ib_sig_handover_wr *sig_handover_wr(struct ib_send_wr *wr)
1345{
1346 return container_of(wr, struct ib_sig_handover_wr, wr);
1347}
1348
1da177e4
LT
1349struct ib_recv_wr {
1350 struct ib_recv_wr *next;
14d3a3b2
CH
1351 union {
1352 u64 wr_id;
1353 struct ib_cqe *wr_cqe;
1354 };
1da177e4
LT
1355 struct ib_sge *sg_list;
1356 int num_sge;
1357};
1358
1359enum ib_access_flags {
1360 IB_ACCESS_LOCAL_WRITE = 1,
1361 IB_ACCESS_REMOTE_WRITE = (1<<1),
1362 IB_ACCESS_REMOTE_READ = (1<<2),
1363 IB_ACCESS_REMOTE_ATOMIC = (1<<3),
7083e42e 1364 IB_ACCESS_MW_BIND = (1<<4),
860f10a7
SG
1365 IB_ZERO_BASED = (1<<5),
1366 IB_ACCESS_ON_DEMAND = (1<<6),
0008b84e 1367 IB_ACCESS_HUGETLB = (1<<7),
1da177e4
LT
1368};
1369
b7d3e0a9
CH
1370/*
1371 * XXX: these are apparently used for ->rereg_user_mr, no idea why they
1372 * are hidden here instead of a uapi header!
1373 */
1da177e4
LT
1374enum ib_mr_rereg_flags {
1375 IB_MR_REREG_TRANS = 1,
1376 IB_MR_REREG_PD = (1<<1),
7e6edb9b
MB
1377 IB_MR_REREG_ACCESS = (1<<2),
1378 IB_MR_REREG_SUPPORTED = ((IB_MR_REREG_ACCESS << 1) - 1)
1da177e4
LT
1379};
1380
1da177e4
LT
1381struct ib_fmr_attr {
1382 int max_pages;
1383 int max_maps;
d36f34aa 1384 u8 page_shift;
1da177e4
LT
1385};
1386
882214e2
HE
1387struct ib_umem;
1388
38321256
MB
1389enum rdma_remove_reason {
1390 /* Userspace requested uobject deletion. Call could fail */
1391 RDMA_REMOVE_DESTROY,
1392 /* Context deletion. This call should delete the actual object itself */
1393 RDMA_REMOVE_CLOSE,
1394 /* Driver is being hot-unplugged. This call should delete the actual object itself */
1395 RDMA_REMOVE_DRIVER_REMOVE,
1396 /* Context is being cleaned-up, but commit was just completed */
1397 RDMA_REMOVE_DURING_CLEANUP,
1398};
1399
43579b5f
PP
1400struct ib_rdmacg_object {
1401#ifdef CONFIG_CGROUP_RDMA
1402 struct rdma_cgroup *cg; /* owner rdma cgroup */
1403#endif
1404};
1405
e2773c06
RD
1406struct ib_ucontext {
1407 struct ib_device *device;
771addf6 1408 struct ib_uverbs_file *ufile;
f7c6a7b5 1409 int closing;
8ada2c1c 1410
38321256
MB
1411 /* locking the uobjects_list */
1412 struct mutex uobjects_lock;
1413 struct list_head uobjects;
1414 /* protects cleanup process from other actions */
1415 struct rw_semaphore cleanup_rwsem;
1416 enum rdma_remove_reason cleanup_reason;
1417
8ada2c1c 1418 struct pid *tgid;
882214e2
HE
1419#ifdef CONFIG_INFINIBAND_ON_DEMAND_PAGING
1420 struct rb_root umem_tree;
1421 /*
1422 * Protects .umem_rbroot and tree, as well as odp_mrs_count and
1423 * mmu notifiers registration.
1424 */
1425 struct rw_semaphore umem_rwsem;
1426 void (*invalidate_range)(struct ib_umem *umem,
1427 unsigned long start, unsigned long end);
1428
1429 struct mmu_notifier mn;
1430 atomic_t notifier_count;
1431 /* A list of umems that don't have private mmu notifier counters yet. */
1432 struct list_head no_private_counters;
1433 int odp_mrs_count;
1434#endif
43579b5f
PP
1435
1436 struct ib_rdmacg_object cg_obj;
e2773c06
RD
1437};
1438
1439struct ib_uobject {
1440 u64 user_handle; /* handle given to us by userspace */
1441 struct ib_ucontext *context; /* associated user context */
9ead190b 1442 void *object; /* containing object */
e2773c06 1443 struct list_head list; /* link to context's list */
43579b5f 1444 struct ib_rdmacg_object cg_obj; /* rdmacg object */
b3d636b0 1445 int id; /* index into kernel idr */
9ead190b 1446 struct kref ref;
38321256 1447 atomic_t usecnt; /* protects exclusive access */
d144da8c 1448 struct rcu_head rcu; /* kfree_rcu() overhead */
38321256
MB
1449
1450 const struct uverbs_obj_type *type;
e2773c06
RD
1451};
1452
cf8966b3
MB
1453struct ib_uobject_file {
1454 struct ib_uobject uobj;
1455 /* ufile contains the lock between context release and file close */
1456 struct ib_uverbs_file *ufile;
e2773c06
RD
1457};
1458
e2773c06 1459struct ib_udata {
309243ec 1460 const void __user *inbuf;
e2773c06
RD
1461 void __user *outbuf;
1462 size_t inlen;
1463 size_t outlen;
1464};
1465
1da177e4 1466struct ib_pd {
96249d70 1467 u32 local_dma_lkey;
ed082d36 1468 u32 flags;
e2773c06
RD
1469 struct ib_device *device;
1470 struct ib_uobject *uobject;
1471 atomic_t usecnt; /* count all resources */
50d46335 1472
ed082d36
CH
1473 u32 unsafe_global_rkey;
1474
50d46335
CH
1475 /*
1476 * Implementation details of the RDMA core, don't use in drivers:
1477 */
1478 struct ib_mr *__internal_mr;
1da177e4
LT
1479};
1480
59991f94
SH
1481struct ib_xrcd {
1482 struct ib_device *device;
d3d72d90 1483 atomic_t usecnt; /* count all exposed resources */
53d0bd1e 1484 struct inode *inode;
d3d72d90
SH
1485
1486 struct mutex tgt_qp_mutex;
1487 struct list_head tgt_qp_list;
59991f94
SH
1488};
1489
1da177e4
LT
1490struct ib_ah {
1491 struct ib_device *device;
1492 struct ib_pd *pd;
e2773c06 1493 struct ib_uobject *uobject;
44c58487 1494 enum rdma_ah_attr_type type;
1da177e4
LT
1495};
1496
1497typedef void (*ib_comp_handler)(struct ib_cq *cq, void *cq_context);
1498
14d3a3b2
CH
1499enum ib_poll_context {
1500 IB_POLL_DIRECT, /* caller context, no hw completions */
1501 IB_POLL_SOFTIRQ, /* poll from softirq context */
1502 IB_POLL_WORKQUEUE, /* poll from workqueue */
1503};
1504
1da177e4 1505struct ib_cq {
e2773c06
RD
1506 struct ib_device *device;
1507 struct ib_uobject *uobject;
1508 ib_comp_handler comp_handler;
1509 void (*event_handler)(struct ib_event *, void *);
4deccd6d 1510 void *cq_context;
e2773c06
RD
1511 int cqe;
1512 atomic_t usecnt; /* count number of work queues */
14d3a3b2
CH
1513 enum ib_poll_context poll_ctx;
1514 struct ib_wc *wc;
1515 union {
1516 struct irq_poll iop;
1517 struct work_struct work;
1518 };
1da177e4
LT
1519};
1520
1521struct ib_srq {
d41fcc67
RD
1522 struct ib_device *device;
1523 struct ib_pd *pd;
1524 struct ib_uobject *uobject;
1525 void (*event_handler)(struct ib_event *, void *);
1526 void *srq_context;
96104eda 1527 enum ib_srq_type srq_type;
1da177e4 1528 atomic_t usecnt;
418d5130
SH
1529
1530 union {
1531 struct {
1532 struct ib_xrcd *xrcd;
1533 struct ib_cq *cq;
1534 u32 srq_num;
1535 } xrc;
1536 } ext;
1da177e4
LT
1537};
1538
ebaaee25
NO
1539enum ib_raw_packet_caps {
1540 /* Strip cvlan from incoming packet and report it in the matching work
1541 * completion is supported.
1542 */
1543 IB_RAW_PACKET_CAP_CVLAN_STRIPPING = (1 << 0),
1544 /* Scatter FCS field of an incoming packet to host memory is supported.
1545 */
1546 IB_RAW_PACKET_CAP_SCATTER_FCS = (1 << 1),
1547 /* Checksum offloads are supported (for both send and receive). */
1548 IB_RAW_PACKET_CAP_IP_CSUM = (1 << 2),
1549};
1550
5fd251c8
YH
1551enum ib_wq_type {
1552 IB_WQT_RQ
1553};
1554
1555enum ib_wq_state {
1556 IB_WQS_RESET,
1557 IB_WQS_RDY,
1558 IB_WQS_ERR
1559};
1560
1561struct ib_wq {
1562 struct ib_device *device;
1563 struct ib_uobject *uobject;
1564 void *wq_context;
1565 void (*event_handler)(struct ib_event *, void *);
1566 struct ib_pd *pd;
1567 struct ib_cq *cq;
1568 u32 wq_num;
1569 enum ib_wq_state state;
1570 enum ib_wq_type wq_type;
1571 atomic_t usecnt;
1572};
1573
10bac72b
NO
1574enum ib_wq_flags {
1575 IB_WQ_FLAGS_CVLAN_STRIPPING = 1 << 0,
27b0df11 1576 IB_WQ_FLAGS_SCATTER_FCS = 1 << 1,
10bac72b
NO
1577};
1578
5fd251c8
YH
1579struct ib_wq_init_attr {
1580 void *wq_context;
1581 enum ib_wq_type wq_type;
1582 u32 max_wr;
1583 u32 max_sge;
1584 struct ib_cq *cq;
1585 void (*event_handler)(struct ib_event *, void *);
10bac72b 1586 u32 create_flags; /* Use enum ib_wq_flags */
5fd251c8
YH
1587};
1588
1589enum ib_wq_attr_mask {
10bac72b
NO
1590 IB_WQ_STATE = 1 << 0,
1591 IB_WQ_CUR_STATE = 1 << 1,
1592 IB_WQ_FLAGS = 1 << 2,
5fd251c8
YH
1593};
1594
1595struct ib_wq_attr {
1596 enum ib_wq_state wq_state;
1597 enum ib_wq_state curr_wq_state;
10bac72b
NO
1598 u32 flags; /* Use enum ib_wq_flags */
1599 u32 flags_mask; /* Use enum ib_wq_flags */
5fd251c8
YH
1600};
1601
6d39786b
YH
1602struct ib_rwq_ind_table {
1603 struct ib_device *device;
1604 struct ib_uobject *uobject;
1605 atomic_t usecnt;
1606 u32 ind_tbl_num;
1607 u32 log_ind_tbl_size;
1608 struct ib_wq **ind_tbl;
1609};
1610
1611struct ib_rwq_ind_table_init_attr {
1612 u32 log_ind_tbl_size;
1613 /* Each entry is a pointer to Receive Work Queue */
1614 struct ib_wq **ind_tbl;
1615};
1616
d291f1a6
DJ
1617enum port_pkey_state {
1618 IB_PORT_PKEY_NOT_VALID = 0,
1619 IB_PORT_PKEY_VALID = 1,
1620 IB_PORT_PKEY_LISTED = 2,
1621};
1622
1623struct ib_qp_security;
1624
1625struct ib_port_pkey {
1626 enum port_pkey_state state;
1627 u16 pkey_index;
1628 u8 port_num;
1629 struct list_head qp_list;
1630 struct list_head to_error_list;
1631 struct ib_qp_security *sec;
1632};
1633
1634struct ib_ports_pkeys {
1635 struct ib_port_pkey main;
1636 struct ib_port_pkey alt;
1637};
1638
1639struct ib_qp_security {
1640 struct ib_qp *qp;
1641 struct ib_device *dev;
1642 /* Hold this mutex when changing port and pkey settings. */
1643 struct mutex mutex;
1644 struct ib_ports_pkeys *ports_pkeys;
1645 /* A list of all open shared QP handles. Required to enforce security
1646 * properly for all users of a shared QP.
1647 */
1648 struct list_head shared_qp_list;
1649 void *security;
1650 bool destroying;
1651 atomic_t error_list_count;
1652 struct completion error_complete;
1653 int error_comps_pending;
1654};
1655
632bc3f6
BVA
1656/*
1657 * @max_write_sge: Maximum SGE elements per RDMA WRITE request.
1658 * @max_read_sge: Maximum SGE elements per RDMA READ request.
