blob: d91dc91f544302a05765126308a345c1dac1c686 [file] [log] [blame]
Andy Groverdb49b9d2009-02-24 15:30:42 +00001/*
2 * Copyright (c) 2008 Oracle. All rights reserved.
3 *
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:
9 *
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
12 * conditions are met:
13 *
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
16 * disclaimer.
17 *
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.
22 *
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
30 * SOFTWARE.
31 *
32 */
33
34#ifndef _LINUX_RDS_H
35#define _LINUX_RDS_H
36
37#include <linux/types.h>
38
39/* These sparse annotated types shouldn't be in any user
40 * visible header file. We should clean this up rather
41 * than kludging around them. */
42#ifndef __KERNEL__
43#define __be16 u_int16_t
44#define __be32 u_int32_t
45#define __be64 u_int64_t
46#endif
47
48#define RDS_IB_ABI_VERSION 0x301
49
50/*
51 * setsockopt/getsockopt for SOL_RDS
52 */
53#define RDS_CANCEL_SENT_TO 1
54#define RDS_GET_MR 2
55#define RDS_FREE_MR 3
56/* deprecated: RDS_BARRIER 4 */
57#define RDS_RECVERR 5
58#define RDS_CONG_MONITOR 6
59
60/*
61 * Control message types for SOL_RDS.
62 *
63 * CMSG_RDMA_ARGS (sendmsg)
64 * Request a RDMA transfer to/from the specified
65 * memory ranges.
66 * The cmsg_data is a struct rds_rdma_args.
67 * RDS_CMSG_RDMA_DEST (recvmsg, sendmsg)
68 * Kernel informs application about intended
69 * source/destination of a RDMA transfer
70 * RDS_CMSG_RDMA_MAP (sendmsg)
71 * Application asks kernel to map the given
72 * memory range into a IB MR, and send the
73 * R_Key along in an RDS extension header.
74 * The cmsg_data is a struct rds_get_mr_args,
75 * the same as for the GET_MR setsockopt.
76 * RDS_CMSG_RDMA_STATUS (recvmsg)
77 * Returns the status of a completed RDMA operation.
78 */
79#define RDS_CMSG_RDMA_ARGS 1
80#define RDS_CMSG_RDMA_DEST 2
81#define RDS_CMSG_RDMA_MAP 3
82#define RDS_CMSG_RDMA_STATUS 4
83#define RDS_CMSG_CONG_UPDATE 5
84
85#define RDS_INFO_FIRST 10000
86#define RDS_INFO_COUNTERS 10000
87#define RDS_INFO_CONNECTIONS 10001
88/* 10002 aka RDS_INFO_FLOWS is deprecated */
89#define RDS_INFO_SEND_MESSAGES 10003
90#define RDS_INFO_RETRANS_MESSAGES 10004
91#define RDS_INFO_RECV_MESSAGES 10005
92#define RDS_INFO_SOCKETS 10006
93#define RDS_INFO_TCP_SOCKETS 10007
94#define RDS_INFO_IB_CONNECTIONS 10008
95#define RDS_INFO_CONNECTION_STATS 10009
96#define RDS_INFO_IWARP_CONNECTIONS 10010
97#define RDS_INFO_LAST 10010
98
99struct rds_info_counter {
100 u_int8_t name[32];
101 u_int64_t value;
102} __attribute__((packed));
103
104#define RDS_INFO_CONNECTION_FLAG_SENDING 0x01
105#define RDS_INFO_CONNECTION_FLAG_CONNECTING 0x02
106#define RDS_INFO_CONNECTION_FLAG_CONNECTED 0x04
107
108#define TRANSNAMSIZ 16
109
110struct rds_info_connection {
111 u_int64_t next_tx_seq;
112 u_int64_t next_rx_seq;
113 __be32 laddr;
114 __be32 faddr;
115 u_int8_t transport[TRANSNAMSIZ]; /* null term ascii */
116 u_int8_t flags;
117} __attribute__((packed));
118
119struct rds_info_flow {
120 __be32 laddr;
121 __be32 faddr;
122 u_int32_t bytes;
123 __be16 lport;
124 __be16 fport;
125} __attribute__((packed));
126
127#define RDS_INFO_MESSAGE_FLAG_ACK 0x01
128#define RDS_INFO_MESSAGE_FLAG_FAST_ACK 0x02
129
130struct rds_info_message {
131 u_int64_t seq;
132 u_int32_t len;
133 __be32 laddr;
134 __be32 faddr;
135 __be16 lport;
136 __be16 fport;
137 u_int8_t flags;
138} __attribute__((packed));
139
140struct rds_info_socket {
141 u_int32_t sndbuf;
142 __be32 bound_addr;
143 __be32 connected_addr;
144 __be16 bound_port;
145 __be16 connected_port;
146 u_int32_t rcvbuf;
147 u_int64_t inum;
148} __attribute__((packed));
149
150#define RDS_IB_GID_LEN 16
151struct rds_info_rdma_connection {
152 __be32 src_addr;
153 __be32 dst_addr;
154 uint8_t src_gid[RDS_IB_GID_LEN];
155 uint8_t dst_gid[RDS_IB_GID_LEN];
156
157 uint32_t max_send_wr;
158 uint32_t max_recv_wr;
159 uint32_t max_send_sge;
160 uint32_t rdma_mr_max;
161 uint32_t rdma_mr_size;
162};
163
164/*
165 * Congestion monitoring.
