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. */
76 enum rdma_transport_type {
80 RDMA_TRANSPORT_USNIC_UDP
83 enum rdma_transport_type
84 rdma_node_get_transport(enum rdma_node_type node_type) __attribute_const__;
86 enum rdma_link_layer {
87 IB_LINK_LAYER_UNSPECIFIED,
88 IB_LINK_LAYER_INFINIBAND,
89 IB_LINK_LAYER_ETHERNET,
92 enum ib_device_cap_flags {
93 IB_DEVICE_RESIZE_MAX_WR = 1,
94 IB_DEVICE_BAD_PKEY_CNTR = (1<<1),
95 IB_DEVICE_BAD_QKEY_CNTR = (1<<2),
96 IB_DEVICE_RAW_MULTI = (1<<3),
97 IB_DEVICE_AUTO_PATH_MIG = (1<<4),
98 IB_DEVICE_CHANGE_PHY_PORT = (1<<5),
99 IB_DEVICE_UD_AV_PORT_ENFORCE = (1<<6),
100 IB_DEVICE_CURR_QP_STATE_MOD = (1<<7),
101 IB_DEVICE_SHUTDOWN_PORT = (1<<8),
102 IB_DEVICE_INIT_TYPE = (1<<9),
103 IB_DEVICE_PORT_ACTIVE_EVENT = (1<<10),
104 IB_DEVICE_SYS_IMAGE_GUID = (1<<11),
105 IB_DEVICE_RC_RNR_NAK_GEN = (1<<12),
106 IB_DEVICE_SRQ_RESIZE = (1<<13),
107 IB_DEVICE_N_NOTIFY_CQ = (1<<14),
108 IB_DEVICE_LOCAL_DMA_LKEY = (1<<15),
109 IB_DEVICE_RESERVED = (1<<16), /* old SEND_W_INV */
110 IB_DEVICE_MEM_WINDOW = (1<<17),
112 * Devices should set IB_DEVICE_UD_IP_SUM if they support
113 * insertion of UDP and TCP checksum on outgoing UD IPoIB
114 * messages and can verify the validity of checksum for
115 * incoming messages. Setting this flag implies that the
116 * IPoIB driver may set NETIF_F_IP_CSUM for datagram mode.
118 IB_DEVICE_UD_IP_CSUM = (1<<18),
119 IB_DEVICE_UD_TSO = (1<<19),
120 IB_DEVICE_XRC = (1<<20),
121 IB_DEVICE_MEM_MGT_EXTENSIONS = (1<<21),
122 IB_DEVICE_BLOCK_MULTICAST_LOOPBACK = (1<<22),
123 IB_DEVICE_MEM_WINDOW_TYPE_2A = (1<<23),
124 IB_DEVICE_MEM_WINDOW_TYPE_2B = (1<<24),
125 IB_DEVICE_MANAGED_FLOW_STEERING = (1<<29)
134 struct ib_device_attr {
136 __be64 sys_image_guid;
144 int device_cap_flags;
154 int max_qp_init_rd_atom;
155 int max_ee_init_rd_atom;
156 enum ib_atomic_cap atomic_cap;
157 enum ib_atomic_cap masked_atomic_cap;
164 int max_mcast_qp_attach;
165 int max_total_mcast_qp_attach;
172 unsigned int max_fast_reg_page_list_len;
174 u8 local_ca_ack_delay;
185 static inline int ib_mtu_enum_to_int(enum ib_mtu mtu)
188 case IB_MTU_256: return 256;
189 case IB_MTU_512: return 512;
190 case IB_MTU_1024: return 1024;
191 case IB_MTU_2048: return 2048;
192 case IB_MTU_4096: return 4096;
203 IB_PORT_ACTIVE_DEFER = 5
206 enum ib_port_cap_flags {
208 IB_PORT_NOTICE_SUP = 1 << 2,
209 IB_PORT_TRAP_SUP = 1 << 3,
210 IB_PORT_OPT_IPD_SUP = 1 << 4,
211 IB_PORT_AUTO_MIGR_SUP = 1 << 5,
212 IB_PORT_SL_MAP_SUP = 1 << 6,
213 IB_PORT_MKEY_NVRAM = 1 << 7,
214 IB_PORT_PKEY_NVRAM = 1 << 8,
215 IB_PORT_LED_INFO_SUP = 1 << 9,
216 IB_PORT_SM_DISABLED = 1 << 10,
217 IB_PORT_SYS_IMAGE_GUID_SUP = 1 << 11,
218 IB_PORT_PKEY_SW_EXT_PORT_TRAP_SUP = 1 << 12,
219 IB_PORT_EXTENDED_SPEEDS_SUP = 1 << 14,
220 IB_PORT_CM_SUP = 1 << 16,
221 IB_PORT_SNMP_TUNNEL_SUP = 1 << 17,
222 IB_PORT_REINIT_SUP = 1 << 18,
223 IB_PORT_DEVICE_MGMT_SUP = 1 << 19,
224 IB_PORT_VENDOR_CLASS_SUP = 1 << 20,
225 IB_PORT_DR_NOTICE_SUP = 1 << 21,
226 IB_PORT_CAP_MASK_NOTICE_SUP = 1 << 22,
227 IB_PORT_BOOT_MGMT_SUP = 1 << 23,
228 IB_PORT_LINK_LATENCY_SUP = 1 << 24,
229 IB_PORT_CLIENT_REG_SUP = 1 << 25,
230 IB_PORT_IP_BASED_GIDS = 1 << 26
240 static inline int ib_width_enum_to_int(enum ib_port_width width)
243 case IB_WIDTH_1X: return 1;
244 case IB_WIDTH_4X: return 4;
245 case IB_WIDTH_8X: return 8;
246 case IB_WIDTH_12X: return 12;
260 struct ib_protocol_stats {
264 struct iw_protocol_stats {
267 u64 ipInTooBigErrors;
270 u64 ipInUnknownProtos;
271 u64 ipInTruncatedPkts;
274 u64 ipOutForwDatagrams;
306 union rdma_protocol_stats {
307 struct ib_protocol_stats ib;
308 struct iw_protocol_stats iw;
311 struct ib_port_attr {
312 enum ib_port_state state;
314 enum ib_mtu active_mtu;
333 enum ib_device_modify_flags {
334 IB_DEVICE_MODIFY_SYS_IMAGE_GUID = 1 << 0,
335 IB_DEVICE_MODIFY_NODE_DESC = 1 << 1
338 struct ib_device_modify {
343 enum ib_port_modify_flags {
344 IB_PORT_SHUTDOWN = 1,
345 IB_PORT_INIT_TYPE = (1<<2),
346 IB_PORT_RESET_QKEY_CNTR = (1<<3)
349 struct ib_port_modify {
350 u32 set_port_cap_mask;
351 u32 clr_port_cap_mask;
359 IB_EVENT_QP_ACCESS_ERR,
363 IB_EVENT_PATH_MIG_ERR,
364 IB_EVENT_DEVICE_FATAL,
365 IB_EVENT_PORT_ACTIVE,
368 IB_EVENT_PKEY_CHANGE,
371 IB_EVENT_SRQ_LIMIT_REACHED,
372 IB_EVENT_QP_LAST_WQE_REACHED,
373 IB_EVENT_CLIENT_REREGISTER,
378 struct ib_device *device;
385 enum ib_event_type event;
388 struct ib_event_handler {
389 struct ib_device *device;
390 void (*handler)(struct ib_event_handler *, struct ib_event *);
391 struct list_head list;
394 #define INIT_IB_EVENT_HANDLER(_ptr, _device, _handler) \
396 (_ptr)->device = _device; \
397 (_ptr)->handler = _handler; \
398 INIT_LIST_HEAD(&(_ptr)->list); \
401 struct ib_global_route {
410 __be32 version_tclass_flow;
419 IB_MULTICAST_QPN = 0xffffff
422 #define IB_LID_PERMISSIVE cpu_to_be16(0xFFFF)
429 IB_RATE_PORT_CURRENT = 0,
430 IB_RATE_2_5_GBPS = 2,
438 IB_RATE_120_GBPS = 10,
439 IB_RATE_14_GBPS = 11,
440 IB_RATE_56_GBPS = 12,
441 IB_RATE_112_GBPS = 13,
442 IB_RATE_168_GBPS = 14,
443 IB_RATE_25_GBPS = 15,
444 IB_RATE_100_GBPS = 16,
445 IB_RATE_200_GBPS = 17,
446 IB_RATE_300_GBPS = 18
450 * ib_rate_to_mult - Convert the IB rate enum to a multiple of the
451 * base rate of 2.5 Gbit/sec. For example, IB_RATE_5_GBPS will be
452 * converted to 2, since 5 Gbit/sec is 2 * 2.5 Gbit/sec.
453 * @rate: rate to convert.
