|
|
1.1 ! root 1: /* ! 2: * Copyright (C) 2009 Michael Brown <[email protected]>. ! 3: * ! 4: * This program is free software; you can redistribute it and/or ! 5: * modify it under the terms of the GNU General Public License as ! 6: * published by the Free Software Foundation; either version 2 of the ! 7: * License, or any later version. ! 8: * ! 9: * This program is distributed in the hope that it will be useful, but ! 10: * WITHOUT ANY WARRANTY; without even the implied warranty of ! 11: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU ! 12: * General Public License for more details. ! 13: * ! 14: * You should have received a copy of the GNU General Public License ! 15: * along with this program; if not, write to the Free Software ! 16: * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. ! 17: */ ! 18: ! 19: FILE_LICENCE ( GPL2_OR_LATER ); ! 20: ! 21: #include <stdint.h> ! 22: #include <stdlib.h> ! 23: #include <string.h> ! 24: #include <byteswap.h> ! 25: #include <errno.h> ! 26: #include <assert.h> ! 27: #include <ipxe/infiniband.h> ! 28: #include <ipxe/ib_mi.h> ! 29: #include <ipxe/ib_pathrec.h> ! 30: #include <ipxe/ib_cm.h> ! 31: ! 32: /** ! 33: * @file ! 34: * ! 35: * Infiniband communication management ! 36: * ! 37: */ ! 38: ! 39: /** List of connections */ ! 40: static LIST_HEAD ( ib_cm_conns ); ! 41: ! 42: /** ! 43: * Find connection by local communication ID ! 44: * ! 45: * @v local_id Local communication ID ! 46: * @ret conn Connection, or NULL ! 47: */ ! 48: static struct ib_connection * ib_cm_find ( uint32_t local_id ) { ! 49: struct ib_connection *conn; ! 50: ! 51: list_for_each_entry ( conn, &ib_cm_conns, list ) { ! 52: if ( conn->local_id == local_id ) ! 53: return conn; ! 54: } ! 55: return NULL; ! 56: } ! 57: ! 58: /** ! 59: * Send "ready to use" response ! 60: * ! 61: * @v ibdev Infiniband device ! 62: * @v mi Management interface ! 63: * @v av Address vector ! 64: * @v local_id Local communication ID ! 65: * @v remote_id Remote communication ID ! 66: * @ret rc Return status code ! 67: */ ! 68: static int ib_cm_send_rtu ( struct ib_device *ibdev, ! 69: struct ib_mad_interface *mi, ! 70: struct ib_address_vector *av, ! 71: uint32_t local_id, uint32_t remote_id ) { ! 72: union ib_mad mad; ! 73: struct ib_cm_ready_to_use *rtu = &mad.cm.cm_data.ready_to_use; ! 74: int rc; ! 75: ! 76: /* Construct "ready to use" response */ ! 77: memset ( &mad, 0, sizeof ( mad ) ); ! 78: mad.hdr.mgmt_class = IB_MGMT_CLASS_CM; ! 79: mad.hdr.class_version = IB_CM_CLASS_VERSION; ! 80: mad.hdr.method = IB_MGMT_METHOD_SEND; ! 81: mad.hdr.attr_id = htons ( IB_CM_ATTR_READY_TO_USE ); ! 82: rtu->local_id = htonl ( local_id ); ! 83: rtu->remote_id = htonl ( remote_id ); ! 84: if ( ( rc = ib_mi_send ( ibdev, mi, &mad, av ) ) != 0 ){ ! 85: DBG ( "CM could not send RTU: %s\n", strerror ( rc ) ); ! 86: return rc; ! 87: } ! 88: ! 89: return 0; ! 90: } ! 91: ! 92: /** ! 93: * Handle duplicate connection replies ! 94: * ! 95: * @v ibdev Infiniband device ! 96: * @v mi Management interface ! 97: * @v mad Received MAD ! 98: * @v av Source address vector ! 99: * @ret rc Return status code ! 100: * ! 101: * If a "ready to use" MAD is lost, the peer may resend the connection ! 102: * reply. We have to respond to these with duplicate "ready to use" ! 103: * MADs, otherwise the peer may time out and drop the connection. ! 104: */ ! 105: static void ib_cm_recv_rep ( struct ib_device *ibdev, ! 