1659 */
1da177e4
LT
1660struct ib_qp {
1661 struct ib_device *device;
1662 struct ib_pd *pd;
1663 struct ib_cq *send_cq;
1664 struct ib_cq *recv_cq;
fffb0383
CH
1665 spinlock_t mr_lock;
1666 int mrs_used;
a060b562 1667 struct list_head rdma_mrs;
0e353e34 1668 struct list_head sig_mrs;
1da177e4 1669 struct ib_srq *srq;
b42b63cf 1670 struct ib_xrcd *xrcd; /* XRC TGT QPs only */
d3d72d90 1671 struct list_head xrcd_list;
fffb0383 1672
319a441d
HHZ
1673 /* count times opened, mcast attaches, flow attaches */
1674 atomic_t usecnt;
0e0ec7e0
SH
1675 struct list_head open_list;
1676 struct ib_qp *real_qp;
e2773c06 1677 struct ib_uobject *uobject;
1da177e4
LT
1678 void (*event_handler)(struct ib_event *, void *);
1679 void *qp_context;
1680 u32 qp_num;
632bc3f6
BVA
1681 u32 max_write_sge;
1682 u32 max_read_sge;
1da177e4 1683 enum ib_qp_type qp_type;
a9017e23 1684 struct ib_rwq_ind_table *rwq_ind_tbl;
d291f1a6 1685 struct ib_qp_security *qp_sec;
1da177e4
LT
1686};
1687
1688struct ib_mr {
e2773c06
RD
1689 struct ib_device *device;
1690 struct ib_pd *pd;
e2773c06
RD
1691 u32 lkey;
1692 u32 rkey;
4c67e2bf
SG
1693 u64 iova;
1694 u32 length;
1695 unsigned int page_size;
d4a85c30 1696 bool need_inval;
fffb0383
CH
1697 union {
1698 struct ib_uobject *uobject; /* user */
1699 struct list_head qp_entry; /* FR */
1700 };
1da177e4
LT
1701};
1702
1703struct ib_mw {
1704 struct ib_device *device;
1705 struct ib_pd *pd;
e2773c06 1706 struct ib_uobject *uobject;
1da177e4 1707 u32 rkey;
7083e42e 1708 enum ib_mw_type type;
1da177e4
LT
1709};
1710
1711struct ib_fmr {
1712 struct ib_device *device;
1713 struct ib_pd *pd;
1714 struct list_head list;
1715 u32 lkey;
1716 u32 rkey;
1717};
1718
319a441d
HHZ
1719/* Supported steering options */
1720enum ib_flow_attr_type {
1721 /* steering according to rule specifications */
1722 IB_FLOW_ATTR_NORMAL = 0x0,
1723 /* default unicast and multicast rule -
1724 * receive all Eth traffic which isn't steered to any QP
1725 */
1726 IB_FLOW_ATTR_ALL_DEFAULT = 0x1,
1727 /* default multicast rule -
1728 * receive all Eth multicast traffic which isn't steered to any QP
1729 */
1730 IB_FLOW_ATTR_MC_DEFAULT = 0x2,
1731 /* sniffer rule - receive all port traffic */
1732 IB_FLOW_ATTR_SNIFFER = 0x3
1733};
1734
1735/* Supported steering header types */
1736enum ib_flow_spec_type {
1737 /* L2 headers*/
76bd23b3
MR
1738 IB_FLOW_SPEC_ETH = 0x20,
1739 IB_FLOW_SPEC_IB = 0x22,
319a441d 1740 /* L3 header*/
76bd23b3
MR
1741 IB_FLOW_SPEC_IPV4 = 0x30,
1742 IB_FLOW_SPEC_IPV6 = 0x31,
319a441d 1743 /* L4 headers*/
76bd23b3
MR
1744 IB_FLOW_SPEC_TCP = 0x40,
1745 IB_FLOW_SPEC_UDP = 0x41,
0dbf3332 1746 IB_FLOW_SPEC_VXLAN_TUNNEL = 0x50,
fbf46860 1747 IB_FLOW_SPEC_INNER = 0x100,
460d0198
MR
1748 /* Actions */
1749 IB_FLOW_SPEC_ACTION_TAG = 0x1000,
483a3966 1750 IB_FLOW_SPEC_ACTION_DROP = 0x1001,
319a441d 1751};
240ae00e 1752#define IB_FLOW_SPEC_LAYER_MASK 0xF0
fbf46860 1753#define IB_FLOW_SPEC_SUPPORT_LAYERS 8
22878dbc 1754
319a441d
HHZ
1755/* Flow steering rule priority is set according to it's domain.
1756 * Lower domain value means higher priority.
1757 */
1758enum ib_flow_domain {
1759 IB_FLOW_DOMAIN_USER,
1760 IB_FLOW_DOMAIN_ETHTOOL,
1761 IB_FLOW_DOMAIN_RFS,
1762 IB_FLOW_DOMAIN_NIC,
1763 IB_FLOW_DOMAIN_NUM /* Must be last */
1764};
1765
a3100a78
MV
1766enum ib_flow_flags {
1767 IB_FLOW_ATTR_FLAGS_DONT_TRAP = 1UL << 1, /* Continue match, no steal */
1768 IB_FLOW_ATTR_FLAGS_RESERVED = 1UL << 2 /* Must be last */
1769};
1770
319a441d
HHZ
1771struct ib_flow_eth_filter {
1772 u8 dst_mac[6];
1773 u8 src_mac[6];
1774 __be16 ether_type;
1775 __be16 vlan_tag;
15dfbd6b
MG
1776 /* Must be last */
1777 u8 real_sz[0];
319a441d
HHZ
1778};
1779
1780struct ib_flow_spec_eth {
fbf46860 1781 u32 type;
319a441d
HHZ
1782 u16 size;
1783 struct ib_flow_eth_filter val;
1784 struct ib_flow_eth_filter mask;
1785};
1786
240ae00e
MB
1787struct ib_flow_ib_filter {
1788 __be16 dlid;
1789 __u8 sl;
15dfbd6b
MG
1790 /* Must be last */
1791 u8 real_sz[0];
240ae00e
MB
1792};
1793
1794struct ib_flow_spec_ib {
fbf46860 1795 u32 type;
240ae00e
MB
1796 u16 size;
1797 struct ib_flow_ib_filter val;
1798 struct ib_flow_ib_filter mask;
1799};
1800
989a3a8f
MG
1801/* IPv4 header flags */
1802enum ib_ipv4_flags {
1803 IB_IPV4_DONT_FRAG = 0x2, /* Don't enable packet fragmentation */
1804 IB_IPV4_MORE_FRAG = 0X4 /* For All fragmented packets except the
1805 last have this flag set */
1806};
1807
319a441d
HHZ
1808struct ib_flow_ipv4_filter {
1809 __be32 src_ip;
1810 __be32 dst_ip;
989a3a8f
MG
1811 u8 proto;
1812 u8 tos;
1813 u8 ttl;
1814 u8 flags;
15dfbd6b
MG
1815 /* Must be last */
1816 u8 real_sz[0];
319a441d
HHZ
1817};
1818
1819struct ib_flow_spec_ipv4 {
fbf46860 1820 u32 type;
319a441d
HHZ
1821 u16 size;
1822 struct ib_flow_ipv4_filter val;
1823 struct ib_flow_ipv4_filter mask;
1824};
1825
4c2aae71
MG
1826struct ib_flow_ipv6_filter {
1827 u8 src_ip[16];
1828 u8 dst_ip[16];
a72c6a2b
MG
1829 __be32 flow_label;
1830 u8 next_hdr;
1831 u8 traffic_class;
1832 u8 hop_limit;
15dfbd6b
MG
1833 /* Must be last */
1834 u8 real_sz[0];
4c2aae71
MG
1835};
1836
1837struct ib_flow_spec_ipv6 {
fbf46860 1838 u32 type;
4c2aae71
MG
1839 u16 size;
1840 struct ib_flow_ipv6_filter val;
1841 struct ib_flow_ipv6_filter mask;
1842};
1843
319a441d
HHZ
1844struct ib_flow_tcp_udp_filter {
1845 __be16 dst_port;
1846 __be16 src_port;
15dfbd6b
MG
1847 /* Must be last */
1848 u8 real_sz[0];
319a441d
HHZ
1849};
1850
1851struct ib_flow_spec_tcp_udp {
fbf46860 1852 u32 type;
319a441d
HHZ
1853 u16 size;
1854 struct ib_flow_tcp_udp_filter val;
1855 struct ib_flow_tcp_udp_filter mask;
1856};
1857
0dbf3332
MR
1858struct ib_flow_tunnel_filter {
1859 __be32 tunnel_id;
1860 u8 real_sz[0];
1861};
1862
1863/* ib_flow_spec_tunnel describes the Vxlan tunnel
1864 * the tunnel_id from val has the vni value
1865 */
1866struct ib_flow_spec_tunnel {
fbf46860 1867 u32 type;
0dbf3332
MR
1868 u16 size;
1869 struct ib_flow_tunnel_filter val;
1870 struct ib_flow_tunnel_filter mask;
1871};
1872
460d0198
MR
1873struct ib_flow_spec_action_tag {
1874 enum ib_flow_spec_type type;
1875 u16 size;
1876 u32 tag_id;
1877};
1878
483a3966
SS
1879struct ib_flow_spec_action_drop {
1880 enum ib_flow_spec_type type;
1881 u16 size;
1882};
1883
319a441d
HHZ
1884union ib_flow_spec {
1885 struct {
fbf46860 1886 u32 type;
319a441d
HHZ
1887 u16 size;
1888 };
1889 struct ib_flow_spec_eth eth;
240ae00e 1890 struct ib_flow_spec_ib ib;
319a441d
HHZ
1891 struct ib_flow_spec_ipv4 ipv4;
1892 struct ib_flow_spec_tcp_udp tcp_udp;
4c2aae71 1893 struct ib_flow_spec_ipv6 ipv6;
0dbf3332 1894 struct ib_flow_spec_tunnel tunnel;
460d0198 1895 struct ib_flow_spec_action_tag flow_tag;
483a3966 1896 struct ib_flow_spec_action_drop drop;
319a441d
HHZ
1897};
1898
1899struct ib_flow_attr {
1900 enum ib_flow_attr_type type;
1901 u16 size;
1902 u16 priority;
1903 u32 flags;
1904 u8 num_of_specs;
1905 u8 port;
1906 /* Following are the optional layers according to user request
1907 * struct ib_flow_spec_xxx
1908 * struct ib_flow_spec_yyy
1909 */
1910};
1911
1912struct ib_flow {
1913 struct ib_qp *qp;
1914 struct ib_uobject *uobject;
1915};
1916
4cd7c947 1917struct ib_mad_hdr;
1da177e4
LT
1918struct ib_grh;
1919
1920enum ib_process_mad_flags {
1921 IB_MAD_IGNORE_MKEY = 1,
1922 IB_MAD_IGNORE_BKEY = 2,
1923 IB_MAD_IGNORE_ALL = IB_MAD_IGNORE_MKEY | IB_MAD_IGNORE_BKEY
1924};
1925
1926enum ib_mad_result {
1927 IB_MAD_RESULT_FAILURE = 0, /* (!SUCCESS is the important flag) */
1928 IB_MAD_RESULT_SUCCESS = 1 << 0, /* MAD was successfully processed */
1929 IB_MAD_RESULT_REPLY = 1 << 1, /* Reply packet needs to be sent */
1930 IB_MAD_RESULT_CONSUMED = 1 << 2 /* Packet consumed: stop processing */
1931};
1932
21d6454a 1933struct ib_port_cache {
883c71fe 1934 u64 subnet_prefix;
21d6454a
JW
1935 struct ib_pkey_cache *pkey;
1936 struct ib_gid_table *gid;
1937 u8 lmc;
1938 enum ib_port_state port_state;
1939};
1940
1da177e4
LT
1941struct ib_cache {
1942 rwlock_t lock;
1943 struct ib_event_handler event_handler;
21d6454a 1944 struct ib_port_cache *ports;
1da177e4
LT
1945};
1946
07ebafba
TT
1947struct iw_cm_verbs;
1948
7738613e
IW
1949struct ib_port_immutable {
1950 int pkey_tbl_len;
1951 int gid_tbl_len;
f9b22e35 1952 u32 core_cap_flags;
337877a4 1953 u32 max_mad_size;
7738613e
IW
1954};
1955
2fc77572
VN
1956/* rdma netdev type - specifies protocol type */
1957enum rdma_netdev_t {
f0ad83ac
NV
1958 RDMA_NETDEV_OPA_VNIC,
1959 RDMA_NETDEV_IPOIB,
2fc77572
VN
1960};
1961
1962/**
1963 * struct rdma_netdev - rdma netdev
1964 * For cases where netstack interfacing is required.