455 int ib_rate_to_mult(enum ib_rate rate) __attribute_const__;
458 * ib_rate_to_mbps - Convert the IB rate enum to Mbps.
459 * For example, IB_RATE_2_5_GBPS will be converted to 2500.
460 * @rate: rate to convert.
462 int ib_rate_to_mbps(enum ib_rate rate) __attribute_const__;
465 * mult_to_ib_rate - Convert a multiple of 2.5 Gbit/sec to an IB rate
467 * @mult: multiple to convert.
469 enum ib_rate mult_to_ib_rate(int mult) __attribute_const__;
472 struct ib_global_route grh;
487 IB_WC_LOC_EEC_OP_ERR,
492 IB_WC_LOC_ACCESS_ERR,
493 IB_WC_REM_INV_REQ_ERR,
494 IB_WC_REM_ACCESS_ERR,
497 IB_WC_RNR_RETRY_EXC_ERR,
498 IB_WC_LOC_RDD_VIOL_ERR,
499 IB_WC_REM_INV_RD_REQ_ERR,
502 IB_WC_INV_EEC_STATE_ERR,
504 IB_WC_RESP_TIMEOUT_ERR,
518 IB_WC_MASKED_COMP_SWAP,
519 IB_WC_MASKED_FETCH_ADD,
521 * Set value of IB_WC_RECV so consumers can test if a completion is a
522 * receive by testing (opcode & IB_WC_RECV).
525 IB_WC_RECV_RDMA_WITH_IMM
530 IB_WC_WITH_IMM = (1<<1),
531 IB_WC_WITH_INVALIDATE = (1<<2),
532 IB_WC_IP_CSUM_OK = (1<<3),
533 IB_WC_WITH_SMAC = (1<<4),
534 IB_WC_WITH_VLAN = (1<<5),
539 enum ib_wc_status status;
540 enum ib_wc_opcode opcode;
554 u8 port_num; /* valid only for DR SMPs on switches */
559 enum ib_cq_notify_flags {
560 IB_CQ_SOLICITED = 1 << 0,
561 IB_CQ_NEXT_COMP = 1 << 1,
562 IB_CQ_SOLICITED_MASK = IB_CQ_SOLICITED | IB_CQ_NEXT_COMP,
563 IB_CQ_REPORT_MISSED_EVENTS = 1 << 2,
571 enum ib_srq_attr_mask {
572 IB_SRQ_MAX_WR = 1 << 0,
573 IB_SRQ_LIMIT = 1 << 1,
582 struct ib_srq_init_attr {
583 void (*event_handler)(struct ib_event *, void *);
585 struct ib_srq_attr attr;
586 enum ib_srq_type srq_type;
590 struct ib_xrcd *xrcd;
611 * IB_QPT_SMI and IB_QPT_GSI have to be the first two entries
612 * here (and in that order) since the MAD layer uses them as
613 * indices into a 2-entry table.
622 IB_QPT_RAW_ETHERTYPE,
623 IB_QPT_RAW_PACKET = 8,
627 /* Reserve a range for qp types internal to the low level driver.
628 * These qp types will not be visible at the IB core layer, so the
629 * IB_QPT_MAX usages should not be affected in the core layer
631 IB_QPT_RESERVED1 = 0x1000,
643 enum ib_qp_create_flags {
644 IB_QP_CREATE_IPOIB_UD_LSO = 1 << 0,
645 IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK = 1 << 1,
646 IB_QP_CREATE_NETIF_QP = 1 << 5,
647 /* reserve bits 26-31 for low level drivers' internal use */
648 IB_QP_CREATE_RESERVED_START = 1 << 26,
649 IB_QP_CREATE_RESERVED_END = 1 << 31,
654 * Note: users may not call ib_close_qp or ib_destroy_qp from the event_handler
655 * callback to destroy the passed in QP.
658 struct ib_qp_init_attr {
659 void (*event_handler)(struct ib_event *, void *);
661 struct ib_cq *send_cq;
662 struct ib_cq *recv_cq;
664 struct ib_xrcd *xrcd; /* XRC TGT QPs only */
665 struct ib_qp_cap cap;
666 enum ib_sig_type sq_sig_type;
667 enum ib_qp_type qp_type;
668 enum ib_qp_create_flags create_flags;
669 u8 port_num; /* special QP types only */
672 struct ib_qp_open_attr {
673 void (*event_handler)(struct ib_event *, void *);
676 enum ib_qp_type qp_type;
679 enum ib_rnr_timeout {
680 IB_RNR_TIMER_655_36 = 0,
681 IB_RNR_TIMER_000_01 = 1,
682 IB_RNR_TIMER_000_02 = 2,
683 IB_RNR_TIMER_000_03 = 3,
684 IB_RNR_TIMER_000_04 = 4,
685 IB_RNR_TIMER_000_06 = 5,
686 IB_RNR_TIMER_000_08 = 6,
687 IB_RNR_TIMER_000_12 = 7,
688 IB_RNR_TIMER_000_16 = 8,
689 IB_RNR_TIMER_000_24 = 9,
690 IB_RNR_TIMER_000_32 = 10,
691 IB_RNR_TIMER_000_48 = 11,
692 IB_RNR_TIMER_000_64 = 12,
693 IB_RNR_TIMER_000_96 = 13,
694 IB_RNR_TIMER_001_28 = 14,
695 IB_RNR_TIMER_001_92 = 15,
696 IB_RNR_TIMER_002_56 = 16,
697 IB_RNR_TIMER_003_84 = 17,
698 IB_RNR_TIMER_005_12 = 18,
699 IB_RNR_TIMER_007_68 = 19,
700 IB_RNR_TIMER_010_24 = 20,
701 IB_RNR_TIMER_015_36 = 21,
702 IB_RNR_TIMER_020_48 = 22,
703 IB_RNR_TIMER_030_72 = 23,
704 IB_RNR_TIMER_040_96 = 24,
705 IB_RNR_TIMER_061_44 = 25,
706 IB_RNR_TIMER_081_92 = 26,
707 IB_RNR_TIMER_122_88 = 27,
708 IB_RNR_TIMER_163_84 = 28,
709 IB_RNR_TIMER_245_76 = 29,
710 IB_RNR_TIMER_327_68 = 30,
711 IB_RNR_TIMER_491_52 = 31
714 enum ib_qp_attr_mask {
716 IB_QP_CUR_STATE = (1<<1),
717 IB_QP_EN_SQD_ASYNC_NOTIFY = (1<<2),
718 IB_QP_ACCESS_FLAGS = (1<<3),
719 IB_QP_PKEY_INDEX = (1<<4),
723 IB_QP_PATH_MTU = (1<<8),
724 IB_QP_TIMEOUT = (1<<9),
725 IB_QP_RETRY_CNT = (1<<10),
726 IB_QP_RNR_RETRY = (1<<11),
727 IB_QP_RQ_PSN = (1<<12),
728 IB_QP_MAX_QP_RD_ATOMIC = (1<<13),
729 IB_QP_ALT_PATH = (1<<14),
730 IB_QP_MIN_RNR_TIMER = (1<<15),
731 IB_QP_SQ_PSN = (1<<16),
732 IB_QP_MAX_DEST_RD_ATOMIC = (1<<17),
733 IB_QP_PATH_MIG_STATE = (1<<18),
735 IB_QP_DEST_QPN = (1<<20),
736 IB_QP_SMAC = (1<<21),
737 IB_QP_ALT_SMAC = (1<<22),
739 IB_QP_ALT_VID = (1<<24),
764 enum ib_qp_state qp_state;
765 enum ib_qp_state cur_qp_state;
766 enum ib_mtu path_mtu;
767 enum ib_mig_state path_mig_state;
773 struct ib_qp_cap cap;
774 struct ib_ah_attr ah_attr;
775 struct ib_ah_attr alt_ah_attr;
778 u8 en_sqd_async_notify;
781 u8 max_dest_rd_atomic;
790 u8 alt_smac[ETH_ALEN];
797 IB_WR_RDMA_WRITE_WITH_IMM,
801 IB_WR_ATOMIC_CMP_AND_SWP,
802 IB_WR_ATOMIC_FETCH_AND_ADD,
805 IB_WR_RDMA_READ_WITH_INV,
808 IB_WR_MASKED_ATOMIC_CMP_AND_SWP,
809 IB_WR_MASKED_ATOMIC_FETCH_AND_ADD,
811 /* reserve values for low level drivers' internal use.
812 * These values will not be used at all in the ib core layer.
814 IB_WR_RESERVED1 = 0xf0,
828 IB_SEND_SIGNALED = (1<<1),
829 IB_SEND_SOLICITED = (1<<2),
830 IB_SEND_INLINE = (1<<3),
831 IB_SEND_IP_CSUM = (1<<4),
833 /* reserve bits 26-31 for low level drivers' internal use */
834 IB_SEND_RESERVED_START = (1 << 26),
835 IB_SEND_RESERVED_END = (1 << 31),
844 struct ib_fast_reg_page_list {
845 struct ib_device *device;
847 unsigned int max_page_list_len;
851 * struct ib_mw_bind_info - Parameters for a memory window bind operation.
