include/linux/rds.h

   1 /*
   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 #define RDS_GET_MR_FOR_DEST             7
  60
  61 /*
  62  * Control message types for SOL_RDS.
  63  *
  64  * CMSG_RDMA_ARGS (sendmsg)
  65  *      Request a RDMA transfer to/from the specified
  66  *      memory ranges.
  67  *      The cmsg_data is a struct rds_rdma_args.
  68  * RDS_CMSG_RDMA_DEST (recvmsg, sendmsg)
  69  *      Kernel informs application about intended
  70  *      source/destination of a RDMA transfer
  71  * RDS_CMSG_RDMA_MAP (sendmsg)
  72  *      Application asks kernel to map the given
  73  *      memory range into a IB MR, and send the
  74  *      R_Key along in an RDS extension header.
  75  *      The cmsg_data is a struct rds_get_mr_args,
  76  *      the same as for the GET_MR setsockopt.
  77  * RDS_CMSG_RDMA_STATUS (recvmsg)
  78  *      Returns the status of a completed RDMA operation.
  79  */
  80 #define RDS_CMSG_RDMA_ARGS              1
  81 #define RDS_CMSG_RDMA_DEST              2
  82 #define RDS_CMSG_RDMA_MAP               3
  83 #define RDS_CMSG_RDMA_STATUS            4
  84 #define RDS_CMSG_CONG_UPDATE            5
  85
  86 #define RDS_INFO_FIRST                  10000
  87 #define RDS_INFO_COUNTERS               10000
  88 #define RDS_INFO_CONNECTIONS            10001
  89 /* 10002 aka RDS_INFO_FLOWS is deprecated */
  90 #define RDS_INFO_SEND_MESSAGES          10003
  91 #define RDS_INFO_RETRANS_MESSAGES       10004
  92 #define RDS_INFO_RECV_MESSAGES          10005
  93 #define RDS_INFO_SOCKETS                10006
  94 #define RDS_INFO_TCP_SOCKETS            10007
  95 #define RDS_INFO_IB_CONNECTIONS         10008
  96 #define RDS_INFO_CONNECTION_STATS       10009
  97 #define RDS_INFO_IWARP_CONNECTIONS      10010
  98 #define RDS_INFO_LAST                   10010
  99
 100 struct rds_info_counter {
 101         u_int8_t        name[32];
 102         u_int64_t       value;
 103 } __attribute__((packed));
 104
 105 #define RDS_INFO_CONNECTION_FLAG_SENDING        0x01
 106 #define RDS_INFO_CONNECTION_FLAG_CONNECTING     0x02
 107 #define RDS_INFO_CONNECTION_FLAG_CONNECTED      0x04
 108
 109 #define TRANSNAMSIZ     16
 110
 111 struct rds_info_connection {
 112         u_int64_t       next_tx_seq;
 113         u_int64_t       next_rx_seq;
 114         __be32          laddr;
 115         __be32          faddr;
 116         u_int8_t        transport[TRANSNAMSIZ];         /* null term ascii */
 117         u_int8_t        flags;
 118 } __attribute__((packed));
 119
 120 struct rds_info_flow {
 121         __be32          laddr;
 122         __be32          faddr;
 123         u_int32_t       bytes;
 124         __be16          lport;
 125         __be16          fport;
 126 } __attribute__((packed));
 127
 128 #define RDS_INFO_MESSAGE_FLAG_ACK               0x01
 129 #define RDS_INFO_MESSAGE_FLAG_FAST_ACK          0x02
 130
 131 struct rds_info_message {
 132         u_int64_t       seq;
 133         u_int32_t       len;
 134         __be32          laddr;
 135         __be32          faddr;
 136         __be16          lport;
 137         __be16          fport;
 138         u_int8_t        flags;
 139 } __attribute__((packed));
 140
 141 struct rds_info_socket {
 142         u_int32_t       sndbuf;
 143         __be32          bound_addr;
 144         __be32          connected_addr;
 145         __be16          bound_port;
 146         __be16          connected_port;
 147         u_int32_t       rcvbuf;
 148         u_int64_t       inum;
 149 } __attribute__((packed));
 150
 151 struct rds_info_tcp_socket {
 152         __be32          local_addr;
 153         __be16          local_port;
 154         __be32          peer_addr;
 155         __be16          peer_port;
 156         u_int64_t       hdr_rem;
 157         u_int64_t       data_rem;
 158         u_int32_t       last_sent_nxt;
 159         u_int32_t       last_expected_una;
 160         u_int32_t       last_seen_una;
 161 } __attribute__((packed));
 162
 163 #define RDS_IB_GID_LEN  16
 164 struct rds_info_rdma_connection {
 165         __be32          src_addr;
 166         __be32          dst_addr;
 167         uint8_t         src_gid[RDS_IB_GID_LEN];
 168         uint8_t         dst_gid[RDS_IB_GID_LEN];
 169
 170         uint32_t        max_send_wr;
 171         uint32_t        max_recv_wr;
 172         uint32_t        max_send_sge;
 173         uint32_t        rdma_mr_max;
 174         uint32_t        rdma_mr_size;
 175 };
 176
 177 /*
 178  * Congestion monitoring.