1965 */
1966struct rdma_netdev {
1967 void *clnt_priv;
1968 struct ib_device *hca;
1969 u8 port_num;
1970
1971 /* control functions */
1972 void (*set_id)(struct net_device *netdev, int id);
f0ad83ac
NV
1973 /* send packet */
1974 int (*send)(struct net_device *dev, struct sk_buff *skb,
1975 struct ib_ah *address, u32 dqpn);
1976 /* multicast */
1977 int (*attach_mcast)(struct net_device *dev, struct ib_device *hca,
1978 union ib_gid *gid, u16 mlid,
1979 int set_qkey, u32 qkey);
1980 int (*detach_mcast)(struct net_device *dev, struct ib_device *hca,
1981 union ib_gid *gid, u16 mlid);
2fc77572
VN
1982};
1983
d291f1a6
DJ
1984struct ib_port_pkey_list {
1985 /* Lock to hold while modifying the list. */
1986 spinlock_t list_lock;
1987 struct list_head pkey_list;
1988};
1989
1da177e4 1990struct ib_device {
0957c29f
BVA
1991 /* Do not access @dma_device directly from ULP nor from HW drivers. */
1992 struct device *dma_device;
1993
1da177e4
LT
1994 char name[IB_DEVICE_NAME_MAX];
1995
1996 struct list_head event_handler_list;
1997 spinlock_t event_handler_lock;
1998
17a55f79 1999 spinlock_t client_data_lock;
1da177e4 2000 struct list_head core_list;
7c1eb45a
HE
2001 /* Access to the client_data_list is protected by the client_data_lock
2002 * spinlock and the lists_rwsem read-write semaphore */
1da177e4 2003 struct list_head client_data_list;
1da177e4
LT
2004
2005 struct ib_cache cache;
7738613e
IW
2006 /**
2007 * port_immutable is indexed by port number
2008 */
2009 struct ib_port_immutable *port_immutable;
1da177e4 2010
f4fd0b22
MT
2011 int num_comp_vectors;
2012
d291f1a6
DJ
2013 struct ib_port_pkey_list *port_pkey_list;
2014
07ebafba
TT
2015 struct iw_cm_verbs *iwcm;
2016
b40f4757
CL
2017 /**
2018 * alloc_hw_stats - Allocate a struct rdma_hw_stats and fill in the
2019 * driver initialized data. The struct is kfree()'ed by the sysfs
2020 * core when the device is removed. A lifespan of -1 in the return
2021 * struct tells the core to set a default lifespan.
2022 */
2023 struct rdma_hw_stats *(*alloc_hw_stats)(struct ib_device *device,
2024 u8 port_num);
2025 /**
2026 * get_hw_stats - Fill in the counter value(s) in the stats struct.
2027 * @index - The index in the value array we wish to have updated, or
2028 * num_counters if we want all stats updated
2029 * Return codes -
2030 * < 0 - Error, no counters updated
2031 * index - Updated the single counter pointed to by index
2032 * num_counters - Updated all counters (will reset the timestamp
2033 * and prevent further calls for lifespan milliseconds)
2034 * Drivers are allowed to update all counters in leiu of just the
2035 * one given in index at their option
2036 */
2037 int (*get_hw_stats)(struct ib_device *device,
2038 struct rdma_hw_stats *stats,
2039 u8 port, int index);
1da177e4 2040 int (*query_device)(struct ib_device *device,
2528e33e
MB
2041 struct ib_device_attr *device_attr,
2042 struct ib_udata *udata);
1da177e4
LT
2043 int (*query_port)(struct ib_device *device,
2044 u8 port_num,
2045 struct ib_port_attr *port_attr);
a3f5adaf
EC
2046 enum rdma_link_layer (*get_link_layer)(struct ib_device *device,
2047 u8 port_num);
03db3a2d
MB
2048 /* When calling get_netdev, the HW vendor's driver should return the
2049 * net device of device @device at port @port_num or NULL if such
2050 * a net device doesn't exist. The vendor driver should call dev_hold
2051 * on this net device. The HW vendor's device driver must guarantee
2052 * that this function returns NULL before the net device reaches
2053 * NETDEV_UNREGISTER_FINAL state.
2054 */
2055 struct net_device *(*get_netdev)(struct ib_device *device,
2056 u8 port_num);
1da177e4
LT
2057 int (*query_gid)(struct ib_device *device,
2058 u8 port_num, int index,
2059 union ib_gid *gid);
03db3a2d
MB
2060 /* When calling add_gid, the HW vendor's driver should
2061 * add the gid of device @device at gid index @index of
2062 * port @port_num to be @gid. Meta-info of that gid (for example,
2063 * the network device related to this gid is available
2064 * at @attr. @context allows the HW vendor driver to store extra
2065 * information together with a GID entry. The HW vendor may allocate
2066 * memory to contain this information and store it in @context when a
2067 * new GID entry is written to. Params are consistent until the next
2068 * call of add_gid or delete_gid. The function should return 0 on
2069 * success or error otherwise. The function could be called
2070 * concurrently for different ports. This function is only called
2071 * when roce_gid_table is used.
2072 */
2073 int (*add_gid)(struct ib_device *device,
2074 u8 port_num,
2075 unsigned int index,
2076 const union ib_gid *gid,
2077 const struct ib_gid_attr *attr,
2078 void **context);
2079 /* When calling del_gid, the HW vendor's driver should delete the
2080 * gid of device @device at gid index @index of port @port_num.
2081 * Upon the deletion of a GID entry, the HW vendor must free any
2082 * allocated memory. The caller will clear @context afterwards.
2083 * This function is only called when roce_gid_table is used.
2084 */
2085 int (*del_gid)(struct ib_device *device,
2086 u8 port_num,
2087 unsigned int index,
2088 void **context);
1da177e4
LT
2089 int (*query_pkey)(struct ib_device *device,
2090 u8 port_num, u16 index, u16 *pkey);
2091 int (*modify_device)(struct ib_device *device,
2092 int device_modify_mask,
2093 struct ib_device_modify *device_modify);
2094 int (*modify_port)(struct ib_device *device,
2095 u8 port_num, int port_modify_mask,
2096 struct ib_port_modify *port_modify);
e2773c06
RD
2097 struct ib_ucontext * (*alloc_ucontext)(struct ib_device *device,
2098 struct ib_udata *udata);
2099 int (*dealloc_ucontext)(struct ib_ucontext *context);
2100 int (*mmap)(struct ib_ucontext *context,
2101 struct vm_area_struct *vma);
2102 struct ib_pd * (*alloc_pd)(struct ib_device *device,
2103 struct ib_ucontext *context,
2104 struct ib_udata *udata);
1da177e4
LT
2105 int (*dealloc_pd)(struct ib_pd *pd);
2106 struct ib_ah * (*create_ah)(struct ib_pd *pd,
90898850 2107 struct rdma_ah_attr *ah_attr,
477864c8 2108 struct ib_udata *udata);
1da177e4 2109 int (*modify_ah)(struct ib_ah *ah,
90898850 2110 struct rdma_ah_attr *ah_attr);
1da177e4 2111 int (*query_ah)(struct ib_ah *ah,
90898850 2112 struct rdma_ah_attr *ah_attr);
1da177e4 2113 int (*destroy_ah)(struct ib_ah *ah);
d41fcc67
RD
2114 struct ib_srq * (*create_srq)(struct ib_pd *pd,
2115 struct ib_srq_init_attr *srq_init_attr,
2116 struct ib_udata *udata);
2117 int (*modify_srq)(struct ib_srq *srq,
2118 struct ib_srq_attr *srq_attr,
9bc57e2d
RC
2119 enum ib_srq_attr_mask srq_attr_mask,
2120 struct ib_udata *udata);
d41fcc67
RD
2121 int (*query_srq)(struct ib_srq *srq,
2122 struct ib_srq_attr *srq_attr);
2123 int (*destroy_srq)(struct ib_srq *srq);
2124 int (*post_srq_recv)(struct ib_srq *srq,
2125 struct ib_recv_wr *recv_wr,
2126 struct ib_recv_wr **bad_recv_wr);
1da177e4 2127 struct ib_qp * (*create_qp)(struct ib_pd *pd,
e2773c06
RD
2128 struct ib_qp_init_attr *qp_init_attr,
2129 struct ib_udata *udata);
1da177e4
LT
2130 int (*modify_qp)(struct ib_qp *qp,
2131 struct ib_qp_attr *qp_attr,
9bc57e2d
RC
2132 int qp_attr_mask,
2133 struct ib_udata *udata);
1da177e4
LT
2134 int (*query_qp)(struct ib_qp *qp,
2135 struct ib_qp_attr *qp_attr,
2136 int qp_attr_mask,
2137 struct ib_qp_init_attr *qp_init_attr);
2138 int (*destroy_qp)(struct ib_qp *qp);
2139 int (*post_send)(struct ib_qp *qp,
2140 struct ib_send_wr *send_wr,
2141 struct ib_send_wr **bad_send_wr);
2142 int (*post_recv)(struct ib_qp *qp,
2143 struct ib_recv_wr *recv_wr,
2144 struct ib_recv_wr **bad_recv_wr);
bcf4c1ea
MB
2145 struct ib_cq * (*create_cq)(struct ib_device *device,
2146 const struct ib_cq_init_attr *attr,
e2773c06
RD
2147 struct ib_ucontext *context,
2148 struct ib_udata *udata);
2dd57162
EC
2149 int (*modify_cq)(struct ib_cq *cq, u16 cq_count,
2150 u16 cq_period);
1da177e4 2151 int (*destroy_cq)(struct ib_cq *cq);
33b9b3ee
RD
2152 int (*resize_cq)(struct ib_cq *cq, int cqe,
2153 struct ib_udata *udata);
1da177e4
LT
2154 int (*poll_cq)(struct ib_cq *cq, int num_entries,
2155 struct ib_wc *wc);
2156 int (*peek_cq)(struct ib_cq *cq, int wc_cnt);
2157 int (*req_notify_cq)(struct ib_cq *cq,
ed23a727 2158 enum ib_cq_notify_flags flags);
1da177e4
LT
2159 int (*req_ncomp_notif)(struct ib_cq *cq,
2160 int wc_cnt);
2161 struct ib_mr * (*get_dma_mr)(struct ib_pd *pd,
2162 int mr_access_flags);
e2773c06 2163 struct ib_mr * (*reg_user_mr)(struct ib_pd *pd,
f7c6a7b5
RD
2164 u64 start, u64 length,
2165 u64 virt_addr,
e2773c06
RD
2166 int mr_access_flags,
2167 struct ib_udata *udata);
7e6edb9b
MB
2168 int (*rereg_user_mr)(struct ib_mr *mr,
2169 int flags,
2170 u64 start, u64 length,
2171 u64 virt_addr,
2172 int mr_access_flags,
2173 struct ib_pd *pd,
2174 struct ib_udata *udata);
1da177e4 2175 int (*dereg_mr)(struct ib_mr *mr);
9bee178b
SG
2176 struct ib_mr * (*alloc_mr)(struct ib_pd *pd,
2177 enum ib_mr_type mr_type,
2178 u32 max_num_sg);
4c67e2bf
SG
2179 int (*map_mr_sg)(struct ib_mr *mr,
2180 struct scatterlist *sg,
ff2ba993 2181 int sg_nents,
9aa8b321 2182 unsigned int *sg_offset);
7083e42e 2183 struct ib_mw * (*alloc_mw)(struct ib_pd *pd,
b2a239df
MB
2184 enum ib_mw_type type,
2185 struct ib_udata *udata);
1da177e4
LT
2186 int (*dealloc_mw)(struct ib_mw *mw);
2187 struct ib_fmr * (*alloc_fmr)(struct ib_pd *pd,
2188 int mr_access_flags,
2189 struct ib_fmr_attr *fmr_attr);
2190 int (*map_phys_fmr)(struct ib_fmr *fmr,
2191 u64 *page_list, int list_len,
2192 u64 iova);
2193 int (*unmap_fmr)(struct list_head *fmr_list);
2194 int (*dealloc_fmr)(struct ib_fmr *fmr);
2195 int (*attach_mcast)(struct ib_qp *qp,
2196 union ib_gid *gid,
2197 u16 lid);
2198 int (*detach_mcast)(struct ib_qp *qp,
2199 union ib_gid *gid,
2200 u16 lid);
2201 int (*process_mad)(struct ib_device *device,
2202 int process_mad_flags,
2203 u8 port_num,
a97e2d86
IW
2204 const struct ib_wc *in_wc,
2205 const struct ib_grh *in_grh,
4cd7c947
IW
2206 const struct ib_mad_hdr *in_mad,
2207 size_t in_mad_size,
2208 struct ib_mad_hdr *out_mad,
2209 size_t *out_mad_size,
2210 u16 *out_mad_pkey_index);
59991f94
SH
2211 struct ib_xrcd * (*alloc_xrcd)(struct ib_device *device,
2212 struct ib_ucontext *ucontext,
2213 struct ib_udata *udata);
2214 int (*dealloc_xrcd)(struct ib_xrcd *xrcd);
319a441d
HHZ
2215 struct ib_flow * (*create_flow)(struct ib_qp *qp,
2216 struct ib_flow_attr
2217 *flow_attr,
2218 int domain);
2219 int (*destroy_flow)(struct ib_flow *flow_id);
1b01d335
SG
2220 int (*check_mr_status)(struct ib_mr *mr, u32 check_mask,
2221 struct ib_mr_status *mr_status);
036b1063 2222 void (*disassociate_ucontext)(struct ib_ucontext *ibcontext);
765d6774
SW
2223 void (*drain_rq)(struct ib_qp *qp);
2224 void (*drain_sq)(struct ib_qp *qp);
50174a7f
EC
2225 int (*set_vf_link_state)(struct ib_device *device, int vf, u8 port,
2226 int state);
2227 int (*get_vf_config)(struct ib_device *device, int vf, u8 port,
2228 struct ifla_vf_info *ivf);
2229 int (*get_vf_stats)(struct ib_device *device, int vf, u8 port,
2230 struct ifla_vf_stats *stats);
2231 int (*set_vf_guid)(struct ib_device *device, int vf, u8 port, u64 guid,
2232 int type);
5fd251c8
YH
2233 struct ib_wq * (*create_wq)(struct ib_pd *pd,
2234 struct ib_wq_init_attr *init_attr,
2235 struct ib_udata *udata);
2236 int (*destroy_wq)(struct ib_wq *wq);
2237 int (*modify_wq)(struct ib_wq *wq,
2238 struct ib_wq_attr *attr,
2239 u32 wq_attr_mask,
2240 struct ib_udata *udata);
6d39786b
YH
2241 struct ib_rwq_ind_table * (*create_rwq_ind_table)(struct ib_device *device,
2242 struct ib_rwq_ind_table_init_attr *init_attr,
2243 struct ib_udata *udata);
2244 int (*destroy_rwq_ind_table)(struct ib_rwq_ind_table *wq_ind_table);
2fc77572
VN
2245 /**
2246 * rdma netdev operations
2247 *
2248 * Driver implementing alloc_rdma_netdev must return -EOPNOTSUPP if it
2249 * doesn't support the specified rdma netdev type.