852 * @mr: A memory region to bind the memory window to.
853 * @addr: The address where the memory window should begin.
854 * @length: The length of the memory window, in bytes.
855 * @mw_access_flags: Access flags from enum ib_access_flags for the window.
857 * This struct contains the shared parameters for type 1 and type 2
858 * memory window bind operations.
860 struct ib_mw_bind_info {
868 struct ib_send_wr *next;
870 struct ib_sge *sg_list;
872 enum ib_wr_opcode opcode;
887 u64 compare_add_mask;
898 u16 pkey_index; /* valid for GSI only */
899 u8 port_num; /* valid for DR SMPs on switch only */
903 struct ib_fast_reg_page_list *page_list;
904 unsigned int page_shift;
905 unsigned int page_list_len;
912 /* The new rkey for the memory window. */
914 struct ib_mw_bind_info bind_info;
917 u32 xrc_remote_srq_num; /* XRC TGT QPs only */
921 struct ib_recv_wr *next;
923 struct ib_sge *sg_list;
927 enum ib_access_flags {
928 IB_ACCESS_LOCAL_WRITE = 1,
929 IB_ACCESS_REMOTE_WRITE = (1<<1),
930 IB_ACCESS_REMOTE_READ = (1<<2),
931 IB_ACCESS_REMOTE_ATOMIC = (1<<3),
932 IB_ACCESS_MW_BIND = (1<<4),
933 IB_ZERO_BASED = (1<<5)
943 u64 device_virt_addr;
950 enum ib_mr_rereg_flags {
951 IB_MR_REREG_TRANS = 1,
952 IB_MR_REREG_PD = (1<<1),
953 IB_MR_REREG_ACCESS = (1<<2)
957 * struct ib_mw_bind - Parameters for a type 1 memory window bind operation.
958 * @wr_id: Work request id.
959 * @send_flags: Flags from ib_send_flags enum.
960 * @bind_info: More parameters of the bind operation.
965 struct ib_mw_bind_info bind_info;
975 struct ib_device *device;
976 struct list_head pd_list;
977 struct list_head mr_list;
978 struct list_head mw_list;
979 struct list_head cq_list;
980 struct list_head qp_list;
981 struct list_head srq_list;
982 struct list_head ah_list;
983 struct list_head xrcd_list;
984 struct list_head rule_list;
989 u64 user_handle; /* handle given to us by userspace */
990 struct ib_ucontext *context; /* associated user context */
991 void *object; /* containing object */
992 struct list_head list; /* link to context's list */
993 int id; /* index into kernel idr */
995 struct rw_semaphore mutex; /* protects .live */
1000 const void __user *inbuf;
1001 void __user *outbuf;
1007 struct ib_device *device;
1008 struct ib_uobject *uobject;
1009 atomic_t usecnt; /* count all resources */
1013 struct ib_device *device;
1014 atomic_t usecnt; /* count all exposed resources */
1015 struct inode *inode;
1017 struct mutex tgt_qp_mutex;
1018 struct list_head tgt_qp_list;
1022 struct ib_device *device;
1024 struct ib_uobject *uobject;
1027 typedef void (*ib_comp_handler)(struct ib_cq *cq, void *cq_context);
1030 struct ib_device *device;
1031 struct ib_uobject *uobject;
1032 ib_comp_handler comp_handler;
1033 void (*event_handler)(struct ib_event *, void *);
1036 atomic_t usecnt; /* count number of work queues */
1040 struct ib_device *device;
1042 struct ib_uobject *uobject;
1043 void (*event_handler)(struct ib_event *, void *);
1045 enum ib_srq_type srq_type;
1050 struct ib_xrcd *xrcd;
1058 struct ib_device *device;
1060 struct ib_cq *send_cq;
1061 struct ib_cq *recv_cq;
1063 struct ib_xrcd *xrcd; /* XRC TGT QPs only */
1064 struct list_head xrcd_list;
1065 /* count times opened, mcast attaches, flow attaches */
1067 struct list_head open_list;
1068 struct ib_qp *real_qp;
1069 struct ib_uobject *uobject;
1070 void (*event_handler)(struct ib_event *, void *);
1073 enum ib_qp_type qp_type;
1077 struct ib_device *device;
1079 struct ib_uobject *uobject;
1082 atomic_t usecnt; /* count number of MWs */
1086 struct ib_device *device;
1088 struct ib_uobject *uobject;
1090 enum ib_mw_type type;
1094 struct ib_device *device;
1096 struct list_head list;
1101 /* Supported steering options */
1102 enum ib_flow_attr_type {
1103 /* steering according to rule specifications */
1104 IB_FLOW_ATTR_NORMAL = 0x0,
1105 /* default unicast and multicast rule -
1106 * receive all Eth traffic which isn't steered to any QP
1108 IB_FLOW_ATTR_ALL_DEFAULT = 0x1,
1109 /* default multicast rule -
1110 * receive all Eth multicast traffic which isn't steered to any QP
1112 IB_FLOW_ATTR_MC_DEFAULT = 0x2,
1113 /* sniffer rule - receive all port traffic */
1114 IB_FLOW_ATTR_SNIFFER = 0x3
1117 /* Supported steering header types */
1118 enum ib_flow_spec_type {
1120 IB_FLOW_SPEC_ETH = 0x20,
1121 IB_FLOW_SPEC_IB = 0x22,
1123 IB_FLOW_SPEC_IPV4 = 0x30,
1125 IB_FLOW_SPEC_TCP = 0x40,
1126 IB_FLOW_SPEC_UDP = 0x41
1128 #define IB_FLOW_SPEC_LAYER_MASK 0xF0
1129 #define IB_FLOW_SPEC_SUPPORT_LAYERS 4
1131 /* Flow steering rule priority is set according to it's domain.
1132 * Lower domain value means higher priority.
1134 enum ib_flow_domain {
1135 IB_FLOW_DOMAIN_USER,
1136 IB_FLOW_DOMAIN_ETHTOOL,
1139 IB_FLOW_DOMAIN_NUM /* Must be last */
1142 struct ib_flow_eth_filter {
1149 struct ib_flow_spec_eth {
1150 enum ib_flow_spec_type type;
1152 struct ib_flow_eth_filter val;
1153 struct ib_flow_eth_filter mask;
1156 struct ib_flow_ib_filter {
1161 struct ib_flow_spec_ib {
1162 enum ib_flow_spec_type type;
1164 struct ib_flow_ib_filter val;
1165 struct ib_flow_ib_filter mask;
1168 struct ib_flow_ipv4_filter {
1173 struct ib_flow_spec_ipv4 {
1174 enum ib_flow_spec_type type;
1176 struct ib_flow_ipv4_filter val;
1177 struct ib_flow_ipv4_filter mask;
1180 struct ib_flow_tcp_udp_filter {
1185 struct ib_flow_spec_tcp_udp {
1186 enum ib_flow_spec_type type;
1188 struct ib_flow_tcp_udp_filter val;
1189 struct ib_flow_tcp_udp_filter mask;
1192 union ib_flow_spec {
1194 enum ib_flow_spec_type type;
1197 struct ib_flow_spec_eth eth;
1198 struct ib_flow_spec_ib ib;
1199 struct ib_flow_spec_ipv4 ipv4;
1200 struct ib_flow_spec_tcp_udp tcp_udp;
1203 struct ib_flow_attr {
1204 enum ib_flow_attr_type type;
1210 /* Following are the optional layers according to user request
1211 * struct ib_flow_spec_xxx
1212 * struct ib_flow_spec_yyy
1218 struct ib_uobject *uobject;
1224 enum ib_process_mad_flags {
1225 IB_MAD_IGNORE_MKEY = 1,
1226 IB_MAD_IGNORE_BKEY = 2,
1227 IB_MAD_IGNORE_ALL = IB_MAD_IGNORE_MKEY | IB_MAD_IGNORE_BKEY
1230 enum ib_mad_result {
1231 IB_MAD_RESULT_FAILURE = 0, /* (!SUCCESS is the important flag) */
1232 IB_MAD_RESULT_SUCCESS = 1 << 0, /* MAD was successfully processed */
1233 IB_MAD_RESULT_REPLY = 1 << 1, /* Reply packet needs to be sent */
1234 IB_MAD_RESULT_CONSUMED = 1 << 2 /* Packet consumed: stop processing */
1237 #define IB_DEVICE_NAME_MAX 64
1241 struct ib_event_handler event_handler;
1242 struct ib_pkey_cache **pkey_cache;
1243 struct ib_gid_cache **gid_cache;
1247 struct ib_dma_mapping_ops {
1248 int (*mapping_error)(struct ib_device *dev,
1250 u64 (*map_single)(struct ib_device *dev,
1251 void *ptr, size_t size,
1252 enum dma_data_direction direction);
1253 void (*unmap_single)(struct ib_device *dev,
1254 u64 addr, size_t size,
1255 enum dma_data_direction direction);
1256 u64 (*map_page)(struct ib_device *dev,
1257 struct page *page, unsigned long offset,
1259 enum dma_data_direction direction);
1260 void (*unmap_page)(struct ib_device *dev,
1261 u64 addr, size_t size,
1262 enum dma_data_direction direction);
1263 int (*map_sg)(struct ib_device *dev,
1264 struct scatterlist *sg, int nents,