 179  * Congestion control in RDS happens at the host connection
 180  * level by exchanging a bitmap marking congested ports.
 181  * By default, a process sleeping in poll() is always woken
 182  * up when the congestion map is updated.
 183  * With explicit monitoring, an application can have more
 184  * fine-grained control.
 185  * The application installs a 64bit mask value in the socket,
 186  * where each bit corresponds to a group of ports.
 187  * When a congestion update arrives, RDS checks the set of
 188  * ports that are now uncongested against the list bit mask
 189  * installed in the socket, and if they overlap, we queue a
 190  * cong_notification on the socket.
 191  *
 192  * To install the congestion monitor bitmask, use RDS_CONG_MONITOR
 193  * with the 64bit mask.
 194  * Congestion updates are received via RDS_CMSG_CONG_UPDATE
 195  * control messages.
 196  *
 197  * The correspondence between bits and ports is
 198  *      1 << (portnum % 64)
 199  */
 200 #define RDS_CONG_MONITOR_SIZE   64
 201 #define RDS_CONG_MONITOR_BIT(port)  (((unsigned int) port) % RDS_CONG_MONITOR_SIZE)
 202 #define RDS_CONG_MONITOR_MASK(port) (1ULL << RDS_CONG_MONITOR_BIT(port))
 203
 204 /*
 205  * RDMA related types
 206  */
 207
 208 /*
 209  * This encapsulates a remote memory location.
 210  * In the current implementation, it contains the R_Key
 211  * of the remote memory region, and the offset into it
 212  * (so that the application does not have to worry about
 213  * alignment).
 214  */
 215 typedef u_int64_t       rds_rdma_cookie_t;
 216
 217 struct rds_iovec {
 218         u_int64_t       addr;
 219         u_int64_t       bytes;
 220 };
 221
 222 struct rds_get_mr_args {
 223         struct rds_iovec vec;
 224         u_int64_t       cookie_addr;
 225         uint64_t        flags;
 226 };
 227
 228 struct rds_get_mr_for_dest_args {
 229         struct sockaddr_storage dest_addr;
 230         struct rds_iovec        vec;
 231         u_int64_t               cookie_addr;
 232         uint64_t                flags;
 233 };
 234
 235 struct rds_free_mr_args {
 236         rds_rdma_cookie_t cookie;
 237         u_int64_t       flags;
 238 };
 239
 240 struct rds_rdma_args {
 241         rds_rdma_cookie_t cookie;
 242         struct rds_iovec remote_vec;
 243         u_int64_t       local_vec_addr;
 244         u_int64_t       nr_local;
 245         u_int64_t       flags;
 246         u_int64_t       user_token;
 247 };
 248
 249 struct rds_rdma_notify {
 250         u_int64_t       user_token;
 251         int32_t         status;
 252 };
 253
 254 #define RDS_RDMA_SUCCESS        0
 255 #define RDS_RDMA_REMOTE_ERROR   1
 256 #define RDS_RDMA_CANCELED       2
 257 #define RDS_RDMA_DROPPED        3
 258 #define RDS_RDMA_OTHER_ERROR    4
 259
 260 /*
 261  * Common set of flags for all RDMA related structs
 262  */
 263 #define RDS_RDMA_READWRITE      0x0001
 264 #define RDS_RDMA_FENCE          0x0002  /* use FENCE for immediate send */
 265 #define RDS_RDMA_INVALIDATE     0x0004  /* invalidate R_Key after freeing MR */
 266 #define RDS_RDMA_USE_ONCE       0x0008  /* free MR after use */
 267 #define RDS_RDMA_DONTWAIT       0x0010  /* Don't wait in SET_BARRIER */
 268 #define RDS_RDMA_NOTIFY_ME      0x0020  /* Notify when operation completes */
 269
 270 #endif /* IB_RDS_H */