1/* 2 * Copyright (c) 2004-2009 Voltaire, Inc. All rights reserved. 3 * Copyright (c) 2002-2005,2008 Mellanox Technologies LTD. All rights reserved. 4 * Copyright (c) 1996-2003 Intel Corporation. All rights reserved. 5 * Copyright (c) 2009 HNR Consulting. All rights reserved. 6 * 7 * This software is available to you under a choice of one of two 8 * licenses. You may choose to be licensed under the terms of the GNU 9 * General Public License (GPL) Version 2, available from the file 10 * COPYING in the main directory of this source tree, or the 11 * OpenIB.org BSD license below: 12 * 13 * Redistribution and use in source and binary forms, with or 14 * without modification, are permitted provided that the following 15 * conditions are met: 16 * 17 * - Redistributions of source code must retain the above 18 * copyright notice, this list of conditions and the following 19 * disclaimer. 20 * 21 * - Redistributions in binary form must reproduce the above 22 * copyright notice, this list of conditions and the following 23 * disclaimer in the documentation and/or other materials 24 * provided with the distribution. 25 * 26 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 27 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 28 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 29 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS 30 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 31 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN 32 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 33 * SOFTWARE. 34 * 35 */ 36 37/* 38 * Abstract: 39 * Implementation of osm_si_rcv_t. 40 * This object represents the SwitchInfo Receiver object. 41 * This object is part of the opensm family of objects. 42 */ 43 44#if HAVE_CONFIG_H 45# include <config.h> 46#endif /* HAVE_CONFIG_H */ 47 48#include <string.h> 49#include <iba/ib_types.h> 50#include <complib/cl_qmap.h> 51#include <complib/cl_passivelock.h> 52#include <complib/cl_debug.h> 53#include <opensm/osm_file_ids.h> 54#define FILE_ID OSM_FILE_SW_INFO_RCV_C 55#include <opensm/osm_log.h> 56#include <opensm/osm_switch.h> 57#include <opensm/osm_subnet.h> 58#include <opensm/osm_helper.h> 59#include <opensm/osm_opensm.h> 60 61#if 0 62/********************************************************************** 63 The plock must be held before calling this function. 64**********************************************************************/ 65static void si_rcv_get_fwd_tbl(IN osm_sm_t * sm, IN osm_switch_t * p_sw) 66{ 67 osm_madw_context_t context; 68 osm_dr_path_t *p_dr_path; 69 osm_physp_t *p_physp; 70 osm_node_t *p_node; 71 uint32_t block_id_ho; 72 uint32_t max_block_id_ho; 73 ib_api_status_t status = IB_SUCCESS; 74 75 OSM_LOG_ENTER(sm->p_log); 76 77 CL_ASSERT(p_sw); 78 79 p_node = p_sw->p_node; 80 81 CL_ASSERT(osm_node_get_type(p_node) == IB_NODE_TYPE_SWITCH); 82 83 context.lft_context.node_guid = osm_node_get_node_guid(p_node); 84 context.lft_context.set_method = FALSE; 85 86 max_block_id_ho = osm_switch_get_max_block_id_in_use(p_sw); 87 88 p_physp = osm_node_get_physp_ptr(p_node, 0); 89 p_dr_path = osm_physp_get_dr_path_ptr(p_physp); 90 91 for (block_id_ho = 0; block_id_ho <= max_block_id_ho; block_id_ho++) { 92 OSM_LOG(sm->p_log, OSM_LOG_DEBUG, 93 "Retrieving FT block %u\n", block_id_ho); 94 95 status = osm_req_get(sm, p_dr_path, IB_MAD_ATTR_LIN_FWD_TBL, 96 cl_hton32(block_id_ho), TRUE, 0, 97 CL_DISP_MSGID_NONE, &context); 98 if (status != IB_SUCCESS) 99 /* continue the loop despite the error */ 100 OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3603: " 101 "Failure initiating PortInfo request (%s)\n", 102 ib_get_err_str(status)); 103 } 104 105 OSM_LOG_EXIT(sm->p_log); 106} 107 108/********************************************************************** 109 The plock must be held before calling this function. 