1265 enum dma_data_direction direction);
1266 void (*unmap_sg)(struct ib_device *dev,
1267 struct scatterlist *sg, int nents,
1268 enum dma_data_direction direction);
1269 void (*sync_single_for_cpu)(struct ib_device *dev,
1272 enum dma_data_direction dir);
1273 void (*sync_single_for_device)(struct ib_device *dev,
1276 enum dma_data_direction dir);
1277 void *(*alloc_coherent)(struct ib_device *dev,
1281 void (*free_coherent)(struct ib_device *dev,
1282 size_t size, void *cpu_addr,
1289 struct device *dma_device;
1291 char name[IB_DEVICE_NAME_MAX];
1293 struct list_head event_handler_list;
1294 spinlock_t event_handler_lock;
1296 spinlock_t client_data_lock;
1297 struct list_head core_list;
1298 struct list_head client_data_list;
1300 struct ib_cache cache;
1304 int num_comp_vectors;
1306 struct iw_cm_verbs *iwcm;
1308 int (*get_protocol_stats)(struct ib_device *device,
1309 union rdma_protocol_stats *stats);
1310 int (*query_device)(struct ib_device *device,
1311 struct ib_device_attr *device_attr);
1312 int (*query_port)(struct ib_device *device,
1314 struct ib_port_attr *port_attr);
1315 enum rdma_link_layer (*get_link_layer)(struct ib_device *device,
1317 int (*query_gid)(struct ib_device *device,
1318 u8 port_num, int index,
1320 int (*query_pkey)(struct ib_device *device,
1321 u8 port_num, u16 index, u16 *pkey);
1322 int (*modify_device)(struct ib_device *device,
1323 int device_modify_mask,
1324 struct ib_device_modify *device_modify);
1325 int (*modify_port)(struct ib_device *device,
1326 u8 port_num, int port_modify_mask,
1327 struct ib_port_modify *port_modify);
1328 struct ib_ucontext * (*alloc_ucontext)(struct ib_device *device,
1329 struct ib_udata *udata);
1330 int (*dealloc_ucontext)(struct ib_ucontext *context);
1331 int (*mmap)(struct ib_ucontext *context,
1332 struct vm_area_struct *vma);
1333 struct ib_pd * (*alloc_pd)(struct ib_device *device,
1334 struct ib_ucontext *context,
1335 struct ib_udata *udata);
1336 int (*dealloc_pd)(struct ib_pd *pd);
1337 struct ib_ah * (*create_ah)(struct ib_pd *pd,
1338 struct ib_ah_attr *ah_attr);
1339 int (*modify_ah)(struct ib_ah *ah,
1340 struct ib_ah_attr *ah_attr);
1341 int (*query_ah)(struct ib_ah *ah,
1342 struct ib_ah_attr *ah_attr);
1343 int (*destroy_ah)(struct ib_ah *ah);
1344 struct ib_srq * (*create_srq)(struct ib_pd *pd,
1345 struct ib_srq_init_attr *srq_init_attr,
1346 struct ib_udata *udata);
1347 int (*modify_srq)(struct ib_srq *srq,
1348 struct ib_srq_attr *srq_attr,
1349 enum ib_srq_attr_mask srq_attr_mask,
1350 struct ib_udata *udata);
1351 int (*query_srq)(struct ib_srq *srq,
1352 struct ib_srq_attr *srq_attr);
1353 int (*destroy_srq)(struct ib_srq *srq);
1354 int (*post_srq_recv)(struct ib_srq *srq,
1355 struct ib_recv_wr *recv_wr,
1356 struct ib_recv_wr **bad_recv_wr);
1357 struct ib_qp * (*create_qp)(struct ib_pd *pd,
1358 struct ib_qp_init_attr *qp_init_attr,
1359 struct ib_udata *udata);
1360 int (*modify_qp)(struct ib_qp *qp,
1361 struct ib_qp_attr *qp_attr,
1363 struct ib_udata *udata);
1364 int (*query_qp)(struct ib_qp *qp,
1365 struct ib_qp_attr *qp_attr,
1367 struct ib_qp_init_attr *qp_init_attr);
1368 int (*destroy_qp)(struct ib_qp *qp);
1369 int (*post_send)(struct ib_qp *qp,
1370 struct ib_send_wr *send_wr,
1371 struct ib_send_wr **bad_send_wr);
1372 int (*post_recv)(struct ib_qp *qp,
1373 struct ib_recv_wr *recv_wr,
1374 struct ib_recv_wr **bad_recv_wr);
1375 struct ib_cq * (*create_cq)(struct ib_device *device, int cqe,
1377 struct ib_ucontext *context,
1378 struct ib_udata *udata);
1379 int (*modify_cq)(struct ib_cq *cq, u16 cq_count,
1381 int (*destroy_cq)(struct ib_cq *cq);
1382 int (*resize_cq)(struct ib_cq *cq, int cqe,
1383 struct ib_udata *udata);
1384 int (*poll_cq)(struct ib_cq *cq, int num_entries,
1386 int (*peek_cq)(struct ib_cq *cq, int wc_cnt);
1387 int (*req_notify_cq)(struct ib_cq *cq,
1388 enum ib_cq_notify_flags flags);
1389 int (*req_ncomp_notif)(struct ib_cq *cq,
1391 struct ib_mr * (*get_dma_mr)(struct ib_pd *pd,
1392 int mr_access_flags);
1393 struct ib_mr * (*reg_phys_mr)(struct ib_pd *pd,
1394 struct ib_phys_buf *phys_buf_array,
1396 int mr_access_flags,
1398 struct ib_mr * (*reg_user_mr)(struct ib_pd *pd,
1399 u64 start, u64 length,
1401 int mr_access_flags,
1402 struct ib_udata *udata);
1403 int (*query_mr)(struct ib_mr *mr,
1404 struct ib_mr_attr *mr_attr);
1405 int (*dereg_mr)(struct ib_mr *mr);
1406 struct ib_mr * (*alloc_fast_reg_mr)(struct ib_pd *pd,
1407 int max_page_list_len);
1408 struct ib_fast_reg_page_list * (*alloc_fast_reg_page_list)(struct ib_device *device,
1410 void (*free_fast_reg_page_list)(struct ib_fast_reg_page_list *page_list);
1411 int (*rereg_phys_mr)(struct ib_mr *mr,
1414 struct ib_phys_buf *phys_buf_array,
1416 int mr_access_flags,
1418 struct ib_mw * (*alloc_mw)(struct ib_pd *pd,
1419 enum ib_mw_type type);
1420 int (*bind_mw)(struct ib_qp *qp,
1422 struct ib_mw_bind *mw_bind);
1423 int (*dealloc_mw)(struct ib_mw *mw);
1424 struct ib_fmr * (*alloc_fmr)(struct ib_pd *pd,
1425 int mr_access_flags,
1426 struct ib_fmr_attr *fmr_attr);
1427 int (*map_phys_fmr)(struct ib_fmr *fmr,
1428 u64 *page_list, int list_len,
1430 int (*unmap_fmr)(struct list_head *fmr_list);
1431 int (*dealloc_fmr)(struct ib_fmr *fmr);
1432 int (*attach_mcast)(struct ib_qp *qp,
1435 int (*detach_mcast)(struct ib_qp *qp,
1438 int (*process_mad)(struct ib_device *device,
1439 int process_mad_flags,
1441 struct ib_wc *in_wc,
1442 struct ib_grh *in_grh,
1443 struct ib_mad *in_mad,
1444 struct ib_mad *out_mad);
1445 struct ib_xrcd * (*alloc_xrcd)(struct ib_device *device,
1446 struct ib_ucontext *ucontext,
1447 struct ib_udata *udata);
1448 int (*dealloc_xrcd)(struct ib_xrcd *xrcd);
1449 struct ib_flow * (*create_flow)(struct ib_qp *qp,
1453 int (*destroy_flow)(struct ib_flow *flow_id);
1455 struct ib_dma_mapping_ops *dma_ops;
1457 struct module *owner;
1459 struct kobject *ports_parent;
1460 struct list_head port_list;
1463 IB_DEV_UNINITIALIZED,
1469 u64 uverbs_cmd_mask;
1470 u64 uverbs_ex_cmd_mask;
1481 void (*add) (struct ib_device *);
1482 void (*remove)(struct ib_device *);
1484 struct list_head list;
1487 struct ib_device *ib_alloc_device(size_t size);
1488 void ib_dealloc_device(struct ib_device *device);
1490 int ib_register_device(struct ib_device *device,
1491 int (*port_callback)(struct ib_device *,
1492 u8, struct kobject *));
1493 void ib_unregister_device(struct ib_device *device);
1495 int ib_register_client (struct ib_client *client);
1496 void ib_unregister_client(struct ib_client *client);
1498 void *ib_get_client_data(struct ib_device *device, struct ib_client *client);
1499 void ib_set_client_data(struct ib_device *device, struct ib_client *client,
1502 static inline int ib_copy_from_udata(void *dest, struct ib_udata *udata, size_t len)
1504 return copy_from_user(dest, udata->inbuf, len) ? -EFAULT : 0;
1507 static inline int ib_copy_to_udata(struct ib_udata *udata, void *src, size_t len)
1509 return copy_to_user(udata->outbuf, src, len) ? -EFAULT : 0;
1513 * ib_modify_qp_is_ok - Check that the supplied attribute mask
1514 * contains all required attributes and no attributes not allowed for
1515 * the given QP state transition.
