2 * Copyright (c) 2004 Mellanox Technologies Ltd. All rights reserved.
3 * Copyright (c) 2004 Infinicon Corporation. All rights reserved.
4 * Copyright (c) 2004 Intel Corporation. All rights reserved.
5 * Copyright (c) 2004 Topspin Corporation. All rights reserved.
6 * Copyright (c) 2004 Voltaire Corporation. All rights reserved.
7 * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
8 * Copyright (c) 2005, 2006, 2007 Cisco Systems. All rights reserved.
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:
16 * Redistribution and use in source and binary forms, with or
17 * without modification, are permitted provided that the following
20 * - Redistributions of source code must retain the above
21 * copyright notice, this list of conditions and the following
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.
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
39 #if !defined(IB_VERBS_H)
42 #include <linux/types.h>
43 #include <linux/device.h>
45 #include <linux/dma-mapping.h>
46 #include <linux/kref.h>
47 #include <linux/list.h>
48 #include <linux/rwsem.h>
49 #include <linux/scatterlist.h>
50 #include <linux/workqueue.h>
51 #include <uapi/linux/if_ether.h>
53 #include <linux/atomic.h>
54 #include <asm/uaccess.h>
56 extern struct workqueue_struct *ib_wq;
67 /* IB values map to NodeInfo:NodeType. */
75 enum rdma_transport_type {
81 enum rdma_transport_type
82 rdma_node_get_transport(enum rdma_node_type node_type) __attribute_const__;
84 enum rdma_link_layer {
85 IB_LINK_LAYER_UNSPECIFIED,
86 IB_LINK_LAYER_INFINIBAND,
87 IB_LINK_LAYER_ETHERNET,
90 enum ib_device_cap_flags {
91 IB_DEVICE_RESIZE_MAX_WR = 1,
92 IB_DEVICE_BAD_PKEY_CNTR = (1<<1),
93 IB_DEVICE_BAD_QKEY_CNTR = (1<<2),
94 IB_DEVICE_RAW_MULTI = (1<<3),
95 IB_DEVICE_AUTO_PATH_MIG = (1<<4),
96 IB_DEVICE_CHANGE_PHY_PORT = (1<<5),
97 IB_DEVICE_UD_AV_PORT_ENFORCE = (1<<6),
98 IB_DEVICE_CURR_QP_STATE_MOD = (1<<7),
99 IB_DEVICE_SHUTDOWN_PORT = (1<<8),
100 IB_DEVICE_INIT_TYPE = (1<<9),
101 IB_DEVICE_PORT_ACTIVE_EVENT = (1<<10),
102 IB_DEVICE_SYS_IMAGE_GUID = (1<<11),
103 IB_DEVICE_RC_RNR_NAK_GEN = (1<<12),
104 IB_DEVICE_SRQ_RESIZE = (1<<13),
105 IB_DEVICE_N_NOTIFY_CQ = (1<<14),
106 IB_DEVICE_LOCAL_DMA_LKEY = (1<<15),
107 IB_DEVICE_RESERVED = (1<<16), /* old SEND_W_INV */
108 IB_DEVICE_MEM_WINDOW = (1<<17),
110 * Devices should set IB_DEVICE_UD_IP_SUM if they support
111 * insertion of UDP and TCP checksum on outgoing UD IPoIB
112 * messages and can verify the validity of checksum for
113 * incoming messages. Setting this flag implies that the
114 * IPoIB driver may set NETIF_F_IP_CSUM for datagram mode.
116 IB_DEVICE_UD_IP_CSUM = (1<<18),
117 IB_DEVICE_UD_TSO = (1<<19),
118 IB_DEVICE_XRC = (1<<20),
119 IB_DEVICE_MEM_MGT_EXTENSIONS = (1<<21),
120 IB_DEVICE_BLOCK_MULTICAST_LOOPBACK = (1<<22),
121 IB_DEVICE_MEM_WINDOW_TYPE_2A = (1<<23),
122 IB_DEVICE_MEM_WINDOW_TYPE_2B = (1<<24),
123 IB_DEVICE_MANAGED_FLOW_STEERING = (1<<29)
132 struct ib_device_attr {
134 __be64 sys_image_guid;
142 int device_cap_flags;
152 int max_qp_init_rd_atom;
153 int max_ee_init_rd_atom;
154 enum ib_atomic_cap atomic_cap;
155 enum ib_atomic_cap masked_atomic_cap;
162 int max_mcast_qp_attach;
163 int max_total_mcast_qp_attach;
170 unsigned int max_fast_reg_page_list_len;
172 u8 local_ca_ack_delay;
183 static inline int ib_mtu_enum_to_int(enum ib_mtu mtu)
186 case IB_MTU_256: return 256;
187 case IB_MTU_512: return 512;
188 case IB_MTU_1024: return 1024;
189 case IB_MTU_2048: return 2048;
190 case IB_MTU_4096: return 4096;
201 IB_PORT_ACTIVE_DEFER = 5
204 enum ib_port_cap_flags {
206 IB_PORT_NOTICE_SUP = 1 << 2,
207 IB_PORT_TRAP_SUP = 1 << 3,
208 IB_PORT_OPT_IPD_SUP = 1 << 4,
209 IB_PORT_AUTO_MIGR_SUP = 1 << 5,
210 IB_PORT_SL_MAP_SUP = 1 << 6,
211 IB_PORT_MKEY_NVRAM = 1 << 7,
212 IB_PORT_PKEY_NVRAM = 1 << 8,
213 IB_PORT_LED_INFO_SUP = 1 << 9,
214 IB_PORT_SM_DISABLED = 1 << 10,
215 IB_PORT_SYS_IMAGE_GUID_SUP = 1 << 11,
216 IB_PORT_PKEY_SW_EXT_PORT_TRAP_SUP = 1 << 12,
217 IB_PORT_EXTENDED_SPEEDS_SUP = 1 << 14,
218 IB_PORT_CM_SUP = 1 << 16,
219 IB_PORT_SNMP_TUNNEL_SUP = 1 << 17,
220 IB_PORT_REINIT_SUP = 1 << 18,
221 IB_PORT_DEVICE_MGMT_SUP = 1 << 19,
222 IB_PORT_VENDOR_CLASS_SUP = 1 << 20,
223 IB_PORT_DR_NOTICE_SUP = 1 << 21,
224 IB_PORT_CAP_MASK_NOTICE_SUP = 1 << 22,
225 IB_PORT_BOOT_MGMT_SUP = 1 << 23,
226 IB_PORT_LINK_LATENCY_SUP = 1 << 24,
227 IB_PORT_CLIENT_REG_SUP = 1 << 25
237 static inline int ib_width_enum_to_int(enum ib_port_width width)
240 case IB_WIDTH_1X: return 1;
241 case IB_WIDTH_4X: return 4;
242 case IB_WIDTH_8X: return 8;
243 case IB_WIDTH_12X: return 12;
257 struct ib_protocol_stats {
261 struct iw_protocol_stats {
264 u64 ipInTooBigErrors;
267 u64 ipInUnknownProtos;
268 u64 ipInTruncatedPkts;
271 u64 ipOutForwDatagrams;
303 union rdma_protocol_stats {
304 struct ib_protocol_stats ib;
305 struct iw_protocol_stats iw;
308 struct ib_port_attr {
309 enum ib_port_state state;
311 enum ib_mtu active_mtu;
330 enum ib_device_modify_flags {
331 IB_DEVICE_MODIFY_SYS_IMAGE_GUID = 1 << 0,
332 IB_DEVICE_MODIFY_NODE_DESC = 1 << 1
335 struct ib_device_modify {
340 enum ib_port_modify_flags {
341 IB_PORT_SHUTDOWN = 1,
342 IB_PORT_INIT_TYPE = (1<<2),
343 IB_PORT_RESET_QKEY_CNTR = (1<<3)
346 struct ib_port_modify {
347 u32 set_port_cap_mask;
348 u32 clr_port_cap_mask;
356 IB_EVENT_QP_ACCESS_ERR,
360 IB_EVENT_PATH_MIG_ERR,
361 IB_EVENT_DEVICE_FATAL,
362 IB_EVENT_PORT_ACTIVE,
365 IB_EVENT_PKEY_CHANGE,
368 IB_EVENT_SRQ_LIMIT_REACHED,
369 IB_EVENT_QP_LAST_WQE_REACHED,
370 IB_EVENT_CLIENT_REREGISTER,
375 struct ib_device *device;
382 enum ib_event_type event;
385 struct ib_event_handler {
386 struct ib_device *device;
387 void (*handler)(struct ib_event_handler *, struct ib_event *);
388 struct list_head list;
391 #define INIT_IB_EVENT_HANDLER(_ptr, _device, _handler) \
393 (_ptr)->device = _device; \
394 (_ptr)->handler = _handler; \
395 INIT_LIST_HEAD(&(_ptr)->list); \
398 struct ib_global_route {
407 __be32 version_tclass_flow;
416 IB_MULTICAST_QPN = 0xffffff
419 #define IB_LID_PERMISSIVE cpu_to_be16(0xFFFF)
426 IB_RATE_PORT_CURRENT = 0,
427 IB_RATE_2_5_GBPS = 2,
435 IB_RATE_120_GBPS = 10,
436 IB_RATE_14_GBPS = 11,
437 IB_RATE_56_GBPS = 12,
438 IB_RATE_112_GBPS = 13,
439 IB_RATE_168_GBPS = 14,
440 IB_RATE_25_GBPS = 15,
441 IB_RATE_100_GBPS = 16,
442 IB_RATE_200_GBPS = 17,
443 IB_RATE_300_GBPS = 18
447 * ib_rate_to_mult - Convert the IB rate enum to a multiple of the
448 * base rate of 2.5 Gbit/sec. For example, IB_RATE_5_GBPS will be
449 * converted to 2, since 5 Gbit/sec is 2 * 2.5 Gbit/sec.