110**********************************************************************/ 111static void si_rcv_get_mcast_fwd_tbl(IN osm_sm_t * sm, IN osm_switch_t * p_sw) 112{ 113 osm_madw_context_t context; 114 osm_dr_path_t *p_dr_path; 115 osm_physp_t *p_physp; 116 osm_node_t *p_node; 117 osm_mcast_tbl_t *p_tbl; 118 uint32_t block_id_ho; 119 uint32_t max_block_id_ho; 120 uint32_t position; 121 uint32_t max_position; 122 uint32_t attr_mod_ho; 123 ib_api_status_t status = IB_SUCCESS; 124 125 OSM_LOG_ENTER(sm->p_log); 126 127 CL_ASSERT(p_sw); 128 129 p_node = p_sw->p_node; 130 131 CL_ASSERT(osm_node_get_type(p_node) == IB_NODE_TYPE_SWITCH); 132 133 if (osm_switch_get_mcast_fwd_tbl_size(p_sw) == 0) { 134 OSM_LOG(sm->p_log, OSM_LOG_DEBUG, 135 "Multicast not supported by switch 0x%016" PRIx64 "\n", 136 cl_ntoh64(osm_node_get_node_guid(p_node))); 137 goto Exit; 138 } 139 140 context.mft_context.node_guid = osm_node_get_node_guid(p_node); 141 context.mft_context.set_method = FALSE; 142 143 p_tbl = osm_switch_get_mcast_tbl_ptr(p_sw); 144 max_block_id_ho = osm_mcast_tbl_get_max_block(p_tbl); 145 146 if (max_block_id_ho > IB_MCAST_MAX_BLOCK_ID) { 147 OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3609: " 148 "Out-of-range mcast block size = %u on switch 0x%016" 149 PRIx64 "\n", max_block_id_ho, 150 cl_ntoh64(osm_node_get_node_guid(p_node))); 151 goto Exit; 152 } 153 154 max_position = osm_mcast_tbl_get_max_position(p_tbl); 155 156 CL_ASSERT(max_position <= IB_MCAST_POSITION_MAX); 157 158 OSM_LOG(sm->p_log, OSM_LOG_DEBUG, 159 "Max MFT block = %u, Max position = %u\n", max_block_id_ho, 160 max_position); 161 162 p_physp = osm_node_get_physp_ptr(p_node, 0); 163 p_dr_path = osm_physp_get_dr_path_ptr(p_physp); 164 165 for (block_id_ho = 0; block_id_ho <= max_block_id_ho; block_id_ho++) { 166 OSM_LOG(sm->p_log, OSM_LOG_DEBUG, 167 "Retrieving MFT block %u\n", block_id_ho); 168 169 for (position = 0; position <= max_position; position++) { 170 OSM_LOG(sm->p_log, OSM_LOG_DEBUG, 171 "Retrieving MFT position %u\n", position); 172 173 attr_mod_ho = 174 block_id_ho | position << IB_MCAST_POSITION_SHIFT; 175 status = 176 osm_req_get(sm, p_dr_path, 177 IB_MAD_ATTR_MCAST_FWD_TBL, 178 cl_hton32(attr_mod_ho), TRUE, 0, 179 CL_DISP_MSGID_NONE, &context); 180 if (status != IB_SUCCESS) 181 /* continue the loop despite the error */ 182 OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3607: " 183 "Failure initiating PortInfo request (%s)\n", 184 ib_get_err_str(status)); 185 } 186 } 187 188Exit: 189 OSM_LOG_EXIT(sm->p_log); 190} 191#endif 192 193/********************************************************************** 194 Lock must be held on entry to this function. 195**********************************************************************/ 196static void si_rcv_process_new(IN osm_sm_t * sm, IN osm_node_t * p_node, 197 IN const osm_madw_t * p_madw) 198{ 199 osm_switch_t *p_sw; 200 osm_switch_t *p_check; 201 ib_switch_info_t *p_si; 202 ib_smp_t *p_smp; 203 cl_qmap_t *p_sw_guid_tbl; 204 205 CL_ASSERT(sm); 206 207 OSM_LOG_ENTER(sm->p_log); 208 209 CL_ASSERT(p_madw); 210 211 p_sw_guid_tbl = &sm->p_subn->sw_guid_tbl; 212 p_smp = osm_madw_get_smp_ptr(p_madw); 213 p_si = ib_smp_get_payload_ptr(p_smp); 214 215 osm_dump_switch_info_v2(sm->p_log, p_si, FILE_ID, OSM_LOG_DEBUG); 216 217 p_sw = osm_switch_new(p_node, p_madw); 218 if (p_sw == NULL) { 219 OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3608: " 220 "Unable to allocate new switch object\n"); 221 goto Exit; 222 } 223 224 /* set subnet max mlid to the minimum MulticastFDBCap of all switches */ 225 if (p_si->mcast_cap && 226 cl_ntoh16(p_si->mcast_cap) + IB_LID_MCAST_START_HO - 1 < 227 