2 * Copyright (c) 2008 Oracle. All rights reserved.
4 * This software is available to you under a choice of one of two
5 * licenses. You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the
8 * OpenIB.org BSD license below:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
18 * - Redistributions in binary form must reproduce the above
19 * copyright notice, this list of conditions and the following
20 * disclaimer in the documentation and/or other materials
21 * provided with the distribution.
23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
37 #include <linux/types.h>
39 #define RDS_IB_ABI_VERSION 0x301
42 * setsockopt/getsockopt for SOL_RDS
44 #define RDS_CANCEL_SENT_TO 1
47 /* deprecated: RDS_BARRIER 4 */
49 #define RDS_CONG_MONITOR 6
50 #define RDS_GET_MR_FOR_DEST 7
53 * Control message types for SOL_RDS.
55 * CMSG_RDMA_ARGS (sendmsg)
56 * Request a RDMA transfer to/from the specified
58 * The cmsg_data is a struct rds_rdma_args.
59 * RDS_CMSG_RDMA_DEST (recvmsg, sendmsg)
60 * Kernel informs application about intended
61 * source/destination of a RDMA transfer
62 * RDS_CMSG_RDMA_MAP (sendmsg)
63 * Application asks kernel to map the given
64 * memory range into a IB MR, and send the
65 * R_Key along in an RDS extension header.
66 * The cmsg_data is a struct rds_get_mr_args,
67 * the same as for the GET_MR setsockopt.
68 * RDS_CMSG_RDMA_STATUS (recvmsg)
69 * Returns the status of a completed RDMA operation.
71 #define RDS_CMSG_RDMA_ARGS 1
72 #define RDS_CMSG_RDMA_DEST 2
73 #define RDS_CMSG_RDMA_MAP 3
74 #define RDS_CMSG_RDMA_STATUS 4
75 #define RDS_CMSG_CONG_UPDATE 5
76 #define RDS_CMSG_ATOMIC_FADD 6
77 #define RDS_CMSG_ATOMIC_CSWP 7
78 #define RDS_CMSG_MASKED_ATOMIC_FADD 8
79 #define RDS_CMSG_MASKED_ATOMIC_CSWP 9
81 #define RDS_INFO_FIRST 10000
82 #define RDS_INFO_COUNTERS 10000
83 #define RDS_INFO_CONNECTIONS 10001
84 /* 10002 aka RDS_INFO_FLOWS is deprecated */
85 #define RDS_INFO_SEND_MESSAGES 10003
86 #define RDS_INFO_RETRANS_MESSAGES 10004
87 #define RDS_INFO_RECV_MESSAGES 10005
88 #define RDS_INFO_SOCKETS 10006
89 #define RDS_INFO_TCP_SOCKETS 10007
90 #define RDS_INFO_IB_CONNECTIONS 10008
91 #define RDS_INFO_CONNECTION_STATS 10009
92 #define RDS_INFO_IWARP_CONNECTIONS 10010
93 #define RDS_INFO_LAST 10010
95 struct rds_info_counter {
98 } __attribute__((packed));
100 #define RDS_INFO_CONNECTION_FLAG_SENDING 0x01
101 #define RDS_INFO_CONNECTION_FLAG_CONNECTING 0x02
102 #define RDS_INFO_CONNECTION_FLAG_CONNECTED 0x04
104 #define TRANSNAMSIZ 16
106 struct rds_info_connection {
107 uint64_t next_tx_seq;
108 uint64_t next_rx_seq;
111 uint8_t transport[TRANSNAMSIZ]; /* null term ascii */
113 } __attribute__((packed));
115 struct rds_info_flow {
121 } __attribute__((packed));
123 #define RDS_INFO_MESSAGE_FLAG_ACK 0x01
124 #define RDS_INFO_MESSAGE_FLAG_FAST_ACK 0x02
126 struct rds_info_message {
134 } __attribute__((packed));
136 struct rds_info_socket {
139 __be32 connected_addr;
141 __be16 connected_port;
144 } __attribute__((packed));
146 struct rds_info_tcp_socket {
153 uint32_t last_sent_nxt;
154 uint32_t last_expected_una;
155 uint32_t last_seen_una;
156 } __attribute__((packed));
158 #define RDS_IB_GID_LEN 16
159 struct rds_info_rdma_connection {
162 uint8_t src_gid[RDS_IB_GID_LEN];
163 uint8_t dst_gid[RDS_IB_GID_LEN];
165 uint32_t max_send_wr;
166 uint32_t max_recv_wr;
167 uint32_t max_send_sge;
168 uint32_t rdma_mr_max;
169 uint32_t rdma_mr_size;
173 * Congestion monitoring.
