Line data    Source code

       1              : /**
       2              :  * Copyright (c) 2026 Huawei Technologies Co., Ltd.
       3              :  * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
       4              :  * CANN Open Software License Agreement Version 2.0 (the "License").
       5              :  * Please refer to the License for details. You may not use this file except in compliance with the License.
       6              :  * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
       7              :  * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
       8              :  * See LICENSE in the root of the software repository for the full text of the License.
       9              :  */
      10              : 
      11              : #include "root_info_detect_bridge.h"
      12              : 
      13              : #include <memory>
      14              : 
      15              : #include "exception_util.h"
      16              : #include "hccl_common_v2.h"
      17              : #include "log.h"
      18              : #include "orion_adapter_rts.h"
      19              : #include "rank_info_detect.h"
      20              : #include "securec.h"
      21              : 
      22              : namespace Hccl {
      23              : namespace {
      24              : 
      25              : // Provider 编译进 hcomm；hccl_v2 中保留的兼容入口只通过 RootInfoDetectBridge 转发到这里。
      26            0 : HcclResult GetRootInfoImpl(HcclRootInfo *rootInfo)
      27              : {
      28            0 :     HCCL_RUN_INFO("Entry-HcclGetRootInfo V950");
      29            0 :     CHK_PTR_NULL(rootInfo);
      30            0 :     HcclUs startut = TIME_NOW();
      31              : 
      32            0 :     HcclRootHandleV2 rootHandle{};
      33            0 :     std::shared_ptr<RankInfoDetect> rankInfoDetectServer;
      34            0 :     EXCEPTION_CATCH((rankInfoDetectServer = std::make_shared<RankInfoDetect>()), return HCCL_E_MEMORY);
      35            0 :     TRY_CATCH_RETURN(rankInfoDetectServer->SetupServer(rootHandle));
      36              : 
      37            0 :     u32 rootHandleLen = sizeof(HcclRootHandleV2);
      38            0 :     CHK_PRT_RET(rootHandleLen > HCCL_ROOT_INFO_BYTES,
      39              :         HCCL_ERROR("[%s] hccl root info overflow. max length: %u, actual:%zu, identifier[%s]",
      40              :             __func__, HCCL_ROOT_INFO_BYTES, rootHandleLen, rootHandle.identifier),
      41              :         HCCL_E_INTERNAL);
      42              : 
      43              :     // 对外 HcclRootInfo 仍使用固定大小的不透明缓冲区，避免迁移实现后改变既有接口布局。
      44            0 :     s32 sRet = memcpy_s(rootInfo->internal, HCCL_ROOT_INFO_BYTES, &rootHandle, rootHandleLen);
      45            0 :     CHK_PRT_RET(sRet != EOK,
      46              :         HCCL_ERROR("[%s] memcpy root info fail. errorno[%d] params:destMaxSize[%u], count[%u]",
      47              :             __func__, sRet, HCCL_ROOT_INFO_BYTES, rootHandleLen),
      48              :         HCCL_E_MEMORY);
      49              : 
      50            0 :     s32 deviceLogicId = HrtGetDevice();
      51            0 :     s32 devPhyId = HrtGetDevicePhyIdByIndex(deviceLogicId);
      52            0 :     HCCL_RUN_INFO("HcclGetRootInfoV2 success, take time [%lld]us, rootinfo: host ip[%s] port[%u] "
      53              :                   "netMode[%s] identifier[%s], deviceLogicId[%d], devPhyId[%d]",
      54              :         DURATION_US(TIME_NOW() - startut), rootHandle.ip, rootHandle.listenPort, rootHandle.netMode.Describe().c_str(),
      55              :         rootHandle.identifier, deviceLogicId, devPhyId);
      56            0 :     return HCCL_SUCCESS;
      57            0 : }
      58              : 
      59            0 : HcclResult DetectRankTableImpl(u32 nRanks, u32 rank, const HcclRootHandleV2 &rootHandle, RankTableInfo &rankTable,
      60              :     RootInfoDetectBridge::DetectContext &detectContext)
      61              : {
      62            0 :     s32 deviceLogicId = HrtGetDevice();
      63            0 :     s32 devPhyId = HrtGetDevicePhyIdByIndex(deviceLogicId);
      64            0 :     HCCL_RUN_INFO("[%s] nRanks[%u], rank[%u] entry flat topo detect, rootinfo: host ip[%s] port[%u] "
      65              :                   "netMode[%s] identifier[%s], deviceLogicId[%d], devPhyId[%d]",
      66              :         __func__, nRanks, rank, rootHandle.ip, rootHandle.listenPort, rootHandle.netMode.Describe().c_str(),
      67              :         rootHandle.identifier, deviceLogicId, devPhyId);
      68              : 
      69            0 :     std::shared_ptr<RankInfoDetect> rankInfoDetectAgent;
      70            0 :     EXCEPTION_CATCH((rankInfoDetectAgent = std::make_shared<RankInfoDetect>()), return HCCL_E_MEMORY);
      71              : 
      72            0 :     bool hasException = false;
      73            0 :     EXCEPTION_CATCH(rankInfoDetectAgent->SetupAgent(nRanks, rank, rootHandle), hasException = true);
      74            0 :     EXCEPTION_CATCH(rankInfoDetectAgent->WaitComplete(rootHandle.listenPort, RANKINFO_DETECT_SERVER_STATUS_IDLE),
      75              :         hasException = true);
      76            0 :     CHK_PRT_RET(hasException,
      77              :         HCCL_ERROR("[%s] RankInfoDetect SetupAgent fail, identifier[%s].", __func__, rootHandle.identifier),
      78              :         HCCL_E_INTERNAL);
      79              : 
      80            0 :     rankInfoDetectAgent->GetRankTable(rankTable);
      81              :     // bridge 不暴露 RankInfoDetect 类型；调用方持有类型擦除后的 shared_ptr，
      82              :     // 使 agent 及其探测资源在后续通信域初始化完成前保持有效。
      83            0 :     detectContext = rankInfoDetectAgent;
      84            0 :     HCCL_RUN_INFO("[%s] end.", __func__);
      85            0 :     return HCCL_SUCCESS;
      86            0 : }
      87              : 
      88              : const RootInfoDetectBridge ROOT_INFO_DETECT_BRIDGE = {GetRootInfoImpl, DetectRankTableImpl};
      89              : 
      90              : // hcomm 链接 hccl_v2，并在自身装载阶段将 provider 回调注册到 hccl_v2 的槽位。
      91              : // 这样兼容入口保持原符号不变，同时不再直接依赖 RankInfoDetect 实现类。
      92              : struct RootInfoDetectBridgeRegistrar {
      93           41 :     RootInfoDetectBridgeRegistrar()
      94              :     {
      95              :         // 静态构造函数无法向加载方返回错误；重复注册由 bridge 侧日志记录。
      96           41 :         (void)RegisterRootInfoDetectBridge(ROOT_INFO_DETECT_BRIDGE);
      97           41 :     }
      98              : };
      99              : 
     100              : RootInfoDetectBridgeRegistrar g_rootInfoDetectBridgeRegistrar;
     101              : 
     102              : } // namespace
     103              : } // namespace Hccl