450 * @rate: rate to convert.
452 int ib_rate_to_mult(enum ib_rate rate) __attribute_const__;
455 * ib_rate_to_mbps - Convert the IB rate enum to Mbps.
456 * For example, IB_RATE_2_5_GBPS will be converted to 2500.
457 * @rate: rate to convert.
459 int ib_rate_to_mbps(enum ib_rate rate) __attribute_const__;
462 * mult_to_ib_rate - Convert a multiple of 2.5 Gbit/sec to an IB rate
464 * @mult: multiple to convert.
466 enum ib_rate mult_to_ib_rate(int mult) __attribute_const__;
469 struct ib_global_route grh;
484 IB_WC_LOC_EEC_OP_ERR,
489 IB_WC_LOC_ACCESS_ERR,
490 IB_WC_REM_INV_REQ_ERR,
491 IB_WC_REM_ACCESS_ERR,
494 IB_WC_RNR_RETRY_EXC_ERR,
495 IB_WC_LOC_RDD_VIOL_ERR,
496 IB_WC_REM_INV_RD_REQ_ERR,
499 IB_WC_INV_EEC_STATE_ERR,
501 IB_WC_RESP_TIMEOUT_ERR,
515 IB_WC_MASKED_COMP_SWAP,
516 IB_WC_MASKED_FETCH_ADD,
518 * Set value of IB_WC_RECV so consumers can test if a completion is a
519 * receive by testing (opcode & IB_WC_RECV).
522 IB_WC_RECV_RDMA_WITH_IMM
527 IB_WC_WITH_IMM = (1<<1),
528 IB_WC_WITH_INVALIDATE = (1<<2),
529 IB_WC_IP_CSUM_OK = (1<<3),
530 IB_WC_WITH_SMAC = (1<<4),
531 IB_WC_WITH_VLAN = (1<<5),
536 enum ib_wc_status status;
537 enum ib_wc_opcode opcode;
551 u8 port_num; /* valid only for DR SMPs on switches */
556 enum ib_cq_notify_flags {
557 IB_CQ_SOLICITED = 1 << 0,
558 IB_CQ_NEXT_COMP = 1 << 1,
559 IB_CQ_SOLICITED_MASK = IB_CQ_SOLICITED | IB_CQ_NEXT_COMP,
560 IB_CQ_REPORT_MISSED_EVENTS = 1 << 2,
568 enum ib_srq_attr_mask {
569 IB_SRQ_MAX_WR = 1 << 0,
570 IB_SRQ_LIMIT = 1 << 1,
579 struct ib_srq_init_attr {
580 void (*event_handler)(struct ib_event *, void *);
582 struct ib_srq_attr attr;
583 enum ib_srq_type srq_type;
587 struct ib_xrcd *xrcd;
608 * IB_QPT_SMI and IB_QPT_GSI have to be the first two entries
609 * here (and in that order) since the MAD layer uses them as
610 * indices into a 2-entry table.
619 IB_QPT_RAW_ETHERTYPE,
620 IB_QPT_RAW_PACKET = 8,
624 /* Reserve a range for qp types internal to the low level driver.
625 * These qp types will not be visible at the IB core layer, so the
626 * IB_QPT_MAX usages should not be affected in the core layer
628 IB_QPT_RESERVED1 = 0x1000,
640 enum ib_qp_create_flags {
641 IB_QP_CREATE_IPOIB_UD_LSO = 1 << 0,
642 IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK = 1 << 1,
643 /* reserve bits 26-31 for low level drivers' internal use */
644 IB_QP_CREATE_RESERVED_START = 1 << 26,
645 IB_QP_CREATE_RESERVED_END = 1 << 31,
650 * Note: users may not call ib_close_qp or ib_destroy_qp from the event_handler
651 * callback to destroy the passed in QP.
654 struct ib_qp_init_attr {
655 void (*event_handler)(struct ib_event *, void *);
657 struct ib_cq *send_cq;
658 struct ib_cq *recv_cq;
660 struct ib_xrcd *xrcd; /* XRC TGT QPs only */
661 struct ib_qp_cap cap;
662 enum ib_sig_type sq_sig_type;
663 enum ib_qp_type qp_type;
664 enum ib_qp_create_flags create_flags;
665 u8 port_num; /* special QP types only */
668 struct ib_qp_open_attr {
669 void (*event_handler)(struct ib_event *, void *);
672 enum ib_qp_type qp_type;
675 enum ib_rnr_timeout {
676 IB_RNR_TIMER_655_36 = 0,
677 IB_RNR_TIMER_000_01 = 1,
678 IB_RNR_TIMER_000_02 = 2,
679 IB_RNR_TIMER_000_03 = 3,
680 IB_RNR_TIMER_000_04 = 4,
681 IB_RNR_TIMER_000_06 = 5,
682 IB_RNR_TIMER_000_08 = 6,
683 IB_RNR_TIMER_000_12 = 7,
684 IB_RNR_TIMER_000_16 = 8,
685 IB_RNR_TIMER_000_24 = 9,
686 IB_RNR_TIMER_000_32 = 10,
687 IB_RNR_TIMER_000_48 = 11,
688 IB_RNR_TIMER_000_64 = 12,
689 IB_RNR_TIMER_000_96 = 13,
690 IB_RNR_TIMER_001_28 = 14,
691 IB_RNR_TIMER_001_92 = 15,
692 IB_RNR_TIMER_002_56 = 16,
693 IB_RNR_TIMER_003_84 = 17,
694 IB_RNR_TIMER_005_12 = 18,
695 IB_RNR_TIMER_007_68 = 19,
696 IB_RNR_TIMER_010_24 = 20,
697 IB_RNR_TIMER_015_36 = 21,
698 IB_RNR_TIMER_020_48 = 22,
699 IB_RNR_TIMER_030_72 = 23,
700 IB_RNR_TIMER_040_96 = 24,
701 IB_RNR_TIMER_061_44 = 25,
702 IB_RNR_TIMER_081_92 = 26,
703 IB_RNR_TIMER_122_88 = 27,
704 IB_RNR_TIMER_163_84 = 28,
705 IB_RNR_TIMER_245_76 = 29,
706 IB_RNR_TIMER_327_68 = 30,
707 IB_RNR_TIMER_491_52 = 31
710 enum ib_qp_attr_mask {
712 IB_QP_CUR_STATE = (1<<1),
713 IB_QP_EN_SQD_ASYNC_NOTIFY = (1<<2),
714 IB_QP_ACCESS_FLAGS = (1<<3),
715 IB_QP_PKEY_INDEX = (1<<4),
719 IB_QP_PATH_MTU = (1<<8),
720 IB_QP_TIMEOUT = (1<<9),
721 IB_QP_RETRY_CNT = (1<<10),
722 IB_QP_RNR_RETRY = (1<<11),
723 IB_QP_RQ_PSN = (1<<12),
724 IB_QP_MAX_QP_RD_ATOMIC = (1<<13),
725 IB_QP_ALT_PATH = (1<<14),
726 IB_QP_MIN_RNR_TIMER = (1<<15),
727 IB_QP_SQ_PSN = (1<<16),
728 IB_QP_MAX_DEST_RD_ATOMIC = (1<<17),
729 IB_QP_PATH_MIG_STATE = (1<<18),
731 IB_QP_DEST_QPN = (1<<20),
732 IB_QP_SMAC = (1<<21),
733 IB_QP_ALT_SMAC = (1<<22),
735 IB_QP_ALT_VID = (1<<24),
760 enum ib_qp_state qp_state;
761 enum ib_qp_state cur_qp_state;
762 enum ib_mtu path_mtu;
763 enum ib_mig_state path_mig_state;
769 struct ib_qp_cap cap;
770 struct ib_ah_attr ah_attr;
771 struct ib_ah_attr alt_ah_attr;
774 u8 en_sqd_async_notify;
777 u8 max_dest_rd_atomic;
786 u8 alt_smac[ETH_ALEN];
793 IB_WR_RDMA_WRITE_WITH_IMM,
797 IB_WR_ATOMIC_CMP_AND_SWP,
798 IB_WR_ATOMIC_FETCH_AND_ADD,
801 IB_WR_RDMA_READ_WITH_INV,
804 IB_WR_MASKED_ATOMIC_CMP_AND_SWP,
805 IB_WR_MASKED_ATOMIC_FETCH_AND_ADD,
807 /* reserve values for low level drivers' internal use.