sm->p_subn->max_mcast_lid_ho) { 228 sm->p_subn->max_mcast_lid_ho = cl_ntoh16(p_si->mcast_cap) + 229 IB_LID_MCAST_START_HO - 1; 230 OSM_LOG(sm->p_log, OSM_LOG_VERBOSE, 231 "Subnet max multicast lid is 0x%X\n", 232 sm->p_subn->max_mcast_lid_ho); 233 } 234 235 /* set subnet max unicast lid to the minimum LinearFDBCap of all switches */ 236 if (cl_ntoh16(p_si->lin_cap) < sm->p_subn->max_ucast_lid_ho) { 237 sm->p_subn->max_ucast_lid_ho = cl_ntoh16(p_si->lin_cap); 238 OSM_LOG(sm->p_log, OSM_LOG_VERBOSE, 239 "Subnet max unicast lid is 0x%X\n", 240 sm->p_subn->max_ucast_lid_ho); 241 } 242 243 p_check = (osm_switch_t *) cl_qmap_insert(p_sw_guid_tbl, 244 osm_node_get_node_guid 245 (p_node), &p_sw->map_item); 246 if (p_check != p_sw) { 247 /* This shouldn't happen since we hold the lock! */ 248 OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3605: " 249 "Unable to add new switch object to database\n"); 250 osm_switch_delete(&p_sw); 251 goto Exit; 252 } 253 254 p_node->sw = p_sw; 255 256 /* Update the switch info according to the info we just received. */ 257 osm_switch_set_switch_info(p_sw, p_si); 258 259#if 0 260 /* Don't bother retrieving the current unicast and multicast tables 261 from the switches. The current version of SM does 262 not support silent take-over of an existing multicast 263 configuration. 264 265 Gathering the multicast tables can also generate large amounts 266 of extra subnet-init traffic. 267 268 The code to retrieve the tables was fully debugged. */ 269 270 si_rcv_get_fwd_tbl(sm, p_sw); 271 if (!sm->p_subn->opt.disable_multicast) 272 si_rcv_get_mcast_fwd_tbl(sm, p_sw); 273#endif 274 275Exit: 276 OSM_LOG_EXIT(sm->p_log); 277} 278 279/********************************************************************** 280 Lock must be held on entry to this function. 281 Return 1 if the caller is expected to send a change_detected event. 282 this can not be done internally as the event needs the lock... 283**********************************************************************/ 284static boolean_t si_rcv_process_existing(IN osm_sm_t * sm, 285 IN osm_node_t * p_node, 286 IN const osm_madw_t * p_madw) 287{ 288 osm_switch_t *p_sw = p_node->sw; 289 ib_switch_info_t *p_si; 290 osm_si_context_t *p_si_context; 291 ib_smp_t *p_smp; 292 osm_epi_lft_change_event_t lft_change; 293 boolean_t is_change_detected = FALSE; 294 295 OSM_LOG_ENTER(sm->p_log); 296 297 CL_ASSERT(p_madw); 298 299 p_smp = osm_madw_get_smp_ptr(p_madw); 300 p_si = ib_smp_get_payload_ptr(p_smp); 301 p_si_context = osm_madw_get_si_context_ptr(p_madw); 302 303 OSM_LOG(sm->p_log, OSM_LOG_DEBUG, "Received logical %cetResp()\n", 304 p_si_context->set_method ? 'S' : 'G'); 305 306 osm_switch_set_switch_info(p_sw, p_si); 307 308 if (p_si_context->light_sweep == TRUE && !p_si_context->set_method) { 309 /* If state changed bit is on the mad was returned with an 310 error - signal a change to the state manager. */ 311 if (ib_smp_get_status(p_smp) != 0) { 312 OSM_LOG(sm->p_log, OSM_LOG_VERBOSE, 313 "GetResp() received with error in light sweep. " 314 "Commencing heavy sweep\n"); 315 is_change_detected = TRUE; 316 } else if (ib_switch_info_get_state_change(p_si)) { 317 osm_dump_switch_info_v2(sm->p_log, p_si, FILE_ID, OSM_LOG_DEBUG); 318 is_change_detected = TRUE; 319 } 320 } 321 322 if (sm->p_subn->first_time_master_sweep == FALSE && 323 p_si_context->set_method && p_si_context->lft_top_change) { 324 lft_change.p_sw = p_sw; 325 lft_change.flags = LFT_CHANGED_LFT_TOP; 326 lft_change.lft_top = cl_ntoh16(p_si->lin_top); 327 lft_change.block_num = 0; 328 osm_opensm_report_event(sm->p_subn->p_osm, 329 OSM_EVENT_ID_LFT_CHANGE, 330 &lft_change); 331 } 332 333 OSM_LOG_EXIT(sm->p_log); 334 return is_change_detected; 335} 336 337static void