1516 * @cur_state: Current QP state
1517 * @next_state: Next QP state
1519 * @mask: Mask of supplied QP attributes
1520 * @ll : link layer of port
1522 * This function is a helper function that a low-level driver's
1523 * modify_qp method can use to validate the consumer's input. It
1524 * checks that cur_state and next_state are valid QP states, that a
1525 * transition from cur_state to next_state is allowed by the IB spec,
1526 * and that the attribute mask supplied is allowed for the transition.
1528 int ib_modify_qp_is_ok(enum ib_qp_state cur_state, enum ib_qp_state next_state,
1529 enum ib_qp_type type, enum ib_qp_attr_mask mask,
1530 enum rdma_link_layer ll);
1532 int ib_register_event_handler (struct ib_event_handler *event_handler);
1533 int ib_unregister_event_handler(struct ib_event_handler *event_handler);
1534 void ib_dispatch_event(struct ib_event *event);
1536 int ib_query_device(struct ib_device *device,
1537 struct ib_device_attr *device_attr);
1539 int ib_query_port(struct ib_device *device,
1540 u8 port_num, struct ib_port_attr *port_attr);
1542 enum rdma_link_layer rdma_port_get_link_layer(struct ib_device *device,
1545 int ib_query_gid(struct ib_device *device,
1546 u8 port_num, int index, union ib_gid *gid);
1548 int ib_query_pkey(struct ib_device *device,
1549 u8 port_num, u16 index, u16 *pkey);
1551 int ib_modify_device(struct ib_device *device,
1552 int device_modify_mask,
1553 struct ib_device_modify *device_modify);
1555 int ib_modify_port(struct ib_device *device,
1556 u8 port_num, int port_modify_mask,
1557 struct ib_port_modify *port_modify);
1559 int ib_find_gid(struct ib_device *device, union ib_gid *gid,
1560 u8 *port_num, u16 *index);
1562 int ib_find_pkey(struct ib_device *device,
1563 u8 port_num, u16 pkey, u16 *index);
1566 * ib_alloc_pd - Allocates an unused protection domain.
1567 * @device: The device on which to allocate the protection domain.
1569 * A protection domain object provides an association between QPs, shared
1570 * receive queues, address handles, memory regions, and memory windows.
1572 struct ib_pd *ib_alloc_pd(struct ib_device *device);
1575 * ib_dealloc_pd - Deallocates a protection domain.
1576 * @pd: The protection domain to deallocate.
1578 int ib_dealloc_pd(struct ib_pd *pd);
1581 * ib_create_ah - Creates an address handle for the given address vector.
1582 * @pd: The protection domain associated with the address handle.
1583 * @ah_attr: The attributes of the address vector.
1585 * The address handle is used to reference a local or global destination
1586 * in all UD QP post sends.
1588 struct ib_ah *ib_create_ah(struct ib_pd *pd, struct ib_ah_attr *ah_attr);
1591 * ib_init_ah_from_wc - Initializes address handle attributes from a
1593 * @device: Device on which the received message arrived.
1594 * @port_num: Port on which the received message arrived.
1595 * @wc: Work completion associated with the received message.
1596 * @grh: References the received global route header. This parameter is
1597 * ignored unless the work completion indicates that the GRH is valid.
1598 * @ah_attr: Returned attributes that can be used when creating an address
1599 * handle for replying to the message.
1601 int ib_init_ah_from_wc(struct ib_device *device, u8 port_num, struct ib_wc *wc,
1602 struct ib_grh *grh, struct ib_ah_attr *ah_attr);
1605 * ib_create_ah_from_wc - Creates an address handle associated with the
1606 * sender of the specified work completion.
1607 * @pd: The protection domain associated with the address handle.
1608 * @wc: Work completion information associated with a received message.
1609 * @grh: References the received global route header. This parameter is
1610 * ignored unless the work completion indicates that the GRH is valid.
1611 * @port_num: The outbound port number to associate with the address.
1613 * The address handle is used to reference a local or global destination
1614 * in all UD QP post sends.
1616 struct ib_ah *ib_create_ah_from_wc(struct ib_pd *pd, struct ib_wc *wc,
1617 struct ib_grh *grh, u8 port_num);
1620 * ib_modify_ah - Modifies the address vector associated with an address
1622 * @ah: The address handle to modify.
1623 * @ah_attr: The new address vector attributes to associate with the
1626 int ib_modify_ah(struct ib_ah *ah, struct ib_ah_attr *ah_attr);
1629 * ib_query_ah - Queries the address vector associated with an address
1631 * @ah: The address handle to query.
1632 * @ah_attr: The address vector attributes associated with the address
1635 int ib_query_ah(struct ib_ah *ah, struct ib_ah_attr *ah_attr);
1638 * ib_destroy_ah - Destroys an address handle.
1639 * @ah: The address handle to destroy.
1641 int ib_destroy_ah(struct ib_ah *ah);
1644 * ib_create_srq - Creates a SRQ associated with the specified protection
1646 * @pd: The protection domain associated with the SRQ.
1647 * @srq_init_attr: A list of initial attributes required to create the
1648 * SRQ. If SRQ creation succeeds, then the attributes are updated to
1649 * the actual capabilities of the created SRQ.
1651 * srq_attr->max_wr and srq_attr->max_sge are read the determine the
1652 * requested size of the SRQ, and set to the actual values allocated
1653 * on return. If ib_create_srq() succeeds, then max_wr and max_sge
1654 * will always be at least as large as the requested values.
1656 struct ib_srq *ib_create_srq(struct ib_pd *pd,
1657 struct ib_srq_init_attr *srq_init_attr);
1660 * ib_modify_srq - Modifies the attributes for the specified SRQ.
1661 * @srq: The SRQ to modify.
1662 * @srq_attr: On input, specifies the SRQ attributes to modify. On output,
1663 * the current values of selected SRQ attributes are returned.
1664 * @srq_attr_mask: A bit-mask used to specify which attributes of the SRQ
1665 * are being modified.
1667 * The mask may contain IB_SRQ_MAX_WR to resize the SRQ and/or
1668 * IB_SRQ_LIMIT to set the SRQ's limit and request notification when
1669 * the number of receives queued drops below the limit.
1671 int ib_modify_srq(struct ib_srq *srq,
1672 struct ib_srq_attr *srq_attr,
1673 enum ib_srq_attr_mask srq_attr_mask);
1676 * ib_query_srq - Returns the attribute list and current values for the
1678 * @srq: The SRQ to query.
1679 * @srq_attr: The attributes of the specified SRQ.
1681 int ib_query_srq(struct ib_srq *srq,
1682 struct ib_srq_attr *srq_attr);
1685 * ib_destroy_srq - Destroys the specified SRQ.
1686 * @srq: The SRQ to destroy.
1688 int ib_destroy_srq(struct ib_srq *srq);
1691 * ib_post_srq_recv - Posts a list of work requests to the specified SRQ.
1692 * @srq: The SRQ to post the work request on.
1693 * @recv_wr: A list of work requests to post on the receive queue.
1694 * @bad_recv_wr: On an immediate failure, this parameter will reference
1695 * the work request that failed to be posted on the QP.
1697 static inline int ib_post_srq_recv(struct ib_srq *srq,
1698 struct ib_recv_wr *recv_wr,
1699 struct ib_recv_wr **bad_recv_wr)
1701 return srq->device->post_srq_recv(srq, recv_wr, bad_recv_wr);
1705 * ib_create_qp - Creates a QP associated with the specified protection
1707 * @pd: The protection domain associated with the QP.