808 * These values will not be used at all in the ib core layer.
810 IB_WR_RESERVED1 = 0xf0,
824 IB_SEND_SIGNALED = (1<<1),
825 IB_SEND_SOLICITED = (1<<2),
826 IB_SEND_INLINE = (1<<3),
827 IB_SEND_IP_CSUM = (1<<4),
829 /* reserve bits 26-31 for low level drivers' internal use */
830 IB_SEND_RESERVED_START = (1 << 26),
831 IB_SEND_RESERVED_END = (1 << 31),
840 struct ib_fast_reg_page_list {
841 struct ib_device *device;
843 unsigned int max_page_list_len;
847 * struct ib_mw_bind_info - Parameters for a memory window bind operation.
848 * @mr: A memory region to bind the memory window to.
849 * @addr: The address where the memory window should begin.
850 * @length: The length of the memory window, in bytes.
851 * @mw_access_flags: Access flags from enum ib_access_flags for the window.
853 * This struct contains the shared parameters for type 1 and type 2
854 * memory window bind operations.
856 struct ib_mw_bind_info {
864 struct ib_send_wr *next;
866 struct ib_sge *sg_list;
868 enum ib_wr_opcode opcode;
883 u64 compare_add_mask;
894 u16 pkey_index; /* valid for GSI only */
895 u8 port_num; /* valid for DR SMPs on switch only */
899 struct ib_fast_reg_page_list *page_list;
900 unsigned int page_shift;
901 unsigned int page_list_len;
908 /* The new rkey for the memory window. */
910 struct ib_mw_bind_info bind_info;
913 u32 xrc_remote_srq_num; /* XRC TGT QPs only */
917 struct ib_recv_wr *next;
919 struct ib_sge *sg_list;
923 enum ib_access_flags {
924 IB_ACCESS_LOCAL_WRITE = 1,
925 IB_ACCESS_REMOTE_WRITE = (1<<1),
926 IB_ACCESS_REMOTE_READ = (1<<2),
927 IB_ACCESS_REMOTE_ATOMIC = (1<<3),
928 IB_ACCESS_MW_BIND = (1<<4),
929 IB_ZERO_BASED = (1<<5)
939 u64 device_virt_addr;
946 enum ib_mr_rereg_flags {
947 IB_MR_REREG_TRANS = 1,
948 IB_MR_REREG_PD = (1<<1),
949 IB_MR_REREG_ACCESS = (1<<2)
953 * struct ib_mw_bind - Parameters for a type 1 memory window bind operation.
954 * @wr_id: Work request id.
955 * @send_flags: Flags from ib_send_flags enum.
956 * @bind_info: More parameters of the bind operation.
961 struct ib_mw_bind_info bind_info;
971 struct ib_device *device;
972 struct list_head pd_list;
973 struct list_head mr_list;
974 struct list_head mw_list;
975 struct list_head cq_list;
976 struct list_head qp_list;
977 struct list_head srq_list;
978 struct list_head ah_list;
979 struct list_head xrcd_list;
980 struct list_head rule_list;
985 u64 user_handle; /* handle given to us by userspace */
986 struct ib_ucontext *context; /* associated user context */
987 void *object; /* containing object */
988 struct list_head list; /* link to context's list */
989 int id; /* index into kernel idr */
991 struct rw_semaphore mutex; /* protects .live */
996 const void __user *inbuf;
1003 struct ib_device *device;
1004 struct ib_uobject *uobject;
1005 atomic_t usecnt; /* count all resources */
1009 struct ib_device *device;
1010 atomic_t usecnt; /* count all exposed resources */
1011 struct inode *inode;
1013 struct mutex tgt_qp_mutex;
1014 struct list_head tgt_qp_list;
1018 struct ib_device *device;
1020 struct ib_uobject *uobject;
1023 typedef void (*ib_comp_handler)(struct ib_cq *cq, void *cq_context);
1026 struct ib_device *device;
1027 struct ib_uobject *uobject;
1028 ib_comp_handler comp_handler;
1029 void (*event_handler)(struct ib_event *, void *);
1032 atomic_t usecnt; /* count number of work queues */
1036 struct ib_device *device;
1038 struct ib_uobject *uobject;
1039 void (*event_handler)(struct ib_event *, void *);
1041 enum ib_srq_type srq_type;
1046 struct ib_xrcd *xrcd;
1054 struct ib_device *device;
1056 struct ib_cq *send_cq;
1057 struct ib_cq *recv_cq;
1059 struct ib_xrcd *xrcd; /* XRC TGT QPs only */
1060 struct list_head xrcd_list;
1061 /* count times opened, mcast attaches, flow attaches */
1063 struct list_head open_list;
1064 struct ib_qp *real_qp;
1065 struct ib_uobject *uobject;
1066 void (*event_handler)(struct ib_event *, void *);
1069 enum ib_qp_type qp_type;
1073 struct ib_device *device;
1075 struct ib_uobject *uobject;
1078 atomic_t usecnt; /* count number of MWs */
1082 struct ib_device *device;
1084 struct ib_uobject *uobject;
1086 enum ib_mw_type type;
1090 struct ib_device *device;
1092 struct list_head list;
1097 /* Supported steering options */
1098 enum ib_flow_attr_type {
1099 /* steering according to rule specifications */
1100 IB_FLOW_ATTR_NORMAL = 0x0,
1101 /* default unicast and multicast rule -
1102 * receive all Eth traffic which isn't steered to any QP
1104 IB_FLOW_ATTR_ALL_DEFAULT = 0x1,
1105 /* default multicast rule -
1106 * receive all Eth multicast traffic which isn't steered to any QP
1108 IB_FLOW_ATTR_MC_DEFAULT = 0x2,
1109 /* sniffer rule - receive all port traffic */
1110 IB_FLOW_ATTR_SNIFFER = 0x3
1113 /* Supported steering header types */
1114 enum ib_flow_spec_type {
1116 IB_FLOW_SPEC_ETH = 0x20,
1118 IB_FLOW_SPEC_IPV4 = 0x30,
1120 IB_FLOW_SPEC_TCP = 0x40,
1121 IB_FLOW_SPEC_UDP = 0x41
1124 #define IB_FLOW_SPEC_SUPPORT_LAYERS 4
1126 /* Flow steering rule priority is set according to it's domain.
1127 * Lower domain value means higher priority.