174 * Congestion control in RDS happens at the host connection
175 * level by exchanging a bitmap marking congested ports.
176 * By default, a process sleeping in poll() is always woken
177 * up when the congestion map is updated.
178 * With explicit monitoring, an application can have more
179 * fine-grained control.
180 * The application installs a 64bit mask value in the socket,
181 * where each bit corresponds to a group of ports.
182 * When a congestion update arrives, RDS checks the set of
183 * ports that are now uncongested against the list bit mask
184 * installed in the socket, and if they overlap, we queue a
185 * cong_notification on the socket.
187 * To install the congestion monitor bitmask, use RDS_CONG_MONITOR
188 * with the 64bit mask.
189 * Congestion updates are received via RDS_CMSG_CONG_UPDATE
192 * The correspondence between bits and ports is
193 * 1 << (portnum % 64)
195 #define RDS_CONG_MONITOR_SIZE 64
196 #define RDS_CONG_MONITOR_BIT(port) (((unsigned int) port) % RDS_CONG_MONITOR_SIZE)
197 #define RDS_CONG_MONITOR_MASK(port) (1ULL << RDS_CONG_MONITOR_BIT(port))
204 * This encapsulates a remote memory location.
205 * In the current implementation, it contains the R_Key
206 * of the remote memory region, and the offset into it
207 * (so that the application does not have to worry about
210 typedef uint64_t rds_rdma_cookie_t;
217 struct rds_get_mr_args {
218 struct rds_iovec vec;
219 uint64_t cookie_addr;
223 struct rds_get_mr_for_dest_args {
224 struct sockaddr_storage dest_addr;
225 struct rds_iovec vec;
226 uint64_t cookie_addr;
230 struct rds_free_mr_args {
231 rds_rdma_cookie_t cookie;
235 struct rds_rdma_args {
236 rds_rdma_cookie_t cookie;
237 struct rds_iovec remote_vec;
238 uint64_t local_vec_addr;
244 struct rds_atomic_args {
245 rds_rdma_cookie_t cookie;
247 uint64_t remote_addr;
259 uint64_t compare_mask;
264 uint64_t nocarry_mask;
271 struct rds_rdma_notify {
276 #define RDS_RDMA_SUCCESS 0
277 #define RDS_RDMA_REMOTE_ERROR 1
278 #define RDS_RDMA_CANCELED 2
279 #define RDS_RDMA_DROPPED 3
280 #define RDS_RDMA_OTHER_ERROR 4
283 * Common set of flags for all RDMA related structs
285 #define RDS_RDMA_READWRITE 0x0001
286 #define RDS_RDMA_FENCE 0x0002 /* use FENCE for immediate send */
287 #define RDS_RDMA_INVALIDATE 0x0004 /* invalidate R_Key after freeing MR */
288 #define RDS_RDMA_USE_ONCE 0x0008 /* free MR after use */
289 #define RDS_RDMA_DONTWAIT 0x0010 /* Don't wait in SET_BARRIER */
290 #define RDS_RDMA_NOTIFY_ME 0x0020 /* Notify when operation completes */
291 #define RDS_RDMA_SILENT 0x0040 /* Do not interrupt remote */
293 #endif /* IB_RDS_H */