1708 * @qp_init_attr: A list of initial attributes required to create the
1709 * QP. If QP creation succeeds, then the attributes are updated to
1710 * the actual capabilities of the created QP.
1712 struct ib_qp *ib_create_qp(struct ib_pd *pd,
1713 struct ib_qp_init_attr *qp_init_attr);
1716 * ib_modify_qp - Modifies the attributes for the specified QP and then
1717 * transitions the QP to the given state.
1718 * @qp: The QP to modify.
1719 * @qp_attr: On input, specifies the QP attributes to modify. On output,
1720 * the current values of selected QP attributes are returned.
1721 * @qp_attr_mask: A bit-mask used to specify which attributes of the QP
1722 * are being modified.
1724 int ib_modify_qp(struct ib_qp *qp,
1725 struct ib_qp_attr *qp_attr,
1729 * ib_query_qp - Returns the attribute list and current values for the
1731 * @qp: The QP to query.
1732 * @qp_attr: The attributes of the specified QP.
1733 * @qp_attr_mask: A bit-mask used to select specific attributes to query.
1734 * @qp_init_attr: Additional attributes of the selected QP.
1736 * The qp_attr_mask may be used to limit the query to gathering only the
1737 * selected attributes.
1739 int ib_query_qp(struct ib_qp *qp,
1740 struct ib_qp_attr *qp_attr,
1742 struct ib_qp_init_attr *qp_init_attr);
1745 * ib_destroy_qp - Destroys the specified QP.
1746 * @qp: The QP to destroy.
1748 int ib_destroy_qp(struct ib_qp *qp);
1751 * ib_open_qp - Obtain a reference to an existing sharable QP.
1752 * @xrcd - XRC domain
1753 * @qp_open_attr: Attributes identifying the QP to open.
1755 * Returns a reference to a sharable QP.
1757 struct ib_qp *ib_open_qp(struct ib_xrcd *xrcd,
1758 struct ib_qp_open_attr *qp_open_attr);
1761 * ib_close_qp - Release an external reference to a QP.
1762 * @qp: The QP handle to release
1764 * The opened QP handle is released by the caller. The underlying
1765 * shared QP is not destroyed until all internal references are released.
1767 int ib_close_qp(struct ib_qp *qp);
1770 * ib_post_send - Posts a list of work requests to the send queue of
1772 * @qp: The QP to post the work request on.
1773 * @send_wr: A list of work requests to post on the send queue.
1774 * @bad_send_wr: On an immediate failure, this parameter will reference
1775 * the work request that failed to be posted on the QP.
1777 * While IBA Vol. 1 section 11.4.1.1 specifies that if an immediate
1778 * error is returned, the QP state shall not be affected,
1779 * ib_post_send() will return an immediate error after queueing any
1780 * earlier work requests in the list.
1782 static inline int ib_post_send(struct ib_qp *qp,
1783 struct ib_send_wr *send_wr,
1784 struct ib_send_wr **bad_send_wr)
1786 return qp->device->post_send(qp, send_wr, bad_send_wr);
1790 * ib_post_recv - Posts a list of work requests to the receive queue of
1792 * @qp: The QP to post the work request on.
1793 * @recv_wr: A list of work requests to post on the receive queue.
1794 * @bad_recv_wr: On an immediate failure, this parameter will reference
1795 * the work request that failed to be posted on the QP.
1797 static inline int ib_post_recv(struct ib_qp *qp,
1798 struct ib_recv_wr *recv_wr,
1799 struct ib_recv_wr **bad_recv_wr)
1801 return qp->device->post_recv(qp, recv_wr, bad_recv_wr);
1805 * ib_create_cq - Creates a CQ on the specified device.
1806 * @device: The device on which to create the CQ.
1807 * @comp_handler: A user-specified callback that is invoked when a
1808 * completion event occurs on the CQ.
1809 * @event_handler: A user-specified callback that is invoked when an
1810 * asynchronous event not associated with a completion occurs on the CQ.
1811 * @cq_context: Context associated with the CQ returned to the user via
1812 * the associated completion and event handlers.
1813 * @cqe: The minimum size of the CQ.
1814 * @comp_vector - Completion vector used to signal completion events.
1815 * Must be >= 0 and < context->num_comp_vectors.
1817 * Users can examine the cq structure to determine the actual CQ size.
1819 struct ib_cq *ib_create_cq(struct ib_device *device,
1820 ib_comp_handler comp_handler,
1821 void (*event_handler)(struct ib_event *, void *),
1822 void *cq_context, int cqe, int comp_vector);
1825 * ib_resize_cq - Modifies the capacity of the CQ.
1826 * @cq: The CQ to resize.
1827 * @cqe: The minimum size of the CQ.
1829 * Users can examine the cq structure to determine the actual CQ size.
1831 int ib_resize_cq(struct ib_cq *cq, int cqe);
1834 * ib_modify_cq - Modifies moderation params of the CQ
1835 * @cq: The CQ to modify.
1836 * @cq_count: number of CQEs that will trigger an event
1837 * @cq_period: max period of time in usec before triggering an event
1840 int ib_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period);
1843 * ib_destroy_cq - Destroys the specified CQ.
1844 * @cq: The CQ to destroy.
1846 int ib_destroy_cq(struct ib_cq *cq);
1849 * ib_poll_cq - poll a CQ for completion(s)
1850 * @cq:the CQ being polled
1851 * @num_entries:maximum number of completions to return
1852 * @wc:array of at least @num_entries &struct ib_wc where completions
1855 * Poll a CQ for (possibly multiple) completions. If the return value
1856 * is < 0, an error occurred. If the return value is >= 0, it is the
1857 * number of completions returned. If the return value is
1858 * non-negative and < num_entries, then the CQ was emptied.
1860 static inline int ib_poll_cq(struct ib_cq *cq, int num_entries,
1863 return cq->device->poll_cq(cq, num_entries, wc);
1867 * ib_peek_cq - Returns the number of unreaped completions currently
1868 * on the specified CQ.
1869 * @cq: The CQ to peek.
1870 * @wc_cnt: A minimum number of unreaped completions to check for.
1872 * If the number of unreaped completions is greater than or equal to wc_cnt,
1873 * this function returns wc_cnt, otherwise, it returns the actual number of
1874 * unreaped completions.
1876 int ib_peek_cq(struct ib_cq *cq, int wc_cnt);
1879 * ib_req_notify_cq - Request completion notification on a CQ.
1880 * @cq: The CQ to generate an event for.
1882 * Must contain exactly one of %IB_CQ_SOLICITED or %IB_CQ_NEXT_COMP
1883 * to request an event on the next solicited event or next work
1884 * completion at any type, respectively. %IB_CQ_REPORT_MISSED_EVENTS
1885 * may also be |ed in to request a hint about missed events, as
1889 * < 0 means an error occurred while requesting notification
1890 * == 0 means notification was requested successfully, and if
1891 * IB_CQ_REPORT_MISSED_EVENTS was passed in, then no events
1892 * were missed and it is safe to wait for another event. In
1893 * this case is it guaranteed that any work completions added
1894 * to the CQ since the last CQ poll will trigger a completion
1895 * notification event.
1896 * > 0 is only returned if IB_CQ_REPORT_MISSED_EVENTS was passed
1897 * in. It means that the consumer must poll the CQ again to
1898 * make sure it is empty to avoid missing an event because of a
1899 * race between requesting notification and an entry being
1900 * added to the CQ. This return value means it is possible
1901 * (but not guaranteed) that a work completion has been added
1902 * to the CQ since the last poll without triggering a
1903 * completion notification event.
1905 static inline int ib_req_notify_cq(struct ib_cq *cq,
1906 enum ib_cq_notify_flags flags)
1908 return cq->device->req_notify_cq(cq, flags);
1912 * ib_req_ncomp_notif - Request completion notification when there are
1913 * at least the specified number of unreaped completions on the CQ.
1914 * @cq: The CQ to generate an event for.
1915 * @wc_cnt: The number of unreaped completions that should be on the
1916 * CQ before an event is generated.
1918 static inline int ib_req_ncomp_notif(struct ib_cq *cq, int wc_cnt)
1920 return cq->device->req_ncomp_notif ?
1921 cq->device->req_ncomp_notif(cq, wc_cnt) :
1926 * ib_get_dma_mr - Returns a memory region for system memory that is
1928 * @pd: The protection domain associated with the memory region.
1929 * @mr_access_flags: Specifies the memory access rights.
1931 * Note that the ib_dma_*() functions defined below must be used
1932 * to create/destroy addresses used with the Lkey or Rkey returned
1933 * by ib_get_dma_mr().