1129 enum ib_flow_domain {
1130 IB_FLOW_DOMAIN_USER,
1131 IB_FLOW_DOMAIN_ETHTOOL,
1134 IB_FLOW_DOMAIN_NUM /* Must be last */
1137 struct ib_flow_eth_filter {
1144 struct ib_flow_spec_eth {
1145 enum ib_flow_spec_type type;
1147 struct ib_flow_eth_filter val;
1148 struct ib_flow_eth_filter mask;
1151 struct ib_flow_ipv4_filter {
1156 struct ib_flow_spec_ipv4 {
1157 enum ib_flow_spec_type type;
1159 struct ib_flow_ipv4_filter val;
1160 struct ib_flow_ipv4_filter mask;
1163 struct ib_flow_tcp_udp_filter {
1168 struct ib_flow_spec_tcp_udp {
1169 enum ib_flow_spec_type type;
1171 struct ib_flow_tcp_udp_filter val;
1172 struct ib_flow_tcp_udp_filter mask;
1175 union ib_flow_spec {
1177 enum ib_flow_spec_type type;
1180 struct ib_flow_spec_eth eth;
1181 struct ib_flow_spec_ipv4 ipv4;
1182 struct ib_flow_spec_tcp_udp tcp_udp;
1185 struct ib_flow_attr {
1186 enum ib_flow_attr_type type;
1192 /* Following are the optional layers according to user request
1193 * struct ib_flow_spec_xxx
1194 * struct ib_flow_spec_yyy
1200 struct ib_uobject *uobject;
1206 enum ib_process_mad_flags {
1207 IB_MAD_IGNORE_MKEY = 1,
1208 IB_MAD_IGNORE_BKEY = 2,
1209 IB_MAD_IGNORE_ALL = IB_MAD_IGNORE_MKEY | IB_MAD_IGNORE_BKEY
1212 enum ib_mad_result {
1213 IB_MAD_RESULT_FAILURE = 0, /* (!SUCCESS is the important flag) */
1214 IB_MAD_RESULT_SUCCESS = 1 << 0, /* MAD was successfully processed */
1215 IB_MAD_RESULT_REPLY = 1 << 1, /* Reply packet needs to be sent */
1216 IB_MAD_RESULT_CONSUMED = 1 << 2 /* Packet consumed: stop processing */
1219 #define IB_DEVICE_NAME_MAX 64
1223 struct ib_event_handler event_handler;
1224 struct ib_pkey_cache **pkey_cache;
1225 struct ib_gid_cache **gid_cache;
1229 struct ib_dma_mapping_ops {
1230 int (*mapping_error)(struct ib_device *dev,
1232 u64 (*map_single)(struct ib_device *dev,
1233 void *ptr, size_t size,
1234 enum dma_data_direction direction);
1235 void (*unmap_single)(struct ib_device *dev,
1236 u64 addr, size_t size,
1237 enum dma_data_direction direction);
1238 u64 (*map_page)(struct ib_device *dev,
1239 struct page *page, unsigned long offset,
1241 enum dma_data_direction direction);
1242 void (*unmap_page)(struct ib_device *dev,
1243 u64 addr, size_t size,
1244 enum dma_data_direction direction);
1245 int (*map_sg)(struct ib_device *dev,
1246 struct scatterlist *sg, int nents,
1247 enum dma_data_direction direction);
1248 void (*unmap_sg)(struct ib_device *dev,
1249 struct scatterlist *sg, int nents,
1250 enum dma_data_direction direction);
1251 u64 (*dma_address)(struct ib_device *dev,
1252 struct scatterlist *sg);
1253 unsigned int (*dma_len)(struct ib_device *dev,
1254 struct scatterlist *sg);
1255 void (*sync_single_for_cpu)(struct ib_device *dev,
1258 enum dma_data_direction dir);
1259 void (*sync_single_for_device)(struct ib_device *dev,
1262 enum dma_data_direction dir);
1263 void *(*alloc_coherent)(struct ib_device *dev,
1267 void (*free_coherent)(struct ib_device *dev,
1268 size_t size, void *cpu_addr,
1275 struct device *dma_device;
1277 char name[IB_DEVICE_NAME_MAX];
1279 struct list_head event_handler_list;
1280 spinlock_t event_handler_lock;
1282 spinlock_t client_data_lock;
1283 struct list_head core_list;
1284 struct list_head client_data_list;
1286 struct ib_cache cache;
1290 int num_comp_vectors;
1292 struct iw_cm_verbs *iwcm;
1294 int (*get_protocol_stats)(struct ib_device *device,
1295 union rdma_protocol_stats *stats);
1296 int (*query_device)(struct ib_device *device,
1297 struct ib_device_attr *device_attr);
1298 int (*query_port)(struct ib_device *device,
1300 struct ib_port_attr *port_attr);
1301 enum rdma_link_layer (*get_link_layer)(struct ib_device *device,
1303 int (*query_gid)(struct ib_device *device,
1304 u8 port_num, int index,
1306 int (*query_pkey)(struct ib_device *device,
1307 u8 port_num, u16 index, u16 *pkey);
1308 int (*modify_device)(struct ib_device *device,
1309 int device_modify_mask,
1310 struct ib_device_modify *device_modify);
1311 int (*modify_port)(struct ib_device *device,
1312 u8 port_num, int port_modify_mask,
1313 struct ib_port_modify *port_modify);
1314 struct ib_ucontext * (*alloc_ucontext)(struct ib_device *device,
1315 struct ib_udata *udata);
1316 int (*dealloc_ucontext)(struct ib_ucontext *context);
1317 int (*mmap)(struct ib_ucontext *context,
1318 struct vm_area_struct *vma);
1319 struct ib_pd * (*alloc_pd)(struct ib_device *device,
1320 struct ib_ucontext *context,
1321 struct ib_udata *udata);
1322 int (*dealloc_pd)(struct ib_pd *pd);
1323 struct ib_ah * (*create_ah)(struct ib_pd *pd,
1324 struct ib_ah_attr *ah_attr);
1325 int (*modify_ah)(struct ib_ah *ah,
1326 struct ib_ah_attr *ah_attr);
1327 int (*query_ah)(struct ib_ah *ah,
1328 struct ib_ah_attr *ah_attr);
1329 int (*destroy_ah)(struct ib_ah *ah);
1330 struct ib_srq * (*create_srq)(struct ib_pd *pd,
1331 struct ib_srq_init_attr *srq_init_attr,
1332 struct ib_udata *udata);
1333 int (*modify_srq)(struct ib_srq *srq,
1334 struct ib_srq_attr *srq_attr,
1335 enum ib_srq_attr_mask srq_attr_mask,
1336 struct ib_udata *udata);
1337 int (*query_srq)(struct ib_srq *srq,
1338 struct ib_srq_attr *srq_attr);
1339 int (*destroy_srq)(struct ib_srq *srq);
1340 int (*post_srq_recv)(struct ib_srq *srq,
1341 struct ib_recv_wr *recv_wr,
1342 struct ib_recv_wr **bad_recv_wr);
1343 struct ib_qp * (*create_qp)(struct ib_pd *pd,
1344 struct ib_qp_init_attr *qp_init_attr,
1345 struct ib_udata *udata);
1346 int (*modify_qp)(struct ib_qp *qp,
1347 struct ib_qp_attr *qp_attr,
1349 struct ib_udata *udata);
1350 int (*query_qp)(struct ib_qp *qp,
1351 struct ib_qp_attr *qp_attr,
1353 struct ib_qp_init_attr *qp_init_attr);
1354 int (*destroy_qp)(struct ib_qp *qp);
1355 int (*post_send)(struct ib_qp *qp,
1356 struct ib_send_wr *send_wr,
1357 struct ib_send_wr **bad_send_wr);
1358 int (*post_recv)(struct ib_qp *qp,
1359 struct ib_recv_wr *recv_wr,
1360 struct ib_recv_wr **bad_recv_wr);
1361 struct ib_cq * (*create_cq)(struct ib_device *device, int cqe,
1363 struct ib_ucontext *context,
1364 struct ib_udata *udata);
1365 int (*modify_cq)(struct ib_cq *cq, u16 cq_count,
1367 int (*destroy_cq)(struct ib_cq *cq);
1368 int (*resize_cq)(struct ib_cq *cq, int cqe,
1369 struct ib_udata *udata);
1370 int (*poll_cq)(struct ib_cq *cq, int num_entries,
1372 int (*peek_cq)(struct ib_cq *cq, int wc_cnt);
1373 int (*req_notify_cq)(struct ib_cq *cq,
1374 enum ib_cq_notify_flags flags);
1375 int (*req_ncomp_notif)(struct ib_cq *cq,
1377 struct ib_mr * (*get_dma_mr)(struct ib_pd *pd,
1378 int mr_access_flags);
1379 struct ib_mr * (*reg_phys_mr)(struct ib_pd *pd,
1380 struct ib_phys_buf *phys_buf_array,
1382 int mr_access_flags,
1384 struct ib_mr * (*reg_user_mr)(struct ib_pd *pd,
1385 u64 start, u64 length,
1387 int mr_access_flags,
1388 struct ib_udata *udata);
1389 int (*query_mr)(struct ib_mr *mr,
1390 struct ib_mr_attr *mr_attr);
1391 int (*dereg_mr)(struct ib_mr *mr);
1392 struct ib_mr * (*alloc_fast_reg_mr)(struct ib_pd *pd,
1393 int max_page_list_len);
1394 struct ib_fast_reg_page_list * (*alloc_fast_reg_page_list)(struct ib_device *device,
1396 void (*free_fast_reg_page_list)(struct ib_fast_reg_page_list *page_list);
1397 int (*rereg_phys_mr)(struct ib_mr *mr,
1400 struct ib_phys_buf *phys_buf_array,
1402 int mr_access_flags,
1404 struct ib_mw * (*alloc_mw)(struct ib_pd *pd,
1405 enum ib_mw_type type);
1406 int (*bind_mw)(struct ib_qp *qp,
1408 struct ib_mw_bind *mw_bind);
1409 int (*dealloc_mw)(struct ib_mw *mw);
1410 struct ib_fmr * (*alloc_fmr)(struct ib_pd *pd,
1411 int mr_access_flags,
1412 struct ib_fmr_attr *fmr_attr);
1413 int (*map_phys_fmr)(struct ib_fmr *fmr,
1414 u64 *page_list, int list_len,
1416 int (*unmap_fmr)(struct list_head *fmr_list);
1417 int (*dealloc_fmr)(struct ib_fmr *fmr);
1418 int (*attach_mcast)(struct ib_qp *qp,
1421 int (*detach_mcast)(struct ib_qp *qp,
1424 int (*process_mad)(struct ib_device *device,
1425 int process_mad_flags,
1427 struct ib_wc *in_wc,
1428 struct ib_grh *in_grh,
1429 struct ib_mad *in_mad,
1430 struct ib_mad *out_mad);
1431 struct ib_xrcd * (*alloc_xrcd)(struct ib_device *device,
1432 struct ib_ucontext *ucontext,
1433 struct ib_udata *udata);
1434 int (*dealloc_xrcd)(struct ib_xrcd *xrcd);
1435 struct ib_flow * (*create_flow)(struct ib_qp *qp,
1439 int (*destroy_flow)(struct ib_flow *flow_id);
1441 struct ib_dma_mapping_ops *dma_ops;
1443 struct module *owner;
1445 struct kobject *ports_parent;
1446 struct list_head port_list;
1449 IB_DEV_UNINITIALIZED,
1455 u64 uverbs_cmd_mask;
1456 u64 uverbs_ex_cmd_mask;
1467 void (*add) (struct ib_device *);
1468 void (*remove)(struct ib_device *);
1470 struct list_head list;
1473 struct ib_device *ib_alloc_device(size_t size);
1474 void ib_dealloc_device(struct ib_device *device);
1476 int ib_register_device(struct ib_device *device,
1477 int (*port_callback)(struct ib_device *,
1478 u8, struct kobject *));
1479 void ib_unregister_device(struct ib_device *device);
1481 int ib_register_client (struct ib_client *client);
1482 void ib_unregister_client(struct ib_client *client);
1484 void *ib_get_client_data(struct ib_device *device, struct ib_client *client);
1485 void ib_set_client_data(struct ib_device *device, struct ib_client *client,
1488 static inline int ib_copy_from_udata(void *dest, struct ib_udata *udata, size_t len)
1490 return copy_from_user(dest, udata->inbuf, len) ? -EFAULT : 0;
1493 static inline int ib_copy_to_udata(struct ib_udata *udata, void *src, size_t len)
1495 return copy_to_user(udata->outbuf, src, len) ? -EFAULT : 0;
1499 * ib_modify_qp_is_ok - Check that the supplied attribute mask
1500 * contains all required attributes and no attributes not allowed for
1501 * the given QP state transition.