106: struct ib_mad_interface *mi, ! 107: union ib_mad *mad, ! 108: struct ib_address_vector *av ) { ! 109: struct ib_cm_connect_reply *rep = &mad->cm.cm_data.connect_reply; ! 110: struct ib_connection *conn; ! 111: uint32_t local_id = ntohl ( rep->remote_id ); ! 112: int rc; ! 113: ! 114: /* Identify connection */ ! 115: conn = ib_cm_find ( local_id ); ! 116: if ( conn ) { ! 117: /* Try to send "ready to use" reply */ ! 118: if ( ( rc = ib_cm_send_rtu ( ibdev, mi, av, conn->local_id, ! 119: conn->remote_id ) ) != 0 ) { ! 120: /* Ignore errors; the remote end will retry */ ! 121: } ! 122: } else { ! 123: DBG ( "CM unidentified connection %08x\n", local_id ); ! 124: } ! 125: } ! 126: ! 127: /** ! 128: * Send reply to disconnection request ! 129: * ! 130: * @v ibdev Infiniband device ! 131: * @v mi Management interface ! 132: * @v av Address vector ! 133: * @v local_id Local communication ID ! 134: * @v remote_id Remote communication ID ! 135: * @ret rc Return status code ! 136: */ ! 137: static int ib_cm_send_drep ( struct ib_device *ibdev, ! 138: struct ib_mad_interface *mi, ! 139: struct ib_address_vector *av, ! 140: uint32_t local_id, uint32_t remote_id ) { ! 141: union ib_mad mad; ! 142: struct ib_cm_disconnect_reply *drep = &mad.cm.cm_data.disconnect_reply; ! 143: int rc; ! 144: ! 145: /* Construct reply to disconnection request */ ! 146: memset ( &mad, 0, sizeof ( mad ) ); ! 147: mad.hdr.mgmt_class = IB_MGMT_CLASS_CM; ! 148: mad.hdr.class_version = IB_CM_CLASS_VERSION; ! 149: mad.hdr.method = IB_MGMT_METHOD_SEND; ! 150: mad.hdr.attr_id = htons ( IB_CM_ATTR_DISCONNECT_REPLY ); ! 151: drep->local_id = htonl ( local_id ); ! 152: drep->remote_id = htonl ( remote_id ); ! 153: if ( ( rc = ib_mi_send ( ibdev, mi, &mad, av ) ) != 0 ){ ! 154: DBG ( "CM could not send DREP: %s\n", strerror ( rc ) ); ! 155: return rc; ! 156: } ! 157: ! 158: return 0; ! 159: } ! 160: ! 161: /** ! 162: * Handle disconnection requests ! 163: * ! 164: * @v ibdev Infiniband device ! 165: * @v mi Management interface ! 166: * @v mad Received MAD ! 167: * @v av Source address vector ! 168: * @ret rc Return status code ! 169: */ ! 170: static void ib_cm_recv_dreq ( struct ib_device *ibdev, ! 171: struct ib_mad_interface *mi, ! 172: union ib_mad *mad, ! 173: struct ib_address_vector *av ) { ! 174: struct ib_cm_disconnect_request *dreq = ! 175: &mad->cm.cm_data.disconnect_request; ! 176: struct ib_connection *conn; ! 177: uint32_t local_id = ntohl ( dreq->remote_id ); ! 178: uint32_t remote_id = ntohl ( dreq->local_id ); ! 179: int rc; ! 180: ! 181: /* Identify connection */ ! 182: conn = ib_cm_find ( local_id ); ! 183: if ( conn ) { ! 184: /* Notify upper layer */ ! 185: conn->op->changed ( ibdev, conn->qp, conn, -ENOTCONN, ! 186: &dreq->private_data, ! 187: sizeof ( dreq->private_data ) ); ! 188: } else { ! 189: DBG ( "CM unidentified connection %08x\n", local_id ); ! 190: } ! 191: ! 192: /* Send reply */ ! 193: if ( ( rc = ib_cm_send_drep ( ibdev, mi, av, local_id, ! 194: remote_id ) ) != 0 ) { ! 195: /* Ignore errors; the remote end will retry */ ! 196: } ! 197: }; ! 198: ! 199: /** Communication management agents */ ! 200: struct ib_mad_agent ib_cm_agent[] __ib_mad_agent = { ! 201: { ! 202: .mgmt_class = IB_MGMT_CLASS_CM, ! 203: .class_version = IB_CM_CLASS_VERSION, ! 204: .attr_id = htons ( IB_CM_ATTR_CONNECT_REPLY ), ! 205: .handle = ib_cm_recv_rep, ! 206: }, ! 207: { ! 208: .mgmt_class = IB_MGMT_CLASS_CM, ! 209: .class_version = IB_CM_CLASS_VERSION, ! 