1935 struct ib_mr *ib_get_dma_mr(struct ib_pd *pd, int mr_access_flags);
1938 * ib_dma_mapping_error - check a DMA addr for error
1939 * @dev: The device for which the dma_addr was created
1940 * @dma_addr: The DMA address to check
1942 static inline int ib_dma_mapping_error(struct ib_device *dev, u64 dma_addr)
1945 return dev->dma_ops->mapping_error(dev, dma_addr);
1946 return dma_mapping_error(dev->dma_device, dma_addr);
1950 * ib_dma_map_single - Map a kernel virtual address to DMA address
1951 * @dev: The device for which the dma_addr is to be created
1952 * @cpu_addr: The kernel virtual address
1953 * @size: The size of the region in bytes
1954 * @direction: The direction of the DMA
1956 static inline u64 ib_dma_map_single(struct ib_device *dev,
1957 void *cpu_addr, size_t size,
1958 enum dma_data_direction direction)
1961 return dev->dma_ops->map_single(dev, cpu_addr, size, direction);
1962 return dma_map_single(dev->dma_device, cpu_addr, size, direction);
1966 * ib_dma_unmap_single - Destroy a mapping created by ib_dma_map_single()
1967 * @dev: The device for which the DMA address was created
1968 * @addr: The DMA address
1969 * @size: The size of the region in bytes
1970 * @direction: The direction of the DMA
1972 static inline void ib_dma_unmap_single(struct ib_device *dev,
1973 u64 addr, size_t size,
1974 enum dma_data_direction direction)
1977 dev->dma_ops->unmap_single(dev, addr, size, direction);
1979 dma_unmap_single(dev->dma_device, addr, size, direction);
1982 static inline u64 ib_dma_map_single_attrs(struct ib_device *dev,
1983 void *cpu_addr, size_t size,
1984 enum dma_data_direction direction,
1985 struct dma_attrs *attrs)
1987 return dma_map_single_attrs(dev->dma_device, cpu_addr, size,
1991 static inline void ib_dma_unmap_single_attrs(struct ib_device *dev,
1992 u64 addr, size_t size,
1993 enum dma_data_direction direction,
1994 struct dma_attrs *attrs)
1996 return dma_unmap_single_attrs(dev->dma_device, addr, size,
2001 * ib_dma_map_page - Map a physical page to DMA address
2002 * @dev: The device for which the dma_addr is to be created
2003 * @page: The page to be mapped
2004 * @offset: The offset within the page
2005 * @size: The size of the region in bytes
2006 * @direction: The direction of the DMA
2008 static inline u64 ib_dma_map_page(struct ib_device *dev,
2010 unsigned long offset,
2012 enum dma_data_direction direction)
2015 return dev->dma_ops->map_page(dev, page, offset, size, direction);
2016 return dma_map_page(dev->dma_device, page, offset, size, direction);
2020 * ib_dma_unmap_page - Destroy a mapping created by ib_dma_map_page()
2021 * @dev: The device for which the DMA address was created
2022 * @addr: The DMA address
2023 * @size: The size of the region in bytes
2024 * @direction: The direction of the DMA
2026 static inline void ib_dma_unmap_page(struct ib_device *dev,
2027 u64 addr, size_t size,
2028 enum dma_data_direction direction)
2031 dev->dma_ops->unmap_page(dev, addr, size, direction);
2033 dma_unmap_page(dev->dma_device, addr, size, direction);
2037 * ib_dma_map_sg - Map a scatter/gather list to DMA addresses
2038 * @dev: The device for which the DMA addresses are to be created
2039 * @sg: The array of scatter/gather entries
2040 * @nents: The number of scatter/gather entries
2041 * @direction: The direction of the DMA
2043 static inline int ib_dma_map_sg(struct ib_device *dev,
2044 struct scatterlist *sg, int nents,
2045 enum dma_data_direction direction)
2048 return dev->dma_ops->map_sg(dev, sg, nents, direction);
2049 return dma_map_sg(dev->dma_device, sg, nents, direction);
2053 * ib_dma_unmap_sg - Unmap a scatter/gather list of DMA addresses
2054 * @dev: The device for which the DMA addresses were created
2055 * @sg: The array of scatter/gather entries
2056 * @nents: The number of scatter/gather entries
2057 * @direction: The direction of the DMA
2059 static inline void ib_dma_unmap_sg(struct ib_device *dev,
2060 struct scatterlist *sg, int nents,
2061 enum dma_data_direction direction)
2064 dev->dma_ops->unmap_sg(dev, sg, nents, direction);
2066 dma_unmap_sg(dev->dma_device, sg, nents, direction);
2069 static inline int ib_dma_map_sg_attrs(struct ib_device *dev,
2070 struct scatterlist *sg, int nents,
2071 enum dma_data_direction direction,
2072 struct dma_attrs *attrs)
2074 return dma_map_sg_attrs(dev->dma_device, sg, nents, direction, attrs);
2077 static inline void ib_dma_unmap_sg_attrs(struct ib_device *dev,
2078 struct scatterlist *sg, int nents,
2079 enum dma_data_direction direction,
2080 struct dma_attrs *attrs)
2082 dma_unmap_sg_attrs(dev->dma_device, sg, nents, direction, attrs);
2085 * ib_sg_dma_address - Return the DMA address from a scatter/gather entry
2086 * @dev: The device for which the DMA addresses were created
2087 * @sg: The scatter/gather entry
2089 * Note: this function is obsolete. To do: change all occurrences of
2090 * ib_sg_dma_address() into sg_dma_address().
2092 static inline u64 ib_sg_dma_address(struct ib_device *dev,
2093 struct scatterlist *sg)
2095 return sg_dma_address(sg);
2099 * ib_sg_dma_len - Return the DMA length from a scatter/gather entry
2100 * @dev: The device for which the DMA addresses were created
2101 * @sg: The scatter/gather entry
2103 * Note: this function is obsolete. To do: change all occurrences of
2104 * ib_sg_dma_len() into sg_dma_len().
2106 static inline unsigned int ib_sg_dma_len(struct ib_device *dev,
2107 struct scatterlist *sg)
2109 return sg_dma_len(sg);
2113 * ib_dma_sync_single_for_cpu - Prepare DMA region to be accessed by CPU
2114 * @dev: The device for which the DMA address was created
2115 * @addr: The DMA address
2116 * @size: The size of the region in bytes
2117 * @dir: The direction of the DMA
2119 static inline void ib_dma_sync_single_for_cpu(struct ib_device *dev,
2122 enum dma_data_direction dir)
2125 dev->dma_ops->sync_single_for_cpu(dev, addr, size, dir);
2127 dma_sync_single_for_cpu(dev->dma_device, addr, size, dir);
2131 * ib_dma_sync_single_for_device - Prepare DMA region to be accessed by device
2132 * @dev: The device for which the DMA address was created
2133 * @addr: The DMA address
2134 * @size: The size of the region in bytes
2135 * @dir: The direction of the DMA
2137 static inline void ib_dma_sync_single_for_device(struct ib_device *dev,
2140 enum dma_data_direction dir)
2143 dev->dma_ops->sync_single_for_device(dev, addr, size, dir);
2145 dma_sync_single_for_device(dev->dma_device, addr, size, dir);
2149 * ib_dma_alloc_coherent - Allocate memory and map it for DMA
2150 * @dev: The device for which the DMA address is requested
2151 * @size: The size of the region to allocate in bytes
2152 * @dma_handle: A pointer for returning the DMA address of the region
2153 * @flag: memory allocator flags
2155 static inline void *ib_dma_alloc_coherent(struct ib_device *dev,
2161 return dev->dma_ops->alloc_coherent(dev, size, dma_handle, flag);
2166 ret = dma_alloc_coherent(dev->dma_device, size, &handle, flag);
2167 *dma_handle = handle;
2173 * ib_dma_free_coherent - Free memory allocated by ib_dma_alloc_coherent()
2174 * @dev: The device for which the DMA addresses were allocated
2175 * @size: The size of the region
2176 * @cpu_addr: the address returned by ib_dma_alloc_coherent()
2177 * @dma_handle: the DMA address returned by ib_dma_alloc_coherent()
2179 static inline void ib_dma_free_coherent(struct ib_device *dev,
2180 size_t size, void *cpu_addr,
2184 dev->dma_ops->free_coherent(dev, size, cpu_addr, dma_handle);
2186 dma_free_coherent(dev->dma_device, size, cpu_addr, dma_handle);
2190 * ib_reg_phys_mr - Prepares a virtually addressed memory region for use
2192 * @pd: The protection domain associated assigned to the registered region.
2193 * @phys_buf_array: Specifies a list of physical buffers to use in the
2195 * @num_phys_buf: Specifies the size of the phys_buf_array.
2196 * @mr_access_flags: Specifies the memory access rights.
2197 * @iova_start: The offset of the region's starting I/O virtual address.
2199 struct ib_mr *ib_reg_phys_mr(struct ib_pd *pd,
2200 struct ib_phys_buf *phys_buf_array,
2202 int mr_access_flags,
2206 * ib_rereg_phys_mr - Modifies the attributes of an existing memory region.