2250 */
2251 struct net_device *(*alloc_rdma_netdev)(
2252 struct ib_device *device,
2253 u8 port_num,
2254 enum rdma_netdev_t type,
2255 const char *name,
2256 unsigned char name_assign_type,
2257 void (*setup)(struct net_device *));
2258 void (*free_rdma_netdev)(struct net_device *netdev);
9b513090 2259
e2773c06 2260 struct module *owner;
f4e91eb4 2261 struct device dev;
35be0681 2262 struct kobject *ports_parent;
1da177e4
LT
2263 struct list_head port_list;
2264
2265 enum {
2266 IB_DEV_UNINITIALIZED,
2267 IB_DEV_REGISTERED,
2268 IB_DEV_UNREGISTERED
2269 } reg_state;
2270
274c0891 2271 int uverbs_abi_ver;
17a55f79 2272 u64 uverbs_cmd_mask;
f21519b2 2273 u64 uverbs_ex_cmd_mask;
274c0891 2274
bd99fdea 2275 char node_desc[IB_DEVICE_NODE_DESC_MAX];
cf311cd4 2276 __be64 node_guid;
96f15c03 2277 u32 local_dma_lkey;
4139032b 2278 u16 is_switch:1;
1da177e4
LT
2279 u8 node_type;
2280 u8 phys_port_cnt;
3e153a93 2281 struct ib_device_attr attrs;
b40f4757
CL
2282 struct attribute_group *hw_stats_ag;
2283 struct rdma_hw_stats *hw_stats;
7738613e 2284
43579b5f
PP
2285#ifdef CONFIG_CGROUP_RDMA
2286 struct rdmacg_device cg_device;
2287#endif
2288
7738613e
IW
2289 /**
2290 * The following mandatory functions are used only at device
2291 * registration. Keep functions such as these at the end of this
2292 * structure to avoid cache line misses when accessing struct ib_device
2293 * in fast paths.
2294 */
2295 int (*get_port_immutable)(struct ib_device *, u8, struct ib_port_immutable *);
5fa76c20 2296 void (*get_dev_fw_str)(struct ib_device *, char *str, size_t str_len);
1da177e4
LT
2297};
2298
2299struct ib_client {
2300 char *name;
2301 void (*add) (struct ib_device *);
7c1eb45a 2302 void (*remove)(struct ib_device *, void *client_data);
1da177e4 2303
9268f72d
YK
2304 /* Returns the net_dev belonging to this ib_client and matching the
2305 * given parameters.
2306 * @dev: An RDMA device that the net_dev use for communication.
2307 * @port: A physical port number on the RDMA device.
2308 * @pkey: P_Key that the net_dev uses if applicable.
2309 * @gid: A GID that the net_dev uses to communicate.
2310 * @addr: An IP address the net_dev is configured with.
2311 * @client_data: The device's client data set by ib_set_client_data().
2312 *
2313 * An ib_client that implements a net_dev on top of RDMA devices
2314 * (such as IP over IB) should implement this callback, allowing the
2315 * rdma_cm module to find the right net_dev for a given request.
2316 *
2317 * The caller is responsible for calling dev_put on the returned
2318 * netdev. */
2319 struct net_device *(*get_net_dev_by_params)(
2320 struct ib_device *dev,
2321 u8 port,
2322 u16 pkey,
2323 const union ib_gid *gid,
2324 const struct sockaddr *addr,
2325 void *client_data);
1da177e4
LT
2326 struct list_head list;
2327};
2328
2329struct ib_device *ib_alloc_device(size_t size);
2330void ib_dealloc_device(struct ib_device *device);
2331
5fa76c20
IW
2332void ib_get_device_fw_str(struct ib_device *device, char *str, size_t str_len);
2333
9a6edb60
RC
2334int ib_register_device(struct ib_device *device,
2335 int (*port_callback)(struct ib_device *,
2336 u8, struct kobject *));
1da177e4
LT
2337void ib_unregister_device(struct ib_device *device);
2338
2339int ib_register_client (struct ib_client *client);
2340void ib_unregister_client(struct ib_client *client);
2341
2342void *ib_get_client_data(struct ib_device *device, struct ib_client *client);
2343void ib_set_client_data(struct ib_device *device, struct ib_client *client,
2344 void *data);
2345
e2773c06
RD
2346static inline int ib_copy_from_udata(void *dest, struct ib_udata *udata, size_t len)
2347{
2348 return copy_from_user(dest, udata->inbuf, len) ? -EFAULT : 0;
2349}
2350
2351static inline int ib_copy_to_udata(struct ib_udata *udata, void *src, size_t len)
2352{
43c61165 2353 return copy_to_user(udata->outbuf, src, len) ? -EFAULT : 0;
e2773c06
RD
2354}
2355
301a721e
MB
2356static inline bool ib_is_udata_cleared(struct ib_udata *udata,
2357 size_t offset,
2358 size_t len)
2359{
2360 const void __user *p = udata->inbuf + offset;
92d27ae6 2361 bool ret;
301a721e
MB
2362 u8 *buf;
2363
2364 if (len > USHRT_MAX)
2365 return false;
2366
92d27ae6
ME
2367 buf = memdup_user(p, len);
2368 if (IS_ERR(buf))
301a721e
MB
2369 return false;
2370
301a721e 2371 ret = !memchr_inv(buf, 0, len);
301a721e
MB
2372 kfree(buf);
2373 return ret;
2374}
2375
8a51866f
RD
2376/**
2377 * ib_modify_qp_is_ok - Check that the supplied attribute mask
2378 * contains all required attributes and no attributes not allowed for
2379 * the given QP state transition.
2380 * @cur_state: Current QP state
2381 * @next_state: Next QP state
2382 * @type: QP type
2383 * @mask: Mask of supplied QP attributes
dd5f03be 2384 * @ll : link layer of port
8a51866f
RD
2385 *
2386 * This function is a helper function that a low-level driver's
2387 * modify_qp method can use to validate the consumer's input. It
2388 * checks that cur_state and next_state are valid QP states, that a
2389 * transition from cur_state to next_state is allowed by the IB spec,
2390 * and that the attribute mask supplied is allowed for the transition.
2391 */
2392int ib_modify_qp_is_ok(enum ib_qp_state cur_state, enum ib_qp_state next_state,
dd5f03be
MB
2393 enum ib_qp_type type, enum ib_qp_attr_mask mask,
2394 enum rdma_link_layer ll);
8a51866f 2395
1da177e4
LT
2396int ib_register_event_handler (struct ib_event_handler *event_handler);
2397int ib_unregister_event_handler(struct ib_event_handler *event_handler);
2398void ib_dispatch_event(struct ib_event *event);
2399
1da177e4
LT
2400int ib_query_port(struct ib_device *device,
2401 u8 port_num, struct ib_port_attr *port_attr);
2402
a3f5adaf
EC
2403enum rdma_link_layer rdma_port_get_link_layer(struct ib_device *device,
2404 u8 port_num);
2405
4139032b
HR
2406/**
2407 * rdma_cap_ib_switch - Check if the device is IB switch
2408 * @device: Device to check
2409 *
2410 * Device driver is responsible for setting is_switch bit on
2411 * in ib_device structure at init time.
2412 *
2413 * Return: true if the device is IB switch.
2414 */
2415static inline bool rdma_cap_ib_switch(const struct ib_device *device)
2416{
2417 return device->is_switch;
2418}
2419
0cf18d77
IW
2420/**
2421 * rdma_start_port - Return the first valid port number for the device
2422 * specified
2423 *
2424 * @device: Device to be checked
2425 *
2426 * Return start port number
2427 */
2428static inline u8 rdma_start_port(const struct ib_device *device)
2429{
4139032b 2430 return rdma_cap_ib_switch(device) ? 0 : 1;
0cf18d77
IW
2431}
2432
2433/**
2434 * rdma_end_port - Return the last valid port number for the device
2435 * specified
2436 *
2437 * @device: Device to be checked
2438 *
2439 * Return last port number
2440 */
2441static inline u8 rdma_end_port(const struct ib_device *device)
2442{
4139032b 2443 return rdma_cap_ib_switch(device) ? 0 : device->phys_port_cnt;
0cf18d77
IW
2444}
2445
24dc831b
YS
2446static inline int rdma_is_port_valid(const struct ib_device *device,
2447 unsigned int port)
2448{
2449 return (port >= rdma_start_port(device) &&
2450 port <= rdma_end_port(device));
2451}
2452
5ede9289 2453static inline bool rdma_protocol_ib(const struct ib_device *device, u8 port_num)
de66be94 2454{
f9b22e35 2455 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_IB;
de66be94
MW
2456}
2457
5ede9289 2458static inline bool rdma_protocol_roce(const struct ib_device *device, u8 port_num)
7766a99f
MB
2459{
2460 return device->port_immutable[port_num].core_cap_flags &
2461 (RDMA_CORE_CAP_PROT_ROCE | RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP);
2462}
2463
2464static inline bool rdma_protocol_roce_udp_encap(const struct ib_device *device, u8 port_num)
2465{
2466 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_ROCE_UDP_ENCAP;
2467}
2468
2469static inline bool rdma_protocol_roce_eth_encap(const struct ib_device *device, u8 port_num)
de66be94 2470{
f9b22e35 2471 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_ROCE;
de66be94
MW
2472}
2473
5ede9289 2474static inline bool rdma_protocol_iwarp(const struct ib_device *device, u8 port_num)
de66be94 2475{
f9b22e35 2476 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_IWARP;
de66be94
MW
2477}
2478
5ede9289 2479static inline bool rdma_ib_or_roce(const struct ib_device *device, u8 port_num)
de66be94 2480{
7766a99f
MB
2481 return rdma_protocol_ib(device, port_num) ||
2482 rdma_protocol_roce(device, port_num);
de66be94
MW
2483}
2484
aa773bd4
OG
2485static inline bool rdma_protocol_raw_packet(const struct ib_device *device, u8 port_num)
2486{
2487 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_RAW_PACKET;
2488}
2489
ce1e055f
OG
2490static inline bool rdma_protocol_usnic(const struct ib_device *device, u8 port_num)
2491{
2492 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_PROT_USNIC;
2493}
2494
c757dea8 2495/**
296ec009 2496 * rdma_cap_ib_mad - Check if the port of a device supports Infiniband
c757dea8 2497 * Management Datagrams.
296ec009
MW
2498 * @device: Device to check
2499 * @port_num: Port number to check
c757dea8 2500 *
296ec009
MW
2501 * Management Datagrams (MAD) are a required part of the InfiniBand
2502 * specification and are supported on all InfiniBand devices. A slightly
2503 * extended version are also supported on OPA interfaces.
c757dea8 2504 *
296ec009 2505 * Return: true if the port supports sending/receiving of MAD packets.
c757dea8 2506 */
5ede9289 2507static inline bool rdma_cap_ib_mad(const struct ib_device *device, u8 port_num)
c757dea8 2508{
f9b22e35 2509 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IB_MAD;
c757dea8
MW
2510}
2511
65995fee
IW
2512/**
2513 * rdma_cap_opa_mad - Check if the port of device provides support for OPA
2514 * Management Datagrams.
2515 * @device: Device to check
2516 * @port_num: Port number to check
2517 *
2518 * Intel OmniPath devices extend and/or replace the InfiniBand Management
2519 * datagrams with their own versions. These OPA MADs share many but not all of
2520 * the characteristics of InfiniBand MADs.