1502 * @cur_state: Current QP state
1503 * @next_state: Next QP state
1505 * @mask: Mask of supplied QP attributes
1506 * @ll : link layer of port
1508 * This function is a helper function that a low-level driver's
1509 * modify_qp method can use to validate the consumer's input. It
1510 * checks that cur_state and next_state are valid QP states, that a
1511 * transition from cur_state to next_state is allowed by the IB spec,
1512 * and that the attribute mask supplied is allowed for the transition.
1514 int ib_modify_qp_is_ok(enum ib_qp_state cur_state, enum ib_qp_state next_state,
1515 enum ib_qp_type type, enum ib_qp_attr_mask mask,
1516 enum rdma_link_layer ll);
1518 int ib_register_event_handler (struct ib_event_handler *event_handler);
1519 int ib_unregister_event_handler(struct ib_event_handler *event_handler);
1520 void ib_dispatch_event(struct ib_event *event);
1522 int ib_query_device(struct ib_device *device,
1523 struct ib_device_attr *device_attr);
1525 int ib_query_port(struct ib_device *device,
1526 u8 port_num, struct ib_port_attr *port_attr);
1528 enum rdma_link_layer rdma_port_get_link_layer(struct ib_device *device,
1531 int ib_query_gid(struct ib_device *device,
1532 u8 port_num, int index, union ib_gid *gid);
1534 int ib_query_pkey(struct ib_device *device,
1535 u8 port_num, u16 index, u16 *pkey);
1537 int ib_modify_device(struct ib_device *device,
1538 int device_modify_mask,
1539 struct ib_device_modify *device_modify);
1541 int ib_modify_port(struct ib_device *device,
1542 u8 port_num, int port_modify_mask,
1543 struct ib_port_modify *port_modify);
1545 int ib_find_gid(struct ib_device *device, union ib_gid *gid,
1546 u8 *port_num, u16 *index);
1548 int ib_find_pkey(struct ib_device *device,
1549 u8 port_num, u16 pkey, u16 *index);
1552 * ib_alloc_pd - Allocates an unused protection domain.
1553 * @device: The device on which to allocate the protection domain.
1555 * A protection domain object provides an association between QPs, shared
1556 * receive queues, address handles, memory regions, and memory windows.
1558 struct ib_pd *ib_alloc_pd(struct ib_device *device);
1561 * ib_dealloc_pd - Deallocates a protection domain.
1562 * @pd: The protection domain to deallocate.
1564 int ib_dealloc_pd(struct ib_pd *pd);
1567 * ib_create_ah - Creates an address handle for the given address vector.
1568 * @pd: The protection domain associated with the address handle.
1569 * @ah_attr: The attributes of the address vector.
1571 * The address handle is used to reference a local or global destination
1572 * in all UD QP post sends.
1574 struct ib_ah *ib_create_ah(struct ib_pd *pd, struct ib_ah_attr *ah_attr);
1577 * ib_init_ah_from_wc - Initializes address handle attributes from a
1579 * @device: Device on which the received message arrived.
1580 * @port_num: Port on which the received message arrived.
1581 * @wc: Work completion associated with the received message.
1582 * @grh: References the received global route header. This parameter is
1583 * ignored unless the work completion indicates that the GRH is valid.
1584 * @ah_attr: Returned attributes that can be used when creating an address
1585 * handle for replying to the message.
1587 int ib_init_ah_from_wc(struct ib_device *device, u8 port_num, struct ib_wc *wc,
1588 struct ib_grh *grh, struct ib_ah_attr *ah_attr);
1591 * ib_create_ah_from_wc - Creates an address handle associated with the
1592 * sender of the specified work completion.
1593 * @pd: The protection domain associated with the address handle.
1594 * @wc: Work completion information associated with a received message.
1595 * @grh: References the received global route header. This parameter is
1596 * ignored unless the work completion indicates that the GRH is valid.
1597 * @port_num: The outbound port number to associate with the address.
1599 * The address handle is used to reference a local or global destination
1600 * in all UD QP post sends.
1602 struct ib_ah *ib_create_ah_from_wc(struct ib_pd *pd, struct ib_wc *wc,
1603 struct ib_grh *grh, u8 port_num);
1606 * ib_modify_ah - Modifies the address vector associated with an address
1608 * @ah: The address handle to modify.
1609 * @ah_attr: The new address vector attributes to associate with the
1612 int ib_modify_ah(struct ib_ah *ah, struct ib_ah_attr *ah_attr);
1615 * ib_query_ah - Queries the address vector associated with an address
1617 * @ah: The address handle to query.
1618 * @ah_attr: The address vector attributes associated with the address
1621 int ib_query_ah(struct ib_ah *ah, struct ib_ah_attr *ah_attr);
1624 * ib_destroy_ah - Destroys an address handle.
1625 * @ah: The address handle to destroy.
1627 int ib_destroy_ah(struct ib_ah *ah);
1630 * ib_create_srq - Creates a SRQ associated with the specified protection
1632 * @pd: The protection domain associated with the SRQ.
1633 * @srq_init_attr: A list of initial attributes required to create the
1634 * SRQ. If SRQ creation succeeds, then the attributes are updated to
1635 * the actual capabilities of the created SRQ.
1637 * srq_attr->max_wr and srq_attr->max_sge are read the determine the
1638 * requested size of the SRQ, and set to the actual values allocated
1639 * on return. If ib_create_srq() succeeds, then max_wr and max_sge
1640 * will always be at least as large as the requested values.
1642 struct ib_srq *ib_create_srq(struct ib_pd *pd,
1643 struct ib_srq_init_attr *srq_init_attr);
1646 * ib_modify_srq - Modifies the attributes for the specified SRQ.
1647 * @srq: The SRQ to modify.
1648 * @srq_attr: On input, specifies the SRQ attributes to modify. On output,
1649 * the current values of selected SRQ attributes are returned.
1650 * @srq_attr_mask: A bit-mask used to specify which attributes of the SRQ
1651 * are being modified.
1653 * The mask may contain IB_SRQ_MAX_WR to resize the SRQ and/or
1654 * IB_SRQ_LIMIT to set the SRQ's limit and request notification when
1655 * the number of receives queued drops below the limit.
1657 int ib_modify_srq(struct ib_srq *srq,
1658 struct ib_srq_attr *srq_attr,
1659 enum ib_srq_attr_mask srq_attr_mask);
1662 * ib_query_srq - Returns the attribute list and current values for the
1664 * @srq: The SRQ to query.
1665 * @srq_attr: The attributes of the specified SRQ.
1667 int ib_query_srq(struct ib_srq *srq,
1668 struct ib_srq_attr *srq_attr);
1671 * ib_destroy_srq - Destroys the specified SRQ.
1672 * @srq: The SRQ to destroy.
1674 int ib_destroy_srq(struct ib_srq *srq);
1677 * ib_post_srq_recv - Posts a list of work requests to the specified SRQ.
1678 * @srq: The SRQ to post the work request on.
1679 * @recv_wr: A list of work requests to post on the receive queue.
1680 * @bad_recv_wr: On an immediate failure, this parameter will reference
1681 * the work request that failed to be posted on the QP.