210: .attr_id = htons ( IB_CM_ATTR_DISCONNECT_REQUEST ), ! 211: .handle = ib_cm_recv_dreq, ! 212: }, ! 213: }; ! 214: ! 215: /** ! 216: * Convert connection rejection reason to return status code ! 217: * ! 218: * @v reason Rejection reason (in network byte order) ! 219: * @ret rc Return status code ! 220: */ ! 221: static int ib_cm_rejection_reason_to_rc ( uint16_t reason ) { ! 222: switch ( reason ) { ! 223: case htons ( IB_CM_REJECT_BAD_SERVICE_ID ) : ! 224: return -ENODEV; ! 225: case htons ( IB_CM_REJECT_STALE_CONN ) : ! 226: return -EALREADY; ! 227: case htons ( IB_CM_REJECT_CONSUMER ) : ! 228: return -ENOTTY; ! 229: default: ! 230: return -EPERM; ! 231: } ! 232: } ! 233: ! 234: /** ! 235: * Handle connection request transaction completion ! 236: * ! 237: * @v ibdev Infiniband device ! 238: * @v mi Management interface ! 239: * @v madx Management transaction ! 240: * @v rc Status code ! 241: * @v mad Received MAD (or NULL on error) ! 242: * @v av Source address vector (or NULL on error) ! 243: */ ! 244: static void ib_cm_req_complete ( struct ib_device *ibdev, ! 245: struct ib_mad_interface *mi, ! 246: struct ib_mad_transaction *madx, ! 247: int rc, union ib_mad *mad, ! 248: struct ib_address_vector *av ) { ! 249: struct ib_connection *conn = ib_madx_get_ownerdata ( madx ); ! 250: struct ib_queue_pair *qp = conn->qp; ! 251: struct ib_cm_common *common = &mad->cm.cm_data.common; ! 252: struct ib_cm_connect_reply *rep = &mad->cm.cm_data.connect_reply; ! 253: struct ib_cm_connect_reject *rej = &mad->cm.cm_data.connect_reject; ! 254: void *private_data = NULL; ! 255: size_t private_data_len = 0; ! 256: ! 257: /* Report failures */ ! 258: if ( ( rc == 0 ) && ( mad->hdr.status != htons ( IB_MGMT_STATUS_OK ) )) ! 259: rc = -EIO; ! 260: if ( rc != 0 ) { ! 261: DBGC ( conn, "CM %p connection request failed: %s\n", ! 262: conn, strerror ( rc ) ); ! 263: goto out; ! 264: } ! 265: ! 266: /* Record remote communication ID */ ! 267: conn->remote_id = ntohl ( common->local_id ); ! 268: ! 269: /* Handle response */ ! 270: switch ( mad->hdr.attr_id ) { ! 271: ! 272: case htons ( IB_CM_ATTR_CONNECT_REPLY ) : ! 273: /* Extract fields */ ! 274: qp->av.qpn = ( ntohl ( rep->local_qpn ) >> 8 ); ! 275: qp->send.psn = ( ntohl ( rep->starting_psn ) >> 8 ); ! 276: private_data = &rep->private_data; ! 277: private_data_len = sizeof ( rep->private_data ); ! 278: DBGC ( conn, "CM %p connected to QPN %lx PSN %x\n", ! 279: conn, qp->av.qpn, qp->send.psn ); ! 280: ! 281: /* Modify queue pair */ ! 282: if ( ( rc = ib_modify_qp ( ibdev, qp ) ) != 0 ) { ! 283: DBGC ( conn, "CM %p could not modify queue pair: %s\n", ! 284: conn, strerror ( rc ) ); ! 285: goto out; ! 286: } ! 287: ! 288: /* Send "ready to use" reply */ ! 289: if ( ( rc = ib_cm_send_rtu ( ibdev, mi, av, conn->local_id, ! 290: conn->remote_id ) ) != 0 ) { ! 291: /* Treat as non-fatal */ ! 292: rc = 0; ! 293: } ! 294: break; ! 295: ! 296: case htons ( IB_CM_ATTR_CONNECT_REJECT ) : ! 297: /* Extract fields */ ! 298: DBGC ( conn, "CM %p connection rejected (reason %d)\n", ! 299: conn, ntohs ( rej->reason ) ); ! 300: /* Private data is valid only for a Consumer Reject */ ! 301: if ( rej->reason == htons ( IB_CM_REJECT_CONSUMER ) ) { ! 302: private_data = &rej->private_data; ! 303: private_data_len = sizeof ( rej->private_data ); ! 304: } ! 305: rc = ib_cm_rejection_reason_to_rc ( rej->reason ); ! 306: break; ! 307: ! 308: default: ! 