2521 *
2522 * OPA MADs differ in the following ways:
2523 *
2524 * 1) MADs are variable size up to 2K
2525 * IBTA defined MADs remain fixed at 256 bytes
2526 * 2) OPA SMPs must carry valid PKeys
2527 * 3) OPA SMP packets are a different format
2528 *
2529 * Return: true if the port supports OPA MAD packet formats.
2530 */
2531static inline bool rdma_cap_opa_mad(struct ib_device *device, u8 port_num)
2532{
2533 return (device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_OPA_MAD)
2534 == RDMA_CORE_CAP_OPA_MAD;
2535}
2536
29541e3a 2537/**
296ec009
MW
2538 * rdma_cap_ib_smi - Check if the port of a device provides an Infiniband
2539 * Subnet Management Agent (SMA) on the Subnet Management Interface (SMI).
2540 * @device: Device to check
2541 * @port_num: Port number to check
29541e3a 2542 *
296ec009
MW
2543 * Each InfiniBand node is required to provide a Subnet Management Agent
2544 * that the subnet manager can access. Prior to the fabric being fully
2545 * configured by the subnet manager, the SMA is accessed via a well known
2546 * interface called the Subnet Management Interface (SMI). This interface
2547 * uses directed route packets to communicate with the SM to get around the
2548 * chicken and egg problem of the SM needing to know what's on the fabric
2549 * in order to configure the fabric, and needing to configure the fabric in
2550 * order to send packets to the devices on the fabric. These directed
2551 * route packets do not need the fabric fully configured in order to reach
2552 * their destination. The SMI is the only method allowed to send
2553 * directed route packets on an InfiniBand fabric.
29541e3a 2554 *
296ec009 2555 * Return: true if the port provides an SMI.
29541e3a 2556 */
5ede9289 2557static inline bool rdma_cap_ib_smi(const struct ib_device *device, u8 port_num)
29541e3a 2558{
f9b22e35 2559 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IB_SMI;
29541e3a
MW
2560}
2561
72219cea
MW
2562/**
2563 * rdma_cap_ib_cm - Check if the port of device has the capability Infiniband
2564 * Communication Manager.
296ec009
MW
2565 * @device: Device to check
2566 * @port_num: Port number to check
72219cea 2567 *
296ec009
MW
2568 * The InfiniBand Communication Manager is one of many pre-defined General
2569 * Service Agents (GSA) that are accessed via the General Service
2570 * Interface (GSI). It's role is to facilitate establishment of connections
2571 * between nodes as well as other management related tasks for established
2572 * connections.
72219cea 2573 *
296ec009
MW
2574 * Return: true if the port supports an IB CM (this does not guarantee that
2575 * a CM is actually running however).
72219cea 2576 */
5ede9289 2577static inline bool rdma_cap_ib_cm(const struct ib_device *device, u8 port_num)
72219cea 2578{
f9b22e35 2579 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IB_CM;
72219cea
MW
2580}
2581
04215330
MW
2582/**
2583 * rdma_cap_iw_cm - Check if the port of device has the capability IWARP
2584 * Communication Manager.
296ec009
MW
2585 * @device: Device to check
2586 * @port_num: Port number to check
04215330 2587 *
296ec009
MW
2588 * Similar to above, but specific to iWARP connections which have a different
2589 * managment protocol than InfiniBand.
04215330 2590 *
296ec009
MW
2591 * Return: true if the port supports an iWARP CM (this does not guarantee that
2592 * a CM is actually running however).
04215330 2593 */
5ede9289 2594static inline bool rdma_cap_iw_cm(const struct ib_device *device, u8 port_num)
04215330 2595{
f9b22e35 2596 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IW_CM;
04215330
MW
2597}
2598
fe53ba2f
MW
2599/**
2600 * rdma_cap_ib_sa - Check if the port of device has the capability Infiniband
2601 * Subnet Administration.
296ec009
MW
2602 * @device: Device to check
2603 * @port_num: Port number to check
fe53ba2f 2604 *
296ec009
MW
2605 * An InfiniBand Subnet Administration (SA) service is a pre-defined General
2606 * Service Agent (GSA) provided by the Subnet Manager (SM). On InfiniBand
2607 * fabrics, devices should resolve routes to other hosts by contacting the
2608 * SA to query the proper route.
fe53ba2f 2609 *
296ec009
MW
2610 * Return: true if the port should act as a client to the fabric Subnet
2611 * Administration interface. This does not imply that the SA service is
2612 * running locally.
fe53ba2f 2613 */
5ede9289 2614static inline bool rdma_cap_ib_sa(const struct ib_device *device, u8 port_num)
fe53ba2f 2615{
f9b22e35 2616 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_IB_SA;
fe53ba2f
MW
2617}
2618
a31ad3b0
MW
2619/**
2620 * rdma_cap_ib_mcast - Check if the port of device has the capability Infiniband
2621 * Multicast.
296ec009
MW
2622 * @device: Device to check
2623 * @port_num: Port number to check
a31ad3b0 2624 *
296ec009
MW
2625 * InfiniBand multicast registration is more complex than normal IPv4 or
2626 * IPv6 multicast registration. Each Host Channel Adapter must register
2627 * with the Subnet Manager when it wishes to join a multicast group. It
2628 * should do so only once regardless of how many queue pairs it subscribes
2629 * to this group. And it should leave the group only after all queue pairs
2630 * attached to the group have been detached.
a31ad3b0 2631 *
296ec009
MW
2632 * Return: true if the port must undertake the additional adminstrative
2633 * overhead of registering/unregistering with the SM and tracking of the
2634 * total number of queue pairs attached to the multicast group.
a31ad3b0 2635 */
5ede9289 2636static inline bool rdma_cap_ib_mcast(const struct ib_device *device, u8 port_num)
a31ad3b0
MW
2637{
2638 return rdma_cap_ib_sa(device, port_num);
2639}
2640
30a74ef4
MW
2641/**
2642 * rdma_cap_af_ib - Check if the port of device has the capability
2643 * Native Infiniband Address.
296ec009
MW
2644 * @device: Device to check
2645 * @port_num: Port number to check
30a74ef4 2646 *
296ec009
MW
2647 * InfiniBand addressing uses a port's GUID + Subnet Prefix to make a default
2648 * GID. RoCE uses a different mechanism, but still generates a GID via
2649 * a prescribed mechanism and port specific data.
30a74ef4 2650 *
296ec009
MW
2651 * Return: true if the port uses a GID address to identify devices on the
2652 * network.
30a74ef4 2653 */
5ede9289 2654static inline bool rdma_cap_af_ib(const struct ib_device *device, u8 port_num)
30a74ef4 2655{
f9b22e35 2656 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_AF_IB;
30a74ef4
MW
2657}
2658
227128fc
MW
2659/**
2660 * rdma_cap_eth_ah - Check if the port of device has the capability
296ec009
MW
2661 * Ethernet Address Handle.
2662 * @device: Device to check
2663 * @port_num: Port number to check
227128fc 2664 *
296ec009
MW
2665 * RoCE is InfiniBand over Ethernet, and it uses a well defined technique
2666 * to fabricate GIDs over Ethernet/IP specific addresses native to the
2667 * port. Normally, packet headers are generated by the sending host
2668 * adapter, but when sending connectionless datagrams, we must manually
2669 * inject the proper headers for the fabric we are communicating over.
227128fc 2670 *
296ec009
MW
2671 * Return: true if we are running as a RoCE port and must force the
2672 * addition of a Global Route Header built from our Ethernet Address
2673 * Handle into our header list for connectionless packets.
227128fc 2674 */
5ede9289 2675static inline bool rdma_cap_eth_ah(const struct ib_device *device, u8 port_num)
227128fc 2676{
f9b22e35 2677 return device->port_immutable[port_num].core_cap_flags & RDMA_CORE_CAP_ETH_AH;
227128fc
MW
2678}
2679
94d595c5
DC
2680/**
2681 * rdma_cap_opa_ah - Check if the port of device supports
2682 * OPA Address handles
2683 * @device: Device to check
2684 * @port_num: Port number to check
2685 *
2686 * Return: true if we are running on an OPA device which supports
2687 * the extended OPA addressing.
2688 */
2689static inline bool rdma_cap_opa_ah(struct ib_device *device, u8 port_num)
2690{
2691 return (device->port_immutable[port_num].core_cap_flags &
2692 RDMA_CORE_CAP_OPA_AH) == RDMA_CORE_CAP_OPA_AH;
2693}
2694
337877a4
IW
2695/**
2696 * rdma_max_mad_size - Return the max MAD size required by this RDMA Port.
2697 *
2698 * @device: Device
2699 * @port_num: Port number
2700 *
2701 * This MAD size includes the MAD headers and MAD payload. No other headers
2702 * are included.
2703 *
2704 * Return the max MAD size required by the Port. Will return 0 if the port
2705 * does not support MADs
2706 */
2707static inline size_t rdma_max_mad_size(const struct ib_device *device, u8 port_num)
2708{
2709 return device->port_immutable[port_num].max_mad_size;
2710}
2711
03db3a2d
MB
2712/**
2713 * rdma_cap_roce_gid_table - Check if the port of device uses roce_gid_table
2714 * @device: Device to check
2715 * @port_num: Port number to check
2716 *
2717 * RoCE GID table mechanism manages the various GIDs for a device.
2718 *
2719 * NOTE: if allocating the port's GID table has failed, this call will still
2720 * return true, but any RoCE GID table API will fail.
2721 *
2722 * Return: true if the port uses RoCE GID table mechanism in order to manage
2723 * its GIDs.
2724 */
2725static inline bool rdma_cap_roce_gid_table(const struct ib_device *device,
2726 u8 port_num)
2727{
2728 return rdma_protocol_roce(device, port_num) &&
2729 device->add_gid && device->del_gid;
2730}
2731
002516ed
CH
2732/*
2733 * Check if the device supports READ W/ INVALIDATE.
2734 */
2735static inline bool rdma_cap_read_inv(struct ib_device *dev, u32 port_num)
2736{
2737 /*
2738 * iWarp drivers must support READ W/ INVALIDATE. No other protocol
2739 * has support for it yet.
2740 */
2741 return rdma_protocol_iwarp(dev, port_num);
2742}
2743
1da177e4 2744int ib_query_gid(struct ib_device *device,
55ee3ab2
MB
2745 u8 port_num, int index, union ib_gid *gid,
2746 struct ib_gid_attr *attr);
1da177e4 2747
50174a7f
EC
2748int ib_set_vf_link_state(struct ib_device *device, int vf, u8 port,
2749 int state);
2750int ib_get_vf_config(struct ib_device *device, int vf, u8 port,
2751 struct ifla_vf_info *info);
2752int ib_get_vf_stats(struct ib_device *device, int vf, u8 port,
2753 struct ifla_vf_stats *stats);
2754int ib_set_vf_guid(struct ib_device *device, int vf, u8 port, u64 guid,
2755 int type);
2756
1da177e4
LT
2757int ib_query_pkey(struct ib_device *device,
2758 u8 port_num, u16 index, u16 *pkey);
2759
2760int ib_modify_device(struct ib_device *device,
2761 int device_modify_mask,
2762 struct ib_device_modify *device_modify);
2763
2764int ib_modify_port(struct ib_device *device,
2765 u8 port_num, int port_modify_mask,
2766 struct ib_port_modify *port_modify);
2767
5eb620c8 2768int ib_find_gid(struct ib_device *device, union ib_gid *gid,
b39ffa1d
MB
2769 enum ib_gid_type gid_type, struct net_device *ndev,
2770 u8 *port_num, u16 *index);
5eb620c8
YE
2771
2772int ib_find_pkey(struct ib_device *device,
2773 u8 port_num, u16 pkey, u16 *index);
2774
ed082d36
CH
2775enum ib_pd_flags {
2776 /*
2777 * Create a memory registration for all memory in the system and place
2778 * the rkey for it into pd->unsafe_global_rkey. This can be used by
2779 * ULPs to avoid the overhead of dynamic MRs.
2780 *
2781 * This flag is generally considered unsafe and must only be used in
2782 * extremly trusted environments. Every use of it will log a warning
2783 * in the kernel log.
2784 */
2785 IB_PD_UNSAFE_GLOBAL_RKEY = 0x01,
2786};
1da177e4 2787
ed082d36
CH
2788struct ib_pd *__ib_alloc_pd(struct ib_device *device, unsigned int flags,
2789 const char *caller);
2790#define ib_alloc_pd(device, flags) \
2791 __ib_alloc_pd((device), (flags), __func__)
7dd78647 2792void ib_dealloc_pd(struct ib_pd *pd);
1da177e4
LT
2793
2794/**
0a18cfe4 2795 * rdma_create_ah - Creates an address handle for the given address vector.
1da177e4
LT
2796 * @pd: The protection domain associated with the address handle.
2797 * @ah_attr: The attributes of the address vector.
2798 *
2799 * The address handle is used to reference a local or global destination
2800 * in all UD QP post sends.
2801 */
0a18cfe4 2802struct ib_ah *rdma_create_ah(struct ib_pd *pd, struct rdma_ah_attr *ah_attr);
1da177e4 2803
850d8fd7
MS
2804/**
2805 * ib_get_gids_from_rdma_hdr - Get sgid and dgid from GRH or IPv4 header
2806 * work completion.