1683 static inline int ib_post_srq_recv(struct ib_srq *srq,
1684 struct ib_recv_wr *recv_wr,
1685 struct ib_recv_wr **bad_recv_wr)
1687 return srq->device->post_srq_recv(srq, recv_wr, bad_recv_wr);
1691 * ib_create_qp - Creates a QP associated with the specified protection
1693 * @pd: The protection domain associated with the QP.
1694 * @qp_init_attr: A list of initial attributes required to create the
1695 * QP. If QP creation succeeds, then the attributes are updated to
1696 * the actual capabilities of the created QP.
1698 struct ib_qp *ib_create_qp(struct ib_pd *pd,
1699 struct ib_qp_init_attr *qp_init_attr);
1702 * ib_modify_qp - Modifies the attributes for the specified QP and then
1703 * transitions the QP to the given state.
1704 * @qp: The QP to modify.
1705 * @qp_attr: On input, specifies the QP attributes to modify. On output,
1706 * the current values of selected QP attributes are returned.
1707 * @qp_attr_mask: A bit-mask used to specify which attributes of the QP
1708 * are being modified.
1710 int ib_modify_qp(struct ib_qp *qp,
1711 struct ib_qp_attr *qp_attr,
1715 * ib_query_qp - Returns the attribute list and current values for the
1717 * @qp: The QP to query.
1718 * @qp_attr: The attributes of the specified QP.
1719 * @qp_attr_mask: A bit-mask used to select specific attributes to query.
1720 * @qp_init_attr: Additional attributes of the selected QP.
1722 * The qp_attr_mask may be used to limit the query to gathering only the
1723 * selected attributes.
1725 int ib_query_qp(struct ib_qp *qp,
1726 struct ib_qp_attr *qp_attr,
1728 struct ib_qp_init_attr *qp_init_attr);
1731 * ib_destroy_qp - Destroys the specified QP.
1732 * @qp: The QP to destroy.
1734 int ib_destroy_qp(struct ib_qp *qp);
1737 * ib_open_qp - Obtain a reference to an existing sharable QP.
1738 * @xrcd - XRC domain
1739 * @qp_open_attr: Attributes identifying the QP to open.
1741 * Returns a reference to a sharable QP.
1743 struct ib_qp *ib_open_qp(struct ib_xrcd *xrcd,
1744 struct ib_qp_open_attr *qp_open_attr);
1747 * ib_close_qp - Release an external reference to a QP.
1748 * @qp: The QP handle to release
1750 * The opened QP handle is released by the caller. The underlying
1751 * shared QP is not destroyed until all internal references are released.
1753 int ib_close_qp(struct ib_qp *qp);
1756 * ib_post_send - Posts a list of work requests to the send queue of
1758 * @qp: The QP to post the work request on.
1759 * @send_wr: A list of work requests to post on the send queue.
1760 * @bad_send_wr: On an immediate failure, this parameter will reference
1761 * the work request that failed to be posted on the QP.
1763 * While IBA Vol. 1 section 11.4.1.1 specifies that if an immediate
1764 * error is returned, the QP state shall not be affected,
1765 * ib_post_send() will return an immediate error after queueing any
1766 * earlier work requests in the list.
1768 static inline int ib_post_send(struct ib_qp *qp,
1769 struct ib_send_wr *send_wr,
1770 struct ib_send_wr **bad_send_wr)
1772 return qp->device->post_send(qp, send_wr, bad_send_wr);
1776 * ib_post_recv - Posts a list of work requests to the receive queue of
1778 * @qp: The QP to post the work request on.
1779 * @recv_wr: A list of work requests to post on the receive queue.
1780 * @bad_recv_wr: On an immediate failure, this parameter will reference
1781 * the work request that failed to be posted on the QP.
1783 static inline int ib_post_recv(struct ib_qp *qp,
1784 struct ib_recv_wr *recv_wr,
1785 struct ib_recv_wr **bad_recv_wr)
1787 return qp->device->post_recv(qp, recv_wr, bad_recv_wr);
1791 * ib_create_cq - Creates a CQ on the specified device.
1792 * @device: The device on which to create the CQ.
1793 * @comp_handler: A user-specified callback that is invoked when a
1794 * completion event occurs on the CQ.
1795 * @event_handler: A user-specified callback that is invoked when an
1796 * asynchronous event not associated with a completion occurs on the CQ.
1797 * @cq_context: Context associated with the CQ returned to the user via
1798 * the associated completion and event handlers.
1799 * @cqe: The minimum size of the CQ.
1800 * @comp_vector - Completion vector used to signal completion events.
1801 * Must be >= 0 and < context->num_comp_vectors.
1803 * Users can examine the cq structure to determine the actual CQ size.
1805 struct ib_cq *ib_create_cq(struct ib_device *device,
1806 ib_comp_handler comp_handler,
1807 void (*event_handler)(struct ib_event *, void *),
1808 void *cq_context, int cqe, int comp_vector);
1811 * ib_resize_cq - Modifies the capacity of the CQ.
1812 * @cq: The CQ to resize.
1813 * @cqe: The minimum size of the CQ.
1815 * Users can examine the cq structure to determine the actual CQ size.
1817 int ib_resize_cq(struct ib_cq *cq, int cqe);
1820 * ib_modify_cq - Modifies moderation params of the CQ
1821 * @cq: The CQ to modify.
1822 * @cq_count: number of CQEs that will trigger an event
1823 * @cq_period: max period of time in usec before triggering an event
1826 int ib_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period);
1829 * ib_destroy_cq - Destroys the specified CQ.
1830 * @cq: The CQ to destroy.
1832 int ib_destroy_cq(struct ib_cq *cq);
1835 * ib_poll_cq - poll a CQ for completion(s)
1836 * @cq:the CQ being polled
1837 * @num_entries:maximum number of completions to return
1838 * @wc:array of at least @num_entries &struct ib_wc where completions
1841 * Poll a CQ for (possibly multiple) completions. If the return value
1842 * is < 0, an error occurred. If the return value is >= 0, it is the
1843 * number of completions returned. If the return value is
1844 * non-negative and < num_entries, then the CQ was emptied.
1846 static inline int ib_poll_cq(struct ib_cq *cq, int num_entries,
1849 return cq->device->poll_cq(cq, num_entries, wc);
1853 * ib_peek_cq - Returns the number of unreaped completions currently
1854 * on the specified CQ.
1855 * @cq: The CQ to peek.
1856 * @wc_cnt: A minimum number of unreaped completions to check for.
1858 * If the number of unreaped completions is greater than or equal to wc_cnt,
1859 * this function returns wc_cnt, otherwise, it returns the actual number of
1860 * unreaped completions.
1862 int ib_peek_cq(struct ib_cq *cq, int wc_cnt);
1865 * ib_req_notify_cq - Request completion notification on a CQ.
1866 * @cq: The CQ to generate an event for.
1868 * Must contain exactly one of %IB_CQ_SOLICITED or %IB_CQ_NEXT_COMP
1869 * to request an event on the next solicited event or next work
1870 * completion at any type, respectively. %IB_CQ_REPORT_MISSED_EVENTS
1871 * may also be |ed in to request a hint about missed events, as
1875 * < 0 means an error occurred while requesting notification
1876 * == 0 means notification was requested successfully, and if
1877 * IB_CQ_REPORT_MISSED_EVENTS was passed in, then no events
1878 * were missed and it is safe to wait for another event. In
1879 * this case is it guaranteed that any work completions added
1880 * to the CQ since the last CQ poll will trigger a completion
1881 * notification event.
1882 * > 0 is only returned if IB_CQ_REPORT_MISSED_EVENTS was passed
1883 * in. It means that the consumer must poll the CQ again to
1884 * make sure it is empty to avoid missing an event because of a
1885 * race between requesting notification and an entry being
1886 * added to the CQ. This return value means it is possible
1887 * (but not guaranteed) that a work completion has been added
1888 * to the CQ since the last poll without triggering a
1889 * completion notification event.
1891 static inline int ib_req_notify_cq(struct ib_cq *cq,
1892 enum ib_cq_notify_flags flags)
1894 return cq->device->req_notify_cq(cq, flags);
1898 * ib_req_ncomp_notif - Request completion notification when there are
1899 * at least the specified number of unreaped completions on the CQ.
1900 * @cq: The CQ to generate an event for.
1901 * @wc_cnt: The number of unreaped completions that should be on the
1902 * CQ before an event is generated.
1904 static inline int ib_req_ncomp_notif(struct ib_cq *cq, int wc_cnt)
1906 return cq->device->req_ncomp_notif ?
1907 cq->device->req_ncomp_notif(cq, wc_cnt) :
1912 * ib_get_dma_mr - Returns a memory region for system memory that is
1914 * @pd: The protection domain associated with the memory region.
1915 * @mr_access_flags: Specifies the memory access rights.