309: DBGC ( conn, "CM %p unexpected response (attribute %04x)\n", ! 310: conn, ntohs ( mad->hdr.attr_id ) ); ! 311: rc = -ENOTSUP; ! 312: break; ! 313: } ! 314: ! 315: out: ! 316: /* Destroy the completed transaction */ ! 317: ib_destroy_madx ( ibdev, ibdev->gsi, madx ); ! 318: conn->madx = NULL; ! 319: ! 320: /* Hand off to the upper completion handler */ ! 321: conn->op->changed ( ibdev, qp, conn, rc, private_data, ! 322: private_data_len ); ! 323: } ! 324: ! 325: /** Connection request operations */ ! 326: static struct ib_mad_transaction_operations ib_cm_req_op = { ! 327: .complete = ib_cm_req_complete, ! 328: }; ! 329: ! 330: /** ! 331: * Handle connection path transaction completion ! 332: * ! 333: * @v ibdev Infiniband device ! 334: * @v path Path ! 335: * @v rc Status code ! 336: * @v av Address vector, or NULL on error ! 337: */ ! 338: static void ib_cm_path_complete ( struct ib_device *ibdev, ! 339: struct ib_path *path, int rc, ! 340: struct ib_address_vector *av ) { ! 341: struct ib_connection *conn = ib_path_get_ownerdata ( path ); ! 342: struct ib_queue_pair *qp = conn->qp; ! 343: union ib_mad mad; ! 344: struct ib_cm_connect_request *req = &mad.cm.cm_data.connect_request; ! 345: size_t private_data_len; ! 346: ! 347: /* Report failures */ ! 348: if ( rc != 0 ) { ! 349: DBGC ( conn, "CM %p path lookup failed: %s\n", ! 350: conn, strerror ( rc ) ); ! 351: conn->op->changed ( ibdev, qp, conn, rc, NULL, 0 ); ! 352: goto out; ! 353: } ! 354: ! 355: /* Update queue pair peer path */ ! 356: memcpy ( &qp->av, av, sizeof ( qp->av ) ); ! 357: ! 358: /* Construct connection request */ ! 359: memset ( &mad, 0, sizeof ( mad ) ); ! 360: mad.hdr.mgmt_class = IB_MGMT_CLASS_CM; ! 361: mad.hdr.class_version = IB_CM_CLASS_VERSION; ! 362: mad.hdr.method = IB_MGMT_METHOD_SEND; ! 363: mad.hdr.attr_id = htons ( IB_CM_ATTR_CONNECT_REQUEST ); ! 364: req->local_id = htonl ( conn->local_id ); ! 365: memcpy ( &req->service_id, &conn->service_id, ! 366: sizeof ( req->service_id ) ); ! 367: memcpy ( &req->local_ca, &ibdev->node_guid, sizeof ( req->local_ca ) ); ! 368: req->local_qpn__responder_resources = htonl ( ( qp->qpn << 8 ) | 1 ); ! 369: req->local_eecn__initiator_depth = htonl ( ( 0 << 8 ) | 1 ); ! 370: req->remote_eecn__remote_timeout__service_type__ee_flow_ctrl = ! 371: htonl ( ( 0x14 << 3 ) | ( IB_CM_TRANSPORT_RC << 1 ) | ! 372: ( 0 << 0 ) ); ! 373: req->starting_psn__local_timeout__retry_count = ! 374: htonl ( ( qp->recv.psn << 8 ) | ( 0x14 << 3 ) | ! 375: ( 0x07 << 0 ) ); ! 376: req->pkey = htons ( ibdev->pkey ); ! 377: req->payload_mtu__rdc_exists__rnr_retry = ! 378: ( ( IB_MTU_2048 << 4 ) | ( 1 << 3 ) | ( 0x07 << 0 ) ); ! 379: req->max_cm_retries__srq = ( ( 0x0f << 4 ) | ( 0 << 3 ) ); ! 380: req->primary.local_lid = htons ( ibdev->lid ); ! 381: req->primary.remote_lid = htons ( conn->qp->av.lid ); ! 382: memcpy ( &req->primary.local_gid, &ibdev->gid, ! 383: sizeof ( req->primary.local_gid ) ); ! 384: memcpy ( &req->primary.remote_gid, &conn->qp->av.gid, ! 385: sizeof ( req->primary.remote_gid ) ); ! 386: req->primary.flow_label__rate = ! 387: htonl ( ( 0 << 12 ) | ( conn->qp->av.rate << 0 ) ); ! 388: req->primary.hop_limit = 0; ! 389: req->primary.sl__subnet_local = ! 390: ( ( conn->qp->av.sl << 4 ) | ( 1 << 3 ) ); ! 391: req->primary.local_ack_timeout = ( 0x13 << 3 ); ! 392: private_data_len = conn->private_data_len; ! 393: if ( private_data_len > sizeof ( req->private_data ) ) ! 