2807 * @hdr: the L3 header to parse
2808 * @net_type: type of header to parse
2809 * @sgid: place to store source gid
2810 * @dgid: place to store destination gid
2811 */
2812int ib_get_gids_from_rdma_hdr(const union rdma_network_hdr *hdr,
2813 enum rdma_network_type net_type,
2814 union ib_gid *sgid, union ib_gid *dgid);
2815
2816/**
2817 * ib_get_rdma_header_version - Get the header version
2818 * @hdr: the L3 header to parse
2819 */
2820int ib_get_rdma_header_version(const union rdma_network_hdr *hdr);
2821
4e00d694
SH
2822/**
2823 * ib_init_ah_from_wc - Initializes address handle attributes from a
2824 * work completion.
2825 * @device: Device on which the received message arrived.
2826 * @port_num: Port on which the received message arrived.
2827 * @wc: Work completion associated with the received message.
2828 * @grh: References the received global route header. This parameter is
2829 * ignored unless the work completion indicates that the GRH is valid.
2830 * @ah_attr: Returned attributes that can be used when creating an address
2831 * handle for replying to the message.
2832 */
73cdaaee
IW
2833int ib_init_ah_from_wc(struct ib_device *device, u8 port_num,
2834 const struct ib_wc *wc, const struct ib_grh *grh,
90898850 2835 struct rdma_ah_attr *ah_attr);
4e00d694 2836
513789ed
HR
2837/**
2838 * ib_create_ah_from_wc - Creates an address handle associated with the
2839 * sender of the specified work completion.
2840 * @pd: The protection domain associated with the address handle.
2841 * @wc: Work completion information associated with a received message.
2842 * @grh: References the received global route header. This parameter is
2843 * ignored unless the work completion indicates that the GRH is valid.
2844 * @port_num: The outbound port number to associate with the address.
2845 *
2846 * The address handle is used to reference a local or global destination
2847 * in all UD QP post sends.
2848 */
73cdaaee
IW
2849struct ib_ah *ib_create_ah_from_wc(struct ib_pd *pd, const struct ib_wc *wc,
2850 const struct ib_grh *grh, u8 port_num);
513789ed 2851
1da177e4 2852/**
67b985b6 2853 * rdma_modify_ah - Modifies the address vector associated with an address
1da177e4
LT
2854 * handle.
2855 * @ah: The address handle to modify.
2856 * @ah_attr: The new address vector attributes to associate with the
2857 * address handle.
2858 */
67b985b6 2859int rdma_modify_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr);
1da177e4
LT
2860
2861/**
bfbfd661 2862 * rdma_query_ah - Queries the address vector associated with an address
1da177e4
LT
2863 * handle.
2864 * @ah: The address handle to query.
2865 * @ah_attr: The address vector attributes associated with the address
2866 * handle.
2867 */
bfbfd661 2868int rdma_query_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr);
1da177e4
LT
2869
2870/**
36523159 2871 * rdma_destroy_ah - Destroys an address handle.
1da177e4
LT
2872 * @ah: The address handle to destroy.
2873 */
36523159 2874int rdma_destroy_ah(struct ib_ah *ah);
1da177e4 2875
d41fcc67
RD
2876/**
2877 * ib_create_srq - Creates a SRQ associated with the specified protection
2878 * domain.
2879 * @pd: The protection domain associated with the SRQ.
abb6e9ba
DB
2880 * @srq_init_attr: A list of initial attributes required to create the
2881 * SRQ. If SRQ creation succeeds, then the attributes are updated to
2882 * the actual capabilities of the created SRQ.
d41fcc67
RD
2883 *
2884 * srq_attr->max_wr and srq_attr->max_sge are read the determine the
2885 * requested size of the SRQ, and set to the actual values allocated
2886 * on return. If ib_create_srq() succeeds, then max_wr and max_sge
2887 * will always be at least as large as the requested values.
2888 */
2889struct ib_srq *ib_create_srq(struct ib_pd *pd,
2890 struct ib_srq_init_attr *srq_init_attr);
2891
2892/**
2893 * ib_modify_srq - Modifies the attributes for the specified SRQ.
2894 * @srq: The SRQ to modify.
2895 * @srq_attr: On input, specifies the SRQ attributes to modify. On output,
2896 * the current values of selected SRQ attributes are returned.
2897 * @srq_attr_mask: A bit-mask used to specify which attributes of the SRQ
2898 * are being modified.
2899 *
2900 * The mask may contain IB_SRQ_MAX_WR to resize the SRQ and/or
2901 * IB_SRQ_LIMIT to set the SRQ's limit and request notification when
2902 * the number of receives queued drops below the limit.
2903 */
2904int ib_modify_srq(struct ib_srq *srq,
2905 struct ib_srq_attr *srq_attr,
2906 enum ib_srq_attr_mask srq_attr_mask);
2907
2908/**
2909 * ib_query_srq - Returns the attribute list and current values for the
2910 * specified SRQ.
2911 * @srq: The SRQ to query.
2912 * @srq_attr: The attributes of the specified SRQ.
2913 */
2914int ib_query_srq(struct ib_srq *srq,
2915 struct ib_srq_attr *srq_attr);
2916
2917/**
2918 * ib_destroy_srq - Destroys the specified SRQ.
2919 * @srq: The SRQ to destroy.
2920 */
2921int ib_destroy_srq(struct ib_srq *srq);
2922
2923/**
2924 * ib_post_srq_recv - Posts a list of work requests to the specified SRQ.
2925 * @srq: The SRQ to post the work request on.
2926 * @recv_wr: A list of work requests to post on the receive queue.
2927 * @bad_recv_wr: On an immediate failure, this parameter will reference
2928 * the work request that failed to be posted on the QP.
2929 */
2930static inline int ib_post_srq_recv(struct ib_srq *srq,
2931 struct ib_recv_wr *recv_wr,
2932 struct ib_recv_wr **bad_recv_wr)
2933{
2934 return srq->device->post_srq_recv(srq, recv_wr, bad_recv_wr);
2935}
2936
1da177e4
LT
2937/**
2938 * ib_create_qp - Creates a QP associated with the specified protection
2939 * domain.
2940 * @pd: The protection domain associated with the QP.
abb6e9ba
DB
2941 * @qp_init_attr: A list of initial attributes required to create the
2942 * QP. If QP creation succeeds, then the attributes are updated to
2943 * the actual capabilities of the created QP.
1da177e4
LT
2944 */
2945struct ib_qp *ib_create_qp(struct ib_pd *pd,
2946 struct ib_qp_init_attr *qp_init_attr);
2947
2948/**
2949 * ib_modify_qp - Modifies the attributes for the specified QP and then
2950 * transitions the QP to the given state.
2951 * @qp: The QP to modify.
2952 * @qp_attr: On input, specifies the QP attributes to modify. On output,
2953 * the current values of selected QP attributes are returned.
2954 * @qp_attr_mask: A bit-mask used to specify which attributes of the QP
2955 * are being modified.
2956 */
2957int ib_modify_qp(struct ib_qp *qp,
2958 struct ib_qp_attr *qp_attr,
2959 int qp_attr_mask);
2960
2961/**
2962 * ib_query_qp - Returns the attribute list and current values for the
2963 * specified QP.
2964 * @qp: The QP to query.
2965 * @qp_attr: The attributes of the specified QP.
2966 * @qp_attr_mask: A bit-mask used to select specific attributes to query.
2967 * @qp_init_attr: Additional attributes of the selected QP.
2968 *
2969 * The qp_attr_mask may be used to limit the query to gathering only the
2970 * selected attributes.
2971 */
2972int ib_query_qp(struct ib_qp *qp,
2973 struct ib_qp_attr *qp_attr,
2974 int qp_attr_mask,
2975 struct ib_qp_init_attr *qp_init_attr);
2976
2977/**
2978 * ib_destroy_qp - Destroys the specified QP.
2979 * @qp: The QP to destroy.
2980 */
2981int ib_destroy_qp(struct ib_qp *qp);
2982
d3d72d90 2983/**
0e0ec7e0
SH
2984 * ib_open_qp - Obtain a reference to an existing sharable QP.
2985 * @xrcd - XRC domain
2986 * @qp_open_attr: Attributes identifying the QP to open.
2987 *
2988 * Returns a reference to a sharable QP.
2989 */
2990struct ib_qp *ib_open_qp(struct ib_xrcd *xrcd,
2991 struct ib_qp_open_attr *qp_open_attr);
2992
2993/**
2994 * ib_close_qp - Release an external reference to a QP.
d3d72d90
SH
2995 * @qp: The QP handle to release
2996 *
0e0ec7e0
SH
2997 * The opened QP handle is released by the caller. The underlying
2998 * shared QP is not destroyed until all internal references are released.
d3d72d90 2999 */
0e0ec7e0 3000int ib_close_qp(struct ib_qp *qp);
d3d72d90 3001
1da177e4
LT
3002/**
3003 * ib_post_send - Posts a list of work requests to the send queue of
3004 * the specified QP.
3005 * @qp: The QP to post the work request on.
3006 * @send_wr: A list of work requests to post on the send queue.
3007 * @bad_send_wr: On an immediate failure, this parameter will reference
3008 * the work request that failed to be posted on the QP.
55464d46
BVA
3009 *
3010 * While IBA Vol. 1 section 11.4.1.1 specifies that if an immediate
3011 * error is returned, the QP state shall not be affected,
3012 * ib_post_send() will return an immediate error after queueing any
3013 * earlier work requests in the list.
1da177e4
LT
3014 */
3015static inline int ib_post_send(struct ib_qp *qp,
3016 struct ib_send_wr *send_wr,
3017 struct ib_send_wr **bad_send_wr)
3018{
3019 return qp->device->post_send(qp, send_wr, bad_send_wr);
3020}
3021
3022/**
3023 * ib_post_recv - Posts a list of work requests to the receive queue of
3024 * the specified QP.
3025 * @qp: The QP to post the work request on.
3026 * @recv_wr: A list of work requests to post on the receive queue.
3027 * @bad_recv_wr: On an immediate failure, this parameter will reference
3028 * the work request that failed to be posted on the QP.
3029 */
3030static inline int ib_post_recv(struct ib_qp *qp,
3031 struct ib_recv_wr *recv_wr,
3032 struct ib_recv_wr **bad_recv_wr)
3033{
3034 return qp->device->post_recv(qp, recv_wr, bad_recv_wr);
3035}
3036
14d3a3b2
CH
3037struct ib_cq *ib_alloc_cq(struct ib_device *dev, void *private,
3038 int nr_cqe, int comp_vector, enum ib_poll_context poll_ctx);
3039void ib_free_cq(struct ib_cq *cq);
3040int ib_process_cq_direct(struct ib_cq *cq, int budget);
3041
1da177e4
LT
3042/**
3043 * ib_create_cq - Creates a CQ on the specified device.
3044 * @device: The device on which to create the CQ.
3045 * @comp_handler: A user-specified callback that is invoked when a
3046 * completion event occurs on the CQ.
3047 * @event_handler: A user-specified callback that is invoked when an
3048 * asynchronous event not associated with a completion occurs on the CQ.
3049 * @cq_context: Context associated with the CQ returned to the user via
3050 * the associated completion and event handlers.
8e37210b 3051 * @cq_attr: The attributes the CQ should be created upon.
1da177e4
LT
3052 *
3053 * Users can examine the cq structure to determine the actual CQ size.
3054 */
3055struct ib_cq *ib_create_cq(struct ib_device *device,
3056 ib_comp_handler comp_handler,
3057 void (*event_handler)(struct ib_event *, void *),
8e37210b
MB
3058 void *cq_context,
3059 const struct ib_cq_init_attr *cq_attr);
1da177e4
LT
3060
3061/**
3062 * ib_resize_cq - Modifies the capacity of the CQ.
3063 * @cq: The CQ to resize.
3064 * @cqe: The minimum size of the CQ.
3065 *
3066 * Users can examine the cq structure to determine the actual CQ size.
3067 */
3068int ib_resize_cq(struct ib_cq *cq, int cqe);
3069
2dd57162
EC
3070/**
3071 * ib_modify_cq - Modifies moderation params of the CQ
3072 * @cq: The CQ to modify.
3073 * @cq_count: number of CQEs that will trigger an event
3074 * @cq_period: max period of time in usec before triggering an event
3075 *
3076 */
3077int ib_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period);
3078
1da177e4
LT
3079/**
3080 * ib_destroy_cq - Destroys the specified CQ.
3081 * @cq: The CQ to destroy.
3082 */
3083int ib_destroy_cq(struct ib_cq *cq);
3084
3085/**
3086 * ib_poll_cq - poll a CQ for completion(s)
3087 * @cq:the CQ being polled
3088 * @num_entries:maximum number of completions to return
3089 * @wc:array of at least @num_entries &struct ib_wc where completions
3090 * will be returned
3091 *
3092 * Poll a CQ for (possibly multiple) completions. If the return value
3093 * is < 0, an error occurred. If the return value is >= 0, it is the
3094 * number of completions returned. If the return value is
3095 * non-negative and < num_entries, then the CQ was emptied.