166 * Congestion control in RDS happens at the host connection
167 * level by exchanging a bitmap marking congested ports.
168 * By default, a process sleeping in poll() is always woken
169 * up when the congestion map is updated.
170 * With explicit monitoring, an application can have more
171 * fine-grained control.
172 * The application installs a 64bit mask value in the socket,
173 * where each bit corresponds to a group of ports.
174 * When a congestion update arrives, RDS checks the set of
175 * ports that are now uncongested against the list bit mask
176 * installed in the socket, and if they overlap, we queue a
177 * cong_notification on the socket.
178 *
179 * To install the congestion monitor bitmask, use RDS_CONG_MONITOR
180 * with the 64bit mask.
181 * Congestion updates are received via RDS_CMSG_CONG_UPDATE
182 * control messages.
183 *
184 * The correspondence between bits and ports is
185 * 1 << (portnum % 64)
186 */
187#define RDS_CONG_MONITOR_SIZE 64
188#define RDS_CONG_MONITOR_BIT(port) (((unsigned int) port) % RDS_CONG_MONITOR_SIZE)
189#define RDS_CONG_MONITOR_MASK(port) (1ULL << RDS_CONG_MONITOR_BIT(port))
190
191/*
192 * RDMA related types
193 */
194
195/*
196 * This encapsulates a remote memory location.
197 * In the current implementation, it contains the R_Key
198 * of the remote memory region, and the offset into it
199 * (so that the application does not have to worry about
200 * alignment).
201 */
202typedef u_int64_t rds_rdma_cookie_t;
203
204struct rds_iovec {
205 u_int64_t addr;
206 u_int64_t bytes;
207};
208
209struct rds_get_mr_args {
210 struct rds_iovec vec;
211 u_int64_t cookie_addr;
212 uint64_t flags;
213};
214
215struct rds_free_mr_args {
216 rds_rdma_cookie_t cookie;
217 u_int64_t flags;
218};
219
220struct rds_rdma_args {
221 rds_rdma_cookie_t cookie;
222 struct rds_iovec remote_vec;
223 u_int64_t local_vec_addr;
224 u_int64_t nr_local;
225 u_int64_t flags;
226 u_int64_t user_token;
227};
228
229struct rds_rdma_notify {
230 u_int64_t user_token;
231 int32_t status;
232};
233
234#define RDS_RDMA_SUCCESS 0
235#define RDS_RDMA_REMOTE_ERROR 1
236#define RDS_RDMA_CANCELED 2
237#define RDS_RDMA_DROPPED 3
238#define RDS_RDMA_OTHER_ERROR 4
239
240/*
241 * Common set of flags for all RDMA related structs
242 */
243#define RDS_RDMA_READWRITE 0x0001
244#define RDS_RDMA_FENCE 0x0002 /* use FENCE for immediate send */
245#define RDS_RDMA_INVALIDATE 0x0004 /* invalidate R_Key after freeing MR */
246#define RDS_RDMA_USE_ONCE 0x0008 /* free MR after use */
247#define RDS_RDMA_DONTWAIT 0x0010 /* Don't wait in SET_BARRIER */
248#define RDS_RDMA_NOTIFY_ME 0x0020 /* Notify when operation completes */
249
250#endif /* IB_RDS_H */