394: private_data_len = sizeof ( req->private_data ); ! 395: memcpy ( &req->private_data, &conn->private_data, private_data_len ); ! 396: ! 397: /* Create connection request */ ! 398: av->qpn = IB_QPN_GSI; ! 399: av->qkey = IB_QKEY_GSI; ! 400: conn->madx = ib_create_madx ( ibdev, ibdev->gsi, &mad, av, ! 401: &ib_cm_req_op ); ! 402: if ( ! conn->madx ) { ! 403: DBGC ( conn, "CM %p could not create connection request\n", ! 404: conn ); ! 405: conn->op->changed ( ibdev, qp, conn, rc, NULL, 0 ); ! 406: goto out; ! 407: } ! 408: ib_madx_set_ownerdata ( conn->madx, conn ); ! 409: ! 410: out: ! 411: /* Destroy the completed transaction */ ! 412: ib_destroy_path ( ibdev, path ); ! 413: conn->path = NULL; ! 414: } ! 415: ! 416: /** Connection path operations */ ! 417: static struct ib_path_operations ib_cm_path_op = { ! 418: .complete = ib_cm_path_complete, ! 419: }; ! 420: ! 421: /** ! 422: * Create connection to remote QP ! 423: * ! 424: * @v ibdev Infiniband device ! 425: * @v qp Queue pair ! 426: * @v dgid Target GID ! 427: * @v service_id Target service ID ! 428: * @v private_data Connection request private data ! 429: * @v private_data_len Length of connection request private data ! 430: * @v op Connection operations ! 431: * @ret conn Connection ! 432: */ ! 433: struct ib_connection * ! 434: ib_create_conn ( struct ib_device *ibdev, struct ib_queue_pair *qp, ! 435: union ib_gid *dgid, union ib_guid *service_id, ! 436: void *private_data, size_t private_data_len, ! 437: struct ib_connection_operations *op ) { ! 438: struct ib_connection *conn; ! 439: ! 440: /* Allocate and initialise request */ ! 441: conn = zalloc ( sizeof ( *conn ) + private_data_len ); ! 442: if ( ! conn ) ! 443: goto err_alloc_conn; ! 444: conn->ibdev = ibdev; ! 445: conn->qp = qp; ! 446: memset ( &qp->av, 0, sizeof ( qp->av ) ); ! 447: qp->av.gid_present = 1; ! 448: memcpy ( &qp->av.gid, dgid, sizeof ( qp->av.gid ) ); ! 449: conn->local_id = random(); ! 450: memcpy ( &conn->service_id, service_id, sizeof ( conn->service_id ) ); ! 451: conn->op = op; ! 452: conn->private_data_len = private_data_len; ! 453: memcpy ( &conn->private_data, private_data, private_data_len ); ! 454: ! 455: /* Create path */ ! 456: conn->path = ib_create_path ( ibdev, &qp->av, &ib_cm_path_op ); ! 457: if ( ! conn->path ) ! 458: goto err_create_path; ! 459: ib_path_set_ownerdata ( conn->path, conn ); ! 460: ! 461: /* Add to list of connections */ ! 462: list_add ( &conn->list, &ib_cm_conns ); ! 463: ! 464: DBGC ( conn, "CM %p created for IBDEV %p QPN %lx\n", ! 465: conn, ibdev, qp->qpn ); ! 466: DBGC ( conn, "CM %p connecting to " IB_GID_FMT " " IB_GUID_FMT "\n", ! 467: conn, IB_GID_ARGS ( dgid ), IB_GUID_ARGS ( service_id ) ); ! 468: ! 469: return conn; ! 470: ! 471: ib_destroy_path ( ibdev, conn->path ); ! 472: err_create_path: ! 473: free ( conn ); ! 474: err_alloc_conn: ! 475: return NULL; ! 476: } ! 477: ! 478: /** ! 479: * Destroy connection to remote QP ! 480: * ! 481: * @v ibdev Infiniband device ! 482: * @v qp Queue pair ! 483: * @v conn Connection ! 484: */ ! 485: void ib_destroy_conn ( struct ib_device *ibdev, ! 486: struct ib_queue_pair *qp __unused, ! 487: struct ib_connection *conn ) { ! 488: ! 489: list_del ( &conn->list ); ! 490: if ( conn->madx ) ! 491: ib_destroy_madx ( ibdev, ibdev->gsi, conn->madx ); ! 492: if ( conn->path ) ! 493: ib_destroy_path ( ibdev, conn->path ); ! 494: free ( conn ); ! 495: }
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.