3096 */
3097static inline int ib_poll_cq(struct ib_cq *cq, int num_entries,
3098 struct ib_wc *wc)
3099{
3100 return cq->device->poll_cq(cq, num_entries, wc);
3101}
3102
3103/**
3104 * ib_peek_cq - Returns the number of unreaped completions currently
3105 * on the specified CQ.
3106 * @cq: The CQ to peek.
3107 * @wc_cnt: A minimum number of unreaped completions to check for.
3108 *
3109 * If the number of unreaped completions is greater than or equal to wc_cnt,
3110 * this function returns wc_cnt, otherwise, it returns the actual number of
3111 * unreaped completions.
3112 */
3113int ib_peek_cq(struct ib_cq *cq, int wc_cnt);
3114
3115/**
3116 * ib_req_notify_cq - Request completion notification on a CQ.
3117 * @cq: The CQ to generate an event for.
ed23a727
RD
3118 * @flags:
3119 * Must contain exactly one of %IB_CQ_SOLICITED or %IB_CQ_NEXT_COMP
3120 * to request an event on the next solicited event or next work
3121 * completion at any type, respectively. %IB_CQ_REPORT_MISSED_EVENTS
3122 * may also be |ed in to request a hint about missed events, as
3123 * described below.
3124 *
3125 * Return Value:
3126 * < 0 means an error occurred while requesting notification
3127 * == 0 means notification was requested successfully, and if
3128 * IB_CQ_REPORT_MISSED_EVENTS was passed in, then no events
3129 * were missed and it is safe to wait for another event. In
3130 * this case is it guaranteed that any work completions added
3131 * to the CQ since the last CQ poll will trigger a completion
3132 * notification event.
3133 * > 0 is only returned if IB_CQ_REPORT_MISSED_EVENTS was passed
3134 * in. It means that the consumer must poll the CQ again to
3135 * make sure it is empty to avoid missing an event because of a
3136 * race between requesting notification and an entry being
3137 * added to the CQ. This return value means it is possible
3138 * (but not guaranteed) that a work completion has been added
3139 * to the CQ since the last poll without triggering a
3140 * completion notification event.
1da177e4
LT
3141 */
3142static inline int ib_req_notify_cq(struct ib_cq *cq,
ed23a727 3143 enum ib_cq_notify_flags flags)
1da177e4 3144{
ed23a727 3145 return cq->device->req_notify_cq(cq, flags);
1da177e4
LT
3146}
3147
3148/**
3149 * ib_req_ncomp_notif - Request completion notification when there are
3150 * at least the specified number of unreaped completions on the CQ.
3151 * @cq: The CQ to generate an event for.
3152 * @wc_cnt: The number of unreaped completions that should be on the
3153 * CQ before an event is generated.
3154 */
3155static inline int ib_req_ncomp_notif(struct ib_cq *cq, int wc_cnt)
3156{
3157 return cq->device->req_ncomp_notif ?
3158 cq->device->req_ncomp_notif(cq, wc_cnt) :
3159 -ENOSYS;
3160}
3161
9b513090
RC
3162/**
3163 * ib_dma_mapping_error - check a DMA addr for error
3164 * @dev: The device for which the dma_addr was created
3165 * @dma_addr: The DMA address to check
3166 */
3167static inline int ib_dma_mapping_error(struct ib_device *dev, u64 dma_addr)
3168{
0957c29f 3169 return dma_mapping_error(dev->dma_device, dma_addr);
9b513090
RC
3170}
3171
3172/**
3173 * ib_dma_map_single - Map a kernel virtual address to DMA address
3174 * @dev: The device for which the dma_addr is to be created
3175 * @cpu_addr: The kernel virtual address
3176 * @size: The size of the region in bytes
3177 * @direction: The direction of the DMA
3178 */
3179static inline u64 ib_dma_map_single(struct ib_device *dev,
3180 void *cpu_addr, size_t size,
3181 enum dma_data_direction direction)
3182{
0957c29f 3183 return dma_map_single(dev->dma_device, cpu_addr, size, direction);
9b513090
RC
3184}
3185
3186/**
3187 * ib_dma_unmap_single - Destroy a mapping created by ib_dma_map_single()
3188 * @dev: The device for which the DMA address was created
3189 * @addr: The DMA address
3190 * @size: The size of the region in bytes
3191 * @direction: The direction of the DMA
3192 */
3193static inline void ib_dma_unmap_single(struct ib_device *dev,
3194 u64 addr, size_t size,
3195 enum dma_data_direction direction)
3196{
0957c29f 3197 dma_unmap_single(dev->dma_device, addr, size, direction);
cb9fbc5c
AK
3198}
3199
9b513090
RC
3200/**
3201 * ib_dma_map_page - Map a physical page to DMA address
3202 * @dev: The device for which the dma_addr is to be created
3203 * @page: The page to be mapped
3204 * @offset: The offset within the page
3205 * @size: The size of the region in bytes
3206 * @direction: The direction of the DMA
3207 */
3208static inline u64 ib_dma_map_page(struct ib_device *dev,
3209 struct page *page,
3210 unsigned long offset,
3211 size_t size,
3212 enum dma_data_direction direction)
3213{
0957c29f 3214 return dma_map_page(dev->dma_device, page, offset, size, direction);
9b513090
RC
3215}
3216
3217/**
3218 * ib_dma_unmap_page - Destroy a mapping created by ib_dma_map_page()
3219 * @dev: The device for which the DMA address was created
3220 * @addr: The DMA address
3221 * @size: The size of the region in bytes
3222 * @direction: The direction of the DMA
3223 */
3224static inline void ib_dma_unmap_page(struct ib_device *dev,
3225 u64 addr, size_t size,
3226 enum dma_data_direction direction)
3227{
0957c29f 3228 dma_unmap_page(dev->dma_device, addr, size, direction);
9b513090
RC
3229}
3230
3231/**
3232 * ib_dma_map_sg - Map a scatter/gather list to DMA addresses
3233 * @dev: The device for which the DMA addresses are to be created
3234 * @sg: The array of scatter/gather entries
3235 * @nents: The number of scatter/gather entries
3236 * @direction: The direction of the DMA
3237 */
3238static inline int ib_dma_map_sg(struct ib_device *dev,
3239 struct scatterlist *sg, int nents,
3240 enum dma_data_direction direction)
3241{
0957c29f 3242 return dma_map_sg(dev->dma_device, sg, nents, direction);
9b513090
RC
3243}
3244
3245/**
3246 * ib_dma_unmap_sg - Unmap a scatter/gather list of DMA addresses
3247 * @dev: The device for which the DMA addresses were created
3248 * @sg: The array of scatter/gather entries
3249 * @nents: The number of scatter/gather entries
3250 * @direction: The direction of the DMA
3251 */
3252static inline void ib_dma_unmap_sg(struct ib_device *dev,
3253 struct scatterlist *sg, int nents,
3254 enum dma_data_direction direction)
3255{
0957c29f 3256 dma_unmap_sg(dev->dma_device, sg, nents, direction);
9b513090
RC
3257}
3258
cb9fbc5c
AK
3259static inline int ib_dma_map_sg_attrs(struct ib_device *dev,
3260 struct scatterlist *sg, int nents,
3261 enum dma_data_direction direction,
00085f1e 3262 unsigned long dma_attrs)
cb9fbc5c 3263{
0957c29f
BVA
3264 return dma_map_sg_attrs(dev->dma_device, sg, nents, direction,
3265 dma_attrs);
cb9fbc5c
AK
3266}
3267
3268static inline void ib_dma_unmap_sg_attrs(struct ib_device *dev,
3269 struct scatterlist *sg, int nents,
3270 enum dma_data_direction direction,
00085f1e 3271 unsigned long dma_attrs)
cb9fbc5c 3272{
0957c29f 3273 dma_unmap_sg_attrs(dev->dma_device, sg, nents, direction, dma_attrs);
cb9fbc5c 3274}
9b513090
RC
3275/**
3276 * ib_sg_dma_address - Return the DMA address from a scatter/gather entry
3277 * @dev: The device for which the DMA addresses were created
3278 * @sg: The scatter/gather entry
ea58a595
MM
3279 *
3280 * Note: this function is obsolete. To do: change all occurrences of
3281 * ib_sg_dma_address() into sg_dma_address().
9b513090
RC
3282 */
3283static inline u64 ib_sg_dma_address(struct ib_device *dev,
3284 struct scatterlist *sg)
3285{
d1998ef3 3286 return sg_dma_address(sg);
9b513090
RC
3287}
3288
3289/**
3290 * ib_sg_dma_len - Return the DMA length from a scatter/gather entry
3291 * @dev: The device for which the DMA addresses were created
3292 * @sg: The scatter/gather entry
ea58a595
MM
3293 *
3294 * Note: this function is obsolete. To do: change all occurrences of
3295 * ib_sg_dma_len() into sg_dma_len().
9b513090
RC
3296 */
3297static inline unsigned int ib_sg_dma_len(struct ib_device *dev,
3298 struct scatterlist *sg)
3299{
d1998ef3 3300 return sg_dma_len(sg);
9b513090
RC
3301}
3302
3303/**
3304 * ib_dma_sync_single_for_cpu - Prepare DMA region to be accessed by CPU
3305 * @dev: The device for which the DMA address was created
3306 * @addr: The DMA address
3307 * @size: The size of the region in bytes
3308 * @dir: The direction of the DMA
3309 */
3310static inline void ib_dma_sync_single_for_cpu(struct ib_device *dev,
3311 u64 addr,
3312 size_t size,
3313 enum dma_data_direction dir)
3314{
0957c29f 3315 dma_sync_single_for_cpu(dev->dma_device, addr, size, dir);
9b513090
RC
3316}
3317
3318/**
3319 * ib_dma_sync_single_for_device - Prepare DMA region to be accessed by device
3320 * @dev: The device for which the DMA address was created
3321 * @addr: The DMA address
3322 * @size: The size of the region in bytes
3323 * @dir: The direction of the DMA
3324 */
3325static inline void ib_dma_sync_single_for_device(struct ib_device *dev,
3326 u64 addr,
3327 size_t size,
3328 enum dma_data_direction dir)
3329{
0957c29f 3330 dma_sync_single_for_device(dev->dma_device, addr, size, dir);
9b513090
RC
3331}
3332
3333/**
3334 * ib_dma_alloc_coherent - Allocate memory and map it for DMA
3335 * @dev: The device for which the DMA address is requested
3336 * @size: The size of the region to allocate in bytes
3337 * @dma_handle: A pointer for returning the DMA address of the region
3338 * @flag: memory allocator flags
3339 */
3340static inline void *ib_dma_alloc_coherent(struct ib_device *dev,
3341 size_t size,
d43dbacf 3342 dma_addr_t *dma_handle,
9b513090
RC
3343 gfp_t flag)
3344{
0957c29f 3345 return dma_alloc_coherent(dev->dma_device, size, dma_handle, flag);
9b513090
RC
3346}
3347
3348/**
3349 * ib_dma_free_coherent - Free memory allocated by ib_dma_alloc_coherent()
3350 * @dev: The device for which the DMA addresses were allocated
3351 * @size: The size of the region
3352 * @cpu_addr: the address returned by ib_dma_alloc_coherent()
3353 * @dma_handle: the DMA address returned by ib_dma_alloc_coherent()
3354 */
3355static inline void ib_dma_free_coherent(struct ib_device *dev,
3356 size_t size, void *cpu_addr,
d43dbacf 3357 dma_addr_t dma_handle)
9b513090 3358{
0957c29f 3359 dma_free_coherent(dev->dma_device, size, cpu_addr, dma_handle);
9b513090
RC
3360}
3361
1da177e4
LT
3362/**
3363 * ib_dereg_mr - Deregisters a memory region and removes it from the
3364 * HCA translation table.
3365 * @mr: The memory region to deregister.
7083e42e
SM
3366 *
3367 * This function can fail, if the memory region has memory windows bound to it.
1da177e4
LT
3368 */
3369int ib_dereg_mr(struct ib_mr *mr);
3370
9bee178b
SG
3371struct ib_mr *ib_alloc_mr(struct ib_pd *pd,
3372 enum ib_mr_type mr_type,
3373 u32 max_num_sg);
00f7ec36 3374
00f7ec36
SW
3375/**
3376 * ib_update_fast_reg_key - updates the key portion of the fast_reg MR
3377 * R_Key and L_Key.
3378 * @mr - struct ib_mr pointer to be updated.
3379 * @newkey - new key to be used.
3380 */
3381static inline void ib_update_fast_reg_key(struct ib_mr *mr, u8 newkey)
3382{
3383 mr->lkey = (mr->lkey & 0xffffff00) | newkey;
3384 mr->rkey = (mr->rkey & 0xffffff00) | newkey;
3385}
3386
7083e42e
SM
3387/**
3388 * ib_inc_rkey - increments the key portion of the given rkey. Can be used
3389 * for calculating a new rkey for type 2 memory windows.