si_rcv_get_sp0_info(IN osm_sm_t * sm, IN osm_node_t * node) 338{ 339 osm_madw_context_t context; 340 osm_physp_t *physp; 341 ib_api_status_t status; 342 int mlnx_epi_supported = 0; 343 344 physp = osm_node_get_physp_ptr(node, 0); 345 346 context.pi_context.node_guid = osm_node_get_node_guid(node); 347 context.pi_context.port_guid = osm_physp_get_port_guid(physp); 348 context.pi_context.set_method = FALSE; 349 context.pi_context.light_sweep = FALSE; 350 context.pi_context.active_transition = FALSE; 351 context.pi_context.client_rereg = FALSE; 352 353 status = osm_req_get(sm, osm_physp_get_dr_path_ptr(physp), 354 IB_MAD_ATTR_PORT_INFO, 0, TRUE, 0, 355 CL_DISP_MSGID_NONE, &context); 356 if (status != IB_SUCCESS) 357 OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3611: " 358 "Failure initiating PortInfo request (%s)\n", 359 ib_get_err_str(status)); 360 361 if (ib_switch_info_is_enhanced_port0(&node->sw->switch_info) && 362 sm->p_subn->opt.fdr10) { 363 mlnx_epi_supported = is_mlnx_ext_port_info_supported( 364 ib_node_info_get_vendor_id(&node->node_info), 365 node->node_info.device_id); 366 if (mlnx_epi_supported) { 367 status = osm_req_get(sm, 368 osm_physp_get_dr_path_ptr(physp), 369 IB_MAD_ATTR_MLNX_EXTENDED_PORT_INFO, 370 0, TRUE, 0, 371 CL_DISP_MSGID_NONE, &context); 372 if (status != IB_SUCCESS) 373 OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3616: " 374 "Failure initiating MLNX ExtPortInfo request (%s)\n", 375 ib_get_err_str(status)); 376 } 377 } 378 379} 380 381void osm_si_rcv_process(IN void *context, IN void *data) 382{ 383 osm_sm_t *sm = context; 384 osm_madw_t *p_madw = data; 385 ib_switch_info_t *p_si; 386 ib_smp_t *p_smp; 387 osm_node_t *p_node; 388 ib_net64_t node_guid; 389 osm_si_context_t *p_context; 390 391 CL_ASSERT(sm); 392 393 OSM_LOG_ENTER(sm->p_log); 394 395 CL_ASSERT(p_madw); 396 397 p_smp = osm_madw_get_smp_ptr(p_madw); 398 p_si = ib_smp_get_payload_ptr(p_smp); 399 p_context = osm_madw_get_si_context_ptr(p_madw); 400 node_guid = p_context->node_guid; 401 402 OSM_LOG(sm->p_log, OSM_LOG_DEBUG, 403 "Switch GUID 0x%016" PRIx64 ", TID 0x%" PRIx64 "\n", 404 cl_ntoh64(node_guid), cl_ntoh64(p_smp->trans_id)); 405 406 if (ib_smp_get_status(p_smp)) { 407 OSM_LOG(sm->p_log, OSM_LOG_DEBUG, 408 "MAD status 0x%x received\n", 409 cl_ntoh16(ib_smp_get_status(p_smp))); 410 goto Exit2; 411 } 412 413 CL_PLOCK_EXCL_ACQUIRE(sm->p_lock); 414 415 p_node = osm_get_node_by_guid(sm->p_subn, node_guid); 416 if (!p_node) { 417 CL_PLOCK_RELEASE(sm->p_lock); 418 OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3606: " 419 "SwitchInfo received for nonexistent node " 420 "with GUID 0x%" PRIx64 "\n", cl_ntoh64(node_guid)); 421 goto Exit; 422 } 423 424 /* Hack for bad value in Mellanox switch */ 425 if (cl_ntoh16(p_si->lin_top) > IB_LID_UCAST_END_HO) { 426 OSM_LOG(sm->p_log, OSM_LOG_ERROR, "ERR 3610: " 427 "\n\t\t\t\tBad LinearFDBTop value = 0x%X " 428 "on switch 0x%" PRIx64 429 "\n\t\t\t\tForcing internal correction to 0x%X\n", 430 cl_ntoh16(p_si->lin_top), 431 cl_ntoh64(osm_node_get_node_guid(p_node)), 0); 432 p_si->lin_top = 0; 433 } 434 435 /* Acquire the switch object for this switch. */ 436 if (!p_node->sw) { 437 si_rcv_process_new(sm, p_node, p_madw); 438 /* A new switch was found during the sweep so we need 439 to ignore the current LFT settings. */ 440 sm->p_subn->ignore_existing_lfts = TRUE; 441 } else if (si_rcv_process_existing(sm, p_node, p_madw)) 442 /* we might get back a request for signaling change was detected */ 443 sm->p_subn->force_heavy_sweep = TRUE; 444 445 if (p_context->light_sweep || p_context->set_method) 446 goto Exit; 447 448 si_rcv_get_sp0_info(sm, p_node); 449 450Exit: 451 CL_PLOCK_RELEASE(sm->p_lock); 452Exit2: 453 OSM_LOG_EXIT(sm->p_log); 454} 455