1917 * Note that the ib_dma_*() functions defined below must be used
1918 * to create/destroy addresses used with the Lkey or Rkey returned
1919 * by ib_get_dma_mr().
1921 struct ib_mr *ib_get_dma_mr(struct ib_pd *pd, int mr_access_flags);
1924 * ib_dma_mapping_error - check a DMA addr for error
1925 * @dev: The device for which the dma_addr was created
1926 * @dma_addr: The DMA address to check
1928 static inline int ib_dma_mapping_error(struct ib_device *dev, u64 dma_addr)
1931 return dev->dma_ops->mapping_error(dev, dma_addr);
1932 return dma_mapping_error(dev->dma_device, dma_addr);
1936 * ib_dma_map_single - Map a kernel virtual address to DMA address
1937 * @dev: The device for which the dma_addr is to be created
1938 * @cpu_addr: The kernel virtual address
1939 * @size: The size of the region in bytes
1940 * @direction: The direction of the DMA
1942 static inline u64 ib_dma_map_single(struct ib_device *dev,
1943 void *cpu_addr, size_t size,
1944 enum dma_data_direction direction)
1947 return dev->dma_ops->map_single(dev, cpu_addr, size, direction);
1948 return dma_map_single(dev->dma_device, cpu_addr, size, direction);
1952 * ib_dma_unmap_single - Destroy a mapping created by ib_dma_map_single()
1953 * @dev: The device for which the DMA address was created
1954 * @addr: The DMA address
1955 * @size: The size of the region in bytes
1956 * @direction: The direction of the DMA
1958 static inline void ib_dma_unmap_single(struct ib_device *dev,
1959 u64 addr, size_t size,
1960 enum dma_data_direction direction)
1963 dev->dma_ops->unmap_single(dev, addr, size, direction);
1965 dma_unmap_single(dev->dma_device, addr, size, direction);
1968 static inline u64 ib_dma_map_single_attrs(struct ib_device *dev,
1969 void *cpu_addr, size_t size,
1970 enum dma_data_direction direction,
1971 struct dma_attrs *attrs)
1973 return dma_map_single_attrs(dev->dma_device, cpu_addr, size,
1977 static inline void ib_dma_unmap_single_attrs(struct ib_device *dev,
1978 u64 addr, size_t size,
1979 enum dma_data_direction direction,
1980 struct dma_attrs *attrs)
1982 return dma_unmap_single_attrs(dev->dma_device, addr, size,
1987 * ib_dma_map_page - Map a physical page to DMA address
1988 * @dev: The device for which the dma_addr is to be created
1989 * @page: The page to be mapped
1990 * @offset: The offset within the page
1991 * @size: The size of the region in bytes
1992 * @direction: The direction of the DMA
1994 static inline u64 ib_dma_map_page(struct ib_device *dev,
1996 unsigned long offset,
1998 enum dma_data_direction direction)
2001 return dev->dma_ops->map_page(dev, page, offset, size, direction);
2002 return dma_map_page(dev->dma_device, page, offset, size, direction);
2006 * ib_dma_unmap_page - Destroy a mapping created by ib_dma_map_page()
2007 * @dev: The device for which the DMA address was created
2008 * @addr: The DMA address
2009 * @size: The size of the region in bytes
2010 * @direction: The direction of the DMA
2012 static inline void ib_dma_unmap_page(struct ib_device *dev,
2013 u64 addr, size_t size,
2014 enum dma_data_direction direction)
2017 dev->dma_ops->unmap_page(dev, addr, size, direction);
2019 dma_unmap_page(dev->dma_device, addr, size, direction);
2023 * ib_dma_map_sg - Map a scatter/gather list to DMA addresses
2024 * @dev: The device for which the DMA addresses are to be created
2025 * @sg: The array of scatter/gather entries
2026 * @nents: The number of scatter/gather entries
2027 * @direction: The direction of the DMA
2029 static inline int ib_dma_map_sg(struct ib_device *dev,
2030 struct scatterlist *sg, int nents,
2031 enum dma_data_direction direction)
2034 return dev->dma_ops->map_sg(dev, sg, nents, direction);
2035 return dma_map_sg(dev->dma_device, sg, nents, direction);
2039 * ib_dma_unmap_sg - Unmap a scatter/gather list of DMA addresses
2040 * @dev: The device for which the DMA addresses were created
2041 * @sg: The array of scatter/gather entries
2042 * @nents: The number of scatter/gather entries
2043 * @direction: The direction of the DMA
2045 static inline void ib_dma_unmap_sg(struct ib_device *dev,
2046 struct scatterlist *sg, int nents,
2047 enum dma_data_direction direction)
2050 dev->dma_ops->unmap_sg(dev, sg, nents, direction);
2052 dma_unmap_sg(dev->dma_device, sg, nents, direction);
2055 static inline int ib_dma_map_sg_attrs(struct ib_device *dev,
2056 struct scatterlist *sg, int nents,
2057 enum dma_data_direction direction,
2058 struct dma_attrs *attrs)
2060 return dma_map_sg_attrs(dev->dma_device, sg, nents, direction, attrs);
2063 static inline void ib_dma_unmap_sg_attrs(struct ib_device *dev,
2064 struct scatterlist *sg, int nents,
2065 enum dma_data_direction direction,
2066 struct dma_attrs *attrs)
2068 dma_unmap_sg_attrs(dev->dma_device, sg, nents, direction, attrs);
2071 * ib_sg_dma_address - Return the DMA address from a scatter/gather entry
2072 * @dev: The device for which the DMA addresses were created
2073 * @sg: The scatter/gather entry
2075 static inline u64 ib_sg_dma_address(struct ib_device *dev,
2076 struct scatterlist *sg)
2079 return dev->dma_ops->dma_address(dev, sg);
2080 return sg_dma_address(sg);
2084 * ib_sg_dma_len - Return the DMA length from a scatter/gather entry
2085 * @dev: The device for which the DMA addresses were created
2086 * @sg: The scatter/gather entry
2088 static inline unsigned int ib_sg_dma_len(struct ib_device *dev,
2089 struct scatterlist *sg)
2092 return dev->dma_ops->dma_len(dev, sg);
2093 return sg_dma_len(sg);
2097 * ib_dma_sync_single_for_cpu - Prepare DMA region to be accessed by CPU
2098 * @dev: The device for which the DMA address was created
2099 * @addr: The DMA address
2100 * @size: The size of the region in bytes
2101 * @dir: The direction of the DMA
2103 static inline void ib_dma_sync_single_for_cpu(struct ib_device *dev,
2106 enum dma_data_direction dir)
2109 dev->dma_ops->sync_single_for_cpu(dev, addr, size, dir);
2111 dma_sync_single_for_cpu(dev->dma_device, addr, size, dir);
2115 * ib_dma_sync_single_for_device - Prepare DMA region to be accessed by device
2116 * @dev: The device for which the DMA address was created
2117 * @addr: The DMA address
2118 * @size: The size of the region in bytes
2119 * @dir: The direction of the DMA
2121 static inline void ib_dma_sync_single_for_device(struct ib_device *dev,
2124 enum dma_data_direction dir)
2127 dev->dma_ops->sync_single_for_device(dev, addr, size, dir);
2129 dma_sync_single_for_device(dev->dma_device, addr, size, dir);
2133 * ib_dma_alloc_coherent - Allocate memory and map it for DMA
2134 * @dev: The device for which the DMA address is requested
2135 * @size: The size of the region to allocate in bytes
2136 * @dma_handle: A pointer for returning the DMA address of the region
2137 * @flag: memory allocator flags
2139 static inline void *ib_dma_alloc_coherent(struct ib_device *dev,
2145 return dev->dma_ops->alloc_coherent(dev, size, dma_handle, flag);
2150 ret = dma_alloc_coherent(dev->dma_device, size, &handle, flag);
2151 *dma_handle = handle;
2157 * ib_dma_free_coherent - Free memory allocated by ib_dma_alloc_coherent()
2158 * @dev: The device for which the DMA addresses were allocated
2159 * @size: The size of the region
2160 * @cpu_addr: the address returned by ib_dma_alloc_coherent()
2161 * @dma_handle: the DMA address returned by ib_dma_alloc_coherent()
2163 static inline void ib_dma_free_coherent(struct ib_device *dev,
2164 size_t size, void *cpu_addr,
2168 dev->dma_ops->free_coherent(dev, size, cpu_addr, dma_handle);
2170 dma_free_coherent(dev->dma_device, size, cpu_addr, dma_handle);
2174 * ib_reg_phys_mr - Prepares a virtually addressed memory region for use
2176 * @pd: The protection domain associated assigned to the registered region.
2177 * @phys_buf_array: Specifies a list of physical buffers to use in the
2179 * @num_phys_buf: Specifies the size of the phys_buf_array.
2180 * @mr_access_flags: Specifies the memory access rights.
2181 * @iova_start: The offset of the region's starting I/O virtual address.