2207 * Conceptually, this call performs the functions deregister memory region
2208 * followed by register physical memory region. Where possible,
2209 * resources are reused instead of deallocated and reallocated.
2210 * @mr: The memory region to modify.
2211 * @mr_rereg_mask: A bit-mask used to indicate which of the following
2212 * properties of the memory region are being modified.
2213 * @pd: If %IB_MR_REREG_PD is set in mr_rereg_mask, this field specifies
2214 * the new protection domain to associated with the memory region,
2215 * otherwise, this parameter is ignored.
2216 * @phys_buf_array: If %IB_MR_REREG_TRANS is set in mr_rereg_mask, this
2217 * field specifies a list of physical buffers to use in the new
2218 * translation, otherwise, this parameter is ignored.
2219 * @num_phys_buf: If %IB_MR_REREG_TRANS is set in mr_rereg_mask, this
2220 * field specifies the size of the phys_buf_array, otherwise, this
2221 * parameter is ignored.
2222 * @mr_access_flags: If %IB_MR_REREG_ACCESS is set in mr_rereg_mask, this
2223 * field specifies the new memory access rights, otherwise, this
2224 * parameter is ignored.
2225 * @iova_start: The offset of the region's starting I/O virtual address.
2227 int ib_rereg_phys_mr(struct ib_mr *mr,
2230 struct ib_phys_buf *phys_buf_array,
2232 int mr_access_flags,
2236 * ib_query_mr - Retrieves information about a specific memory region.
2237 * @mr: The memory region to retrieve information about.
2238 * @mr_attr: The attributes of the specified memory region.
2240 int ib_query_mr(struct ib_mr *mr, struct ib_mr_attr *mr_attr);
2243 * ib_dereg_mr - Deregisters a memory region and removes it from the
2244 * HCA translation table.
2245 * @mr: The memory region to deregister.
2247 * This function can fail, if the memory region has memory windows bound to it.
2249 int ib_dereg_mr(struct ib_mr *mr);
2252 * ib_alloc_fast_reg_mr - Allocates memory region usable with the
2253 * IB_WR_FAST_REG_MR send work request.
2254 * @pd: The protection domain associated with the region.
2255 * @max_page_list_len: requested max physical buffer list length to be
2256 * used with fast register work requests for this MR.
2258 struct ib_mr *ib_alloc_fast_reg_mr(struct ib_pd *pd, int max_page_list_len);
2261 * ib_alloc_fast_reg_page_list - Allocates a page list array
2262 * @device - ib device pointer.
2263 * @page_list_len - size of the page list array to be allocated.
2265 * This allocates and returns a struct ib_fast_reg_page_list * and a
2266 * page_list array that is at least page_list_len in size. The actual
2267 * size is returned in max_page_list_len. The caller is responsible
2268 * for initializing the contents of the page_list array before posting
2269 * a send work request with the IB_WC_FAST_REG_MR opcode.
2271 * The page_list array entries must be translated using one of the
2272 * ib_dma_*() functions just like the addresses passed to
2273 * ib_map_phys_fmr(). Once the ib_post_send() is issued, the struct
2274 * ib_fast_reg_page_list must not be modified by the caller until the
2275 * IB_WC_FAST_REG_MR work request completes.
2277 struct ib_fast_reg_page_list *ib_alloc_fast_reg_page_list(
2278 struct ib_device *device, int page_list_len);
2281 * ib_free_fast_reg_page_list - Deallocates a previously allocated
2283 * @page_list - struct ib_fast_reg_page_list pointer to be deallocated.
2285 void ib_free_fast_reg_page_list(struct ib_fast_reg_page_list *page_list);
2288 * ib_update_fast_reg_key - updates the key portion of the fast_reg MR
2290 * @mr - struct ib_mr pointer to be updated.
2291 * @newkey - new key to be used.
2293 static inline void ib_update_fast_reg_key(struct ib_mr *mr, u8 newkey)
2295 mr->lkey = (mr->lkey & 0xffffff00) | newkey;
2296 mr->rkey = (mr->rkey & 0xffffff00) | newkey;
2300 * ib_inc_rkey - increments the key portion of the given rkey. Can be used
2301 * for calculating a new rkey for type 2 memory windows.
2302 * @rkey - the rkey to increment.
2304 static inline u32 ib_inc_rkey(u32 rkey)
2306 const u32 mask = 0x000000ff;
2307 return ((rkey + 1) & mask) | (rkey & ~mask);
2311 * ib_alloc_mw - Allocates a memory window.
2312 * @pd: The protection domain associated with the memory window.
2313 * @type: The type of the memory window (1 or 2).
2315 struct ib_mw *ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type);
2318 * ib_bind_mw - Posts a work request to the send queue of the specified
2319 * QP, which binds the memory window to the given address range and
2320 * remote access attributes.
2321 * @qp: QP to post the bind work request on.
2322 * @mw: The memory window to bind.
2323 * @mw_bind: Specifies information about the memory window, including
2324 * its address range, remote access rights, and associated memory region.
2326 * If there is no immediate error, the function will update the rkey member
2327 * of the mw parameter to its new value. The bind operation can still fail
2330 static inline int ib_bind_mw(struct ib_qp *qp,
2332 struct ib_mw_bind *mw_bind)
2334 /* XXX reference counting in corresponding MR? */
2335 return mw->device->bind_mw ?
2336 mw->device->bind_mw(qp, mw, mw_bind) :
2341 * ib_dealloc_mw - Deallocates a memory window.
2342 * @mw: The memory window to deallocate.
2344 int ib_dealloc_mw(struct ib_mw *mw);
2347 * ib_alloc_fmr - Allocates a unmapped fast memory region.
2348 * @pd: The protection domain associated with the unmapped region.
2349 * @mr_access_flags: Specifies the memory access rights.
2350 * @fmr_attr: Attributes of the unmapped region.
2352 * A fast memory region must be mapped before it can be used as part of
2355 struct ib_fmr *ib_alloc_fmr(struct ib_pd *pd,
2356 int mr_access_flags,
2357 struct ib_fmr_attr *fmr_attr);
2360 * ib_map_phys_fmr - Maps a list of physical pages to a fast memory region.
2361 * @fmr: The fast memory region to associate with the pages.
2362 * @page_list: An array of physical pages to map to the fast memory region.
2363 * @list_len: The number of pages in page_list.
2364 * @iova: The I/O virtual address to use with the mapped region.
2366 static inline int ib_map_phys_fmr(struct ib_fmr *fmr,
2367 u64 *page_list, int list_len,
2370 return fmr->device->map_phys_fmr(fmr, page_list, list_len, iova);
2374 * ib_unmap_fmr - Removes the mapping from a list of fast memory regions.
2375 * @fmr_list: A linked list of fast memory regions to unmap.
2377 int ib_unmap_fmr(struct list_head *fmr_list);
2380 * ib_dealloc_fmr - Deallocates a fast memory region.
2381 * @fmr: The fast memory region to deallocate.
2383 int ib_dealloc_fmr(struct ib_fmr *fmr);
2386 * ib_attach_mcast - Attaches the specified QP to a multicast group.
2387 * @qp: QP to attach to the multicast group. The QP must be type
2389 * @gid: Multicast group GID.
2390 * @lid: Multicast group LID in host byte order.
2392 * In order to send and receive multicast packets, subnet
2393 * administration must have created the multicast group and configured
2394 * the fabric appropriately. The port associated with the specified
2395 * QP must also be a member of the multicast group.
2397 int ib_attach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid);
2400 * ib_detach_mcast - Detaches the specified QP from a multicast group.
2401 * @qp: QP to detach from the multicast group.
2402 * @gid: Multicast group GID.
2403 * @lid: Multicast group LID in host byte order.
2405 int ib_detach_mcast(struct ib_qp *qp, union ib_gid *gid, u16 lid);
2408 * ib_alloc_xrcd - Allocates an XRC domain.
2409 * @device: The device on which to allocate the XRC domain.
2411 struct ib_xrcd *ib_alloc_xrcd(struct ib_device *device);
2414 * ib_dealloc_xrcd - Deallocates an XRC domain.
2415 * @xrcd: The XRC domain to deallocate.
2417 int ib_dealloc_xrcd(struct ib_xrcd *xrcd);
2419 struct ib_flow *ib_create_flow(struct ib_qp *qp,
2420 struct ib_flow_attr *flow_attr, int domain);
2421 int ib_destroy_flow(struct ib_flow *flow_id);
2423 static inline int ib_check_mr_access(int flags)
2426 * Local write permission is required if remote write or
2427 * remote atomic permission is also requested.
2429 if (flags & (IB_ACCESS_REMOTE_ATOMIC | IB_ACCESS_REMOTE_WRITE) &&
2430 !(flags & IB_ACCESS_LOCAL_WRITE))
2436 #endif /* IB_VERBS_H */