3390 * @rkey - the rkey to increment.
3391 */
3392static inline u32 ib_inc_rkey(u32 rkey)
3393{
3394 const u32 mask = 0x000000ff;
3395 return ((rkey + 1) & mask) | (rkey & ~mask);
3396}
3397
1da177e4
LT
3398/**
3399 * ib_alloc_fmr - Allocates a unmapped fast memory region.
3400 * @pd: The protection domain associated with the unmapped region.
3401 * @mr_access_flags: Specifies the memory access rights.
3402 * @fmr_attr: Attributes of the unmapped region.
3403 *
3404 * A fast memory region must be mapped before it can be used as part of
3405 * a work request.
3406 */
3407struct ib_fmr *ib_alloc_fmr(struct ib_pd *pd,
3408 int mr_access_flags,
3409 struct ib_fmr_attr *fmr_attr);
3410
3411/**
3412 * ib_map_phys_fmr - Maps a list of physical pages to a fast memory region.
3413 * @fmr: The fast memory region to associate with the pages.
3414 * @page_list: An array of physical pages to map to the fast memory region.
3415 * @list_len: The number of pages in page_list.
3416 * @iova: The I/O virtual address to use with the mapped region.
3417 */
3418static inline int ib_map_phys_fmr(struct ib_fmr *fmr,
3419 u64 *page_list, int list_len,
3420 u64 iova)
3421{
3422 return fmr->device->map_phys_fmr(fmr, page_list, list_len, iova);
3423}
3424
3425/**
3426 * ib_unmap_fmr - Removes the mapping from a list of fast memory regions.
3427 * @fmr_list: A linked list of fast memory regions to unmap.
3428 */
3429int ib_unmap_fmr(struct list_head *fmr_list);
3430
3431/**
3432 * ib_dealloc_fmr - Deallocates a fast memory region.
3433 * @fmr: The fast memory region to deallocate.
3434 */
3435int ib_dealloc_fmr(struct ib_fmr *fmr);
3436
3437/**
3438 * ib_attach_mcast - Attaches the specified QP to a multicast group.
3439 * @qp: QP to attach to the multicast group. The QP must be type
3440 * IB_QPT_UD.
3441 * @gid: Multicast group GID.
3442 * @lid: Multicast group LID in host byte order.
3443 *
3444 * In order to send and receive multicast packets, subnet
3445 * administration must have created the multicast group and configured
3446 * the fabric appropriately. The port associated with the specified
3447 * QP must also be a member of the multicast group.
3448 */
3449int ib_attach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid);
3450
3451/**
3452 * ib_detach_mcast - Detaches the specified QP from a multicast group.
3453 * @qp: QP to detach from the multicast group.
3454 * @gid: Multicast group GID.
3455 * @lid: Multicast group LID in host byte order.
3456 */
3457int ib_detach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid);
3458
59991f94
SH
3459/**
3460 * ib_alloc_xrcd - Allocates an XRC domain.
3461 * @device: The device on which to allocate the XRC domain.
3462 */
3463struct ib_xrcd *ib_alloc_xrcd(struct ib_device *device);
3464
3465/**
3466 * ib_dealloc_xrcd - Deallocates an XRC domain.
3467 * @xrcd: The XRC domain to deallocate.
3468 */
3469int ib_dealloc_xrcd(struct ib_xrcd *xrcd);
3470
319a441d
HHZ
3471struct ib_flow *ib_create_flow(struct ib_qp *qp,
3472 struct ib_flow_attr *flow_attr, int domain);
3473int ib_destroy_flow(struct ib_flow *flow_id);
3474
1c636f80
EC
3475static inline int ib_check_mr_access(int flags)
3476{
3477 /*
3478 * Local write permission is required if remote write or
3479 * remote atomic permission is also requested.
3480 */
3481 if (flags & (IB_ACCESS_REMOTE_ATOMIC | IB_ACCESS_REMOTE_WRITE) &&
3482 !(flags & IB_ACCESS_LOCAL_WRITE))
3483 return -EINVAL;
3484
3485 return 0;
3486}
3487
1b01d335
SG
3488/**
3489 * ib_check_mr_status: lightweight check of MR status.
3490 * This routine may provide status checks on a selected
3491 * ib_mr. first use is for signature status check.
3492 *
3493 * @mr: A memory region.
3494 * @check_mask: Bitmask of which checks to perform from
3495 * ib_mr_status_check enumeration.
3496 * @mr_status: The container of relevant status checks.
3497 * failed checks will be indicated in the status bitmask
3498 * and the relevant info shall be in the error item.
3499 */
3500int ib_check_mr_status(struct ib_mr *mr, u32 check_mask,
3501 struct ib_mr_status *mr_status);
3502
9268f72d
YK
3503struct net_device *ib_get_net_dev_by_params(struct ib_device *dev, u8 port,
3504 u16 pkey, const union ib_gid *gid,
3505 const struct sockaddr *addr);
5fd251c8
YH
3506struct ib_wq *ib_create_wq(struct ib_pd *pd,
3507 struct ib_wq_init_attr *init_attr);
3508int ib_destroy_wq(struct ib_wq *wq);
3509int ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *attr,
3510 u32 wq_attr_mask);
6d39786b
YH
3511struct ib_rwq_ind_table *ib_create_rwq_ind_table(struct ib_device *device,
3512 struct ib_rwq_ind_table_init_attr*
3513 wq_ind_table_init_attr);
3514int ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *wq_ind_table);
9268f72d 3515
ff2ba993 3516int ib_map_mr_sg(struct ib_mr *mr, struct scatterlist *sg, int sg_nents,
9aa8b321 3517 unsigned int *sg_offset, unsigned int page_size);
4c67e2bf
SG
3518
3519static inline int
ff2ba993 3520ib_map_mr_sg_zbva(struct ib_mr *mr, struct scatterlist *sg, int sg_nents,
9aa8b321 3521 unsigned int *sg_offset, unsigned int page_size)
4c67e2bf
SG
3522{
3523 int n;
3524
ff2ba993 3525 n = ib_map_mr_sg(mr, sg, sg_nents, sg_offset, page_size);
4c67e2bf
SG
3526 mr->iova = 0;
3527
3528 return n;
3529}
3530
ff2ba993 3531int ib_sg_to_pages(struct ib_mr *mr, struct scatterlist *sgl, int sg_nents,
9aa8b321 3532 unsigned int *sg_offset, int (*set_page)(struct ib_mr *, u64));
4c67e2bf 3533
765d6774
SW
3534void ib_drain_rq(struct ib_qp *qp);
3535void ib_drain_sq(struct ib_qp *qp);
3536void ib_drain_qp(struct ib_qp *qp);
850d8fd7 3537
c90ea9d8 3538int ib_resolve_eth_dmac(struct ib_device *device,
90898850 3539 struct rdma_ah_attr *ah_attr);
2224c47a
DC
3540
3541static inline u8 *rdma_ah_retrieve_dmac(struct rdma_ah_attr *attr)
3542{
44c58487
DC
3543 if (attr->type == RDMA_AH_ATTR_TYPE_ROCE)
3544 return attr->roce.dmac;
3545 return NULL;
2224c47a
DC
3546}
3547
64b4646e 3548static inline void rdma_ah_set_dlid(struct rdma_ah_attr *attr, u32 dlid)
2224c47a 3549{
44c58487 3550 if (attr->type == RDMA_AH_ATTR_TYPE_IB)
64b4646e
DC
3551 attr->ib.dlid = (u16)dlid;
3552 else if (attr->type == RDMA_AH_ATTR_TYPE_OPA)
3553 attr->opa.dlid = dlid;
2224c47a
DC
3554}
3555
64b4646e 3556static inline u32 rdma_ah_get_dlid(const struct rdma_ah_attr *attr)
2224c47a 3557{
44c58487
DC
3558 if (attr->type == RDMA_AH_ATTR_TYPE_IB)
3559 return attr->ib.dlid;
64b4646e
DC
3560 else if (attr->type == RDMA_AH_ATTR_TYPE_OPA)
3561 return attr->opa.dlid;
44c58487 3562 return 0;
2224c47a
DC
3563}
3564
3565static inline void rdma_ah_set_sl(struct rdma_ah_attr *attr, u8 sl)
3566{
3567 attr->sl = sl;
3568}
3569
3570static inline u8 rdma_ah_get_sl(const struct rdma_ah_attr *attr)
3571{
3572 return attr->sl;
3573}
3574
3575static inline void rdma_ah_set_path_bits(struct rdma_ah_attr *attr,
3576 u8 src_path_bits)
3577{
44c58487
DC
3578 if (attr->type == RDMA_AH_ATTR_TYPE_IB)
3579 attr->ib.src_path_bits = src_path_bits;
64b4646e
DC
3580 else if (attr->type == RDMA_AH_ATTR_TYPE_OPA)
3581 attr->opa.src_path_bits = src_path_bits;
2224c47a
DC
3582}
3583
3584static inline u8 rdma_ah_get_path_bits(const struct rdma_ah_attr *attr)
3585{
44c58487
DC
3586 if (attr->type == RDMA_AH_ATTR_TYPE_IB)
3587 return attr->ib.src_path_bits;
64b4646e
DC
3588 else if (attr->type == RDMA_AH_ATTR_TYPE_OPA)
3589 return attr->opa.src_path_bits;
44c58487 3590 return 0;
2224c47a
DC
3591}
3592
3593static inline void rdma_ah_set_port_num(struct rdma_ah_attr *attr, u8 port_num)
3594{
3595 attr->port_num = port_num;
3596}
3597
3598static inline u8 rdma_ah_get_port_num(const struct rdma_ah_attr *attr)
3599{
3600 return attr->port_num;
3601}
3602
3603static inline void rdma_ah_set_static_rate(struct rdma_ah_attr *attr,
3604 u8 static_rate)
3605{
3606 attr->static_rate = static_rate;
3607}
3608
3609static inline u8 rdma_ah_get_static_rate(const struct rdma_ah_attr *attr)
3610{
3611 return attr->static_rate;
3612}
3613
3614static inline void rdma_ah_set_ah_flags(struct rdma_ah_attr *attr,
3615 enum ib_ah_flags flag)
3616{
3617 attr->ah_flags = flag;
3618}
3619
3620static inline enum ib_ah_flags
3621 rdma_ah_get_ah_flags(const struct rdma_ah_attr *attr)
3622{
3623 return attr->ah_flags;
3624}
3625
3626static inline const struct ib_global_route
3627 *rdma_ah_read_grh(const struct rdma_ah_attr *attr)
3628{
3629 return &attr->grh;
3630}
3631
3632/*To retrieve and modify the grh */
3633static inline struct ib_global_route
3634 *rdma_ah_retrieve_grh(struct rdma_ah_attr *attr)
3635{
3636 return &attr->grh;
3637}
3638
3639static inline void rdma_ah_set_dgid_raw(struct rdma_ah_attr *attr, void *dgid)
3640{
3641 struct ib_global_route *grh = rdma_ah_retrieve_grh(attr);
3642
3643 memcpy(grh->dgid.raw, dgid, sizeof(grh->dgid));
3644}
3645
3646static inline void rdma_ah_set_subnet_prefix(struct rdma_ah_attr *attr,
3647 __be64 prefix)
3648{
3649 struct ib_global_route *grh = rdma_ah_retrieve_grh(attr);
3650
3651 grh->dgid.global.subnet_prefix = prefix;
3652}
3653
3654static inline void rdma_ah_set_interface_id(struct rdma_ah_attr *attr,
3655 __be64 if_id)
3656{
3657 struct ib_global_route *grh = rdma_ah_retrieve_grh(attr);
3658
3659 grh->dgid.global.interface_id = if_id;
3660}
3661
3662static inline void rdma_ah_set_grh(struct rdma_ah_attr *attr,
3663 union ib_gid *dgid, u32 flow_label,
3664 u8 sgid_index, u8 hop_limit,
3665 u8 traffic_class)
3666{
3667 struct ib_global_route *grh = rdma_ah_retrieve_grh(attr);
3668
3669 attr->ah_flags = IB_AH_GRH;
3670 if (dgid)
3671 grh->dgid = *dgid;
3672 grh->flow_label = flow_label;
3673 grh->sgid_index = sgid_index;
3674 grh->hop_limit = hop_limit;
3675 grh->traffic_class = traffic_class;
3676}
44c58487
DC
3677
3678/*Get AH type */
3679static inline enum rdma_ah_attr_type rdma_ah_find_type(struct ib_device *dev,
3680 u32 port_num)
3681{
3682 if ((rdma_protocol_roce(dev, port_num)) ||
3683 (rdma_protocol_iwarp(dev, port_num)))
3684 return RDMA_AH_ATTR_TYPE_ROCE;
64b4646e
DC
3685 else if ((rdma_protocol_ib(dev, port_num)) &&
3686 (rdma_cap_opa_ah(dev, port_num)))
3687 return RDMA_AH_ATTR_TYPE_OPA;
44c58487
DC
3688 else
3689 return RDMA_AH_ATTR_TYPE_IB;
3690}
1da177e4 3691#endif /* IB_VERBS_H */