2183 struct ib_mr *ib_reg_phys_mr(struct ib_pd *pd,
2184 struct ib_phys_buf *phys_buf_array,
2186 int mr_access_flags,
2190 * ib_rereg_phys_mr - Modifies the attributes of an existing memory region.
2191 * Conceptually, this call performs the functions deregister memory region
2192 * followed by register physical memory region. Where possible,
2193 * resources are reused instead of deallocated and reallocated.
2194 * @mr: The memory region to modify.
2195 * @mr_rereg_mask: A bit-mask used to indicate which of the following
2196 * properties of the memory region are being modified.
2197 * @pd: If %IB_MR_REREG_PD is set in mr_rereg_mask, this field specifies
2198 * the new protection domain to associated with the memory region,
2199 * otherwise, this parameter is ignored.
2200 * @phys_buf_array: If %IB_MR_REREG_TRANS is set in mr_rereg_mask, this
2201 * field specifies a list of physical buffers to use in the new
2202 * translation, otherwise, this parameter is ignored.
2203 * @num_phys_buf: If %IB_MR_REREG_TRANS is set in mr_rereg_mask, this
2204 * field specifies the size of the phys_buf_array, otherwise, this
2205 * parameter is ignored.
2206 * @mr_access_flags: If %IB_MR_REREG_ACCESS is set in mr_rereg_mask, this
2207 * field specifies the new memory access rights, otherwise, this
2208 * parameter is ignored.
2209 * @iova_start: The offset of the region's starting I/O virtual address.
2211 int ib_rereg_phys_mr(struct ib_mr *mr,
2214 struct ib_phys_buf *phys_buf_array,
2216 int mr_access_flags,
2220 * ib_query_mr - Retrieves information about a specific memory region.
2221 * @mr: The memory region to retrieve information about.
2222 * @mr_attr: The attributes of the specified memory region.
2224 int ib_query_mr(struct ib_mr *mr, struct ib_mr_attr *mr_attr);
2227 * ib_dereg_mr - Deregisters a memory region and removes it from the
2228 * HCA translation table.
2229 * @mr: The memory region to deregister.
2231 * This function can fail, if the memory region has memory windows bound to it.
2233 int ib_dereg_mr(struct ib_mr *mr);
2236 * ib_alloc_fast_reg_mr - Allocates memory region usable with the
2237 * IB_WR_FAST_REG_MR send work request.
2238 * @pd: The protection domain associated with the region.
2239 * @max_page_list_len: requested max physical buffer list length to be
2240 * used with fast register work requests for this MR.
2242 struct ib_mr *ib_alloc_fast_reg_mr(struct ib_pd *pd, int max_page_list_len);
2245 * ib_alloc_fast_reg_page_list - Allocates a page list array
2246 * @device - ib device pointer.
2247 * @page_list_len - size of the page list array to be allocated.
2249 * This allocates and returns a struct ib_fast_reg_page_list * and a
2250 * page_list array that is at least page_list_len in size. The actual
2251 * size is returned in max_page_list_len. The caller is responsible
2252 * for initializing the contents of the page_list array before posting
2253 * a send work request with the IB_WC_FAST_REG_MR opcode.
2255 * The page_list array entries must be translated using one of the
2256 * ib_dma_*() functions just like the addresses passed to
2257 * ib_map_phys_fmr(). Once the ib_post_send() is issued, the struct
2258 * ib_fast_reg_page_list must not be modified by the caller until the
2259 * IB_WC_FAST_REG_MR work request completes.
2261 struct ib_fast_reg_page_list *ib_alloc_fast_reg_page_list(
2262 struct ib_device *device, int page_list_len);
2265 * ib_free_fast_reg_page_list - Deallocates a previously allocated
2267 * @page_list - struct ib_fast_reg_page_list pointer to be deallocated.
2269 void ib_free_fast_reg_page_list(struct ib_fast_reg_page_list *page_list);
2272 * ib_update_fast_reg_key - updates the key portion of the fast_reg MR
2274 * @mr - struct ib_mr pointer to be updated.
2275 * @newkey - new key to be used.
2277 static inline void ib_update_fast_reg_key(struct ib_mr *mr, u8 newkey)
2279 mr->lkey = (mr->lkey & 0xffffff00) | newkey;
2280 mr->rkey = (mr->rkey & 0xffffff00) | newkey;
2284 * ib_inc_rkey - increments the key portion of the given rkey. Can be used
2285 * for calculating a new rkey for type 2 memory windows.
2286 * @rkey - the rkey to increment.
2288 static inline u32 ib_inc_rkey(u32 rkey)
2290 const u32 mask = 0x000000ff;
2291 return ((rkey + 1) & mask) | (rkey & ~mask);
2295 * ib_alloc_mw - Allocates a memory window.
2296 * @pd: The protection domain associated with the memory window.
2297 * @type: The type of the memory window (1 or 2).
2299 struct ib_mw *ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type);
2302 * ib_bind_mw - Posts a work request to the send queue of the specified
2303 * QP, which binds the memory window to the given address range and
2304 * remote access attributes.
2305 * @qp: QP to post the bind work request on.
2306 * @mw: The memory window to bind.
2307 * @mw_bind: Specifies information about the memory window, including
2308 * its address range, remote access rights, and associated memory region.
2310 * If there is no immediate error, the function will update the rkey member
2311 * of the mw parameter to its new value. The bind operation can still fail
2314 static inline int ib_bind_mw(struct ib_qp *qp,
2316 struct ib_mw_bind *mw_bind)
2318 /* XXX reference counting in corresponding MR? */
2319 return mw->device->bind_mw ?
2320 mw->device->bind_mw(qp, mw, mw_bind) :
2325 * ib_dealloc_mw - Deallocates a memory window.
2326 * @mw: The memory window to deallocate.
2328 int ib_dealloc_mw(struct ib_mw *mw);
2331 * ib_alloc_fmr - Allocates a unmapped fast memory region.
2332 * @pd: The protection domain associated with the unmapped region.
2333 * @mr_access_flags: Specifies the memory access rights.
2334 * @fmr_attr: Attributes of the unmapped region.
2336 * A fast memory region must be mapped before it can be used as part of
2339 struct ib_fmr *ib_alloc_fmr(struct ib_pd *pd,
2340 int mr_access_flags,
2341 struct ib_fmr_attr *fmr_attr);
2344 * ib_map_phys_fmr - Maps a list of physical pages to a fast memory region.
2345 * @fmr: The fast memory region to associate with the pages.
2346 * @page_list: An array of physical pages to map to the fast memory region.
2347 * @list_len: The number of pages in page_list.
2348 * @iova: The I/O virtual address to use with the mapped region.
2350 static inline int ib_map_phys_fmr(struct ib_fmr *fmr,
2351 u64 *page_list, int list_len,
2354 return fmr->device->map_phys_fmr(fmr, page_list, list_len, iova);
2358 * ib_unmap_fmr - Removes the mapping from a list of fast memory regions.
2359 * @fmr_list: A linked list of fast memory regions to unmap.
2361 int ib_unmap_fmr(struct list_head *fmr_list);
2364 * ib_dealloc_fmr - Deallocates a fast memory region.
2365 * @fmr: The fast memory region to deallocate.
2367 int ib_dealloc_fmr(struct ib_fmr *fmr);
2370 * ib_attach_mcast - Attaches the specified QP to a multicast group.
2371 * @qp: QP to attach to the multicast group. The QP must be type
2373 * @gid: Multicast group GID.
2374 * @lid: Multicast group LID in host byte order.
2376 * In order to send and receive multicast packets, subnet
2377 * administration must have created the multicast group and configured
2378 * the fabric appropriately. The port associated with the specified
2379 * QP must also be a member of the multicast group.
2381 int ib_attach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid);
2384 * ib_detach_mcast - Detaches the specified QP from a multicast group.
2385 * @qp: QP to detach from the multicast group.
2386 * @gid: Multicast group GID.
2387 * @lid: Multicast group LID in host byte order.
2389 int ib_detach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid);
2392 * ib_alloc_xrcd - Allocates an XRC domain.
2393 * @device: The device on which to allocate the XRC domain.
2395 struct ib_xrcd *ib_alloc_xrcd(struct ib_device *device);
2398 * ib_dealloc_xrcd - Deallocates an XRC domain.
2399 * @xrcd: The XRC domain to deallocate.
2401 int ib_dealloc_xrcd(struct ib_xrcd *xrcd);
2403 struct ib_flow *ib_create_flow(struct ib_qp *qp,
2404 struct ib_flow_attr *flow_attr, int domain);
2405 int ib_destroy_flow(struct ib_flow *flow_id);
2407 static inline int ib_check_mr_access(int flags)
2410 * Local write permission is required if remote write or
2411 * remote atomic permission is also requested.
2413 if (flags & (IB_ACCESS_REMOTE_ATOMIC | IB_ACCESS_REMOTE_WRITE) &&
2414 !(flags & IB_ACCESS_LOCAL_WRITE))
2420 #endif /* IB_VERBS_H */