已关闭
[Bug-Report|缺陷反馈]: A3 createEndpoint报错 #860
hu-yiliang11创建于  21 天前关闭于  21 天前
hu-yiliang11
21 天前 创建

Thanks for sending an issue! Please fill in the following template to help quickly solve your problem.

Describe the current behavior / 问题描述 (Mandatory / 必填)

单边通信调用CreateEndpoint报错,芯片类型不支持

Environment / 环境信息 (Mandatory / 必填)

A3 Ascend910_9382 CANN hcomm master分支

Steps to reproduce the issue / 重现步骤 (Mandatory / 必填)

CreateEndpoint 创建ROCE链路

Describe the expected behavior / 预期结果 (Mandatory / 必填)

CreateEndpoint 无报错

545: [ERROR] HCCL(3374811,hixlCS_N2_dfx_compatibility_002_0.bin):2026-09-07-14:19:25.305.467 [orion_adapter_rts.cc:75][3374811][Get][DeviceType]errNo[0x000000000500000f] rtGetSocVersion get illegal chipver, chip_ver[Ascend910_9382].
546: [ERROR] HCCL(3374811,hixlCS_N2_dfx_compatibility_002_0.bin):2026-09-07-14:19:25.305.814 [cpu_roce_endpoint.cc:98] [3374811][Init]Failed, exception caught:Runtime api exception: [Get][DeviceType]errNo[0x000000000500000f] rtGetSocVersion get illegal chipver, chip_ver[Ascend910_9382].
547: [ERROR] HCCL(3374811,hixlCS_N2_dfx_compatibility_002_0.bin):2026-09-07-14:19:25.305.821 [hcomm_endpoint_c_adpt.cc:195] [3374811]call endpointPtr->Init failed
550: [ERROR] HCCL(3374811,hixlCS_N2_dfx_compatibility_002_0.bin):2026-09-07-14:19:25.305.830 [hcomm_endpoint_c_adpt.cc:367] [3374811][HcommEndpointCreate]call trace: hcclRet -> 4
552: [ERROR] GE(3374811,hixlCS_N2_dfx_compatibility_002_0.bin):2026-09-07-14:19:25.305.871 [endpoint.cc:100][HIXL] 3374811 Initialize: ErrorNo: 4 Call hccl api:HcommProxy::EndpointCreate(&endpoint_, &handle_) failed. HcommEndpointCreate failed, endpoint=[protocol=roce, addr=IPv4:10.0.103.21, hostId=0]. Please check whether the endpoint address is valid and available in the current environment.
554: [ERROR] GE(3374811,hixlCS_N2_dfx_compatibility_002_0.bin):2026-09-07-14:19:25.305.882 [hixl_cs_client.cc:292][HIXL] 3374811 InitBaseClient: ErrorNo: 503900 [HixlClient] Failed to initialize src endpoint. Check Config: [Loc:1, protocol:roce, AddrVal:0x1567000a]
556: [ERROR] GE(3374811,hixlCS_N2_dfx_compatibility_002_0.bin):2026-09-07-14:19:25.305.890 [hixl_cs_client.cc:370][HIXL] 3374811 Create: ErrorNo: 503900 [HixlClient] InitBaseClient failed
558: [ERROR] GE(3374811,hixlCS_N2_dfx_compatibility_002_0.bin):2026-09-07-14:19:25.305.898 [hixl_cs.cc:102][HIXL] 3374811 HixlCSClientCreate: ErrorNo: 503900 Failed to create hixl cs client, server_ip:10.0.103.21, server_port:22946
564: [ERROR] HCCL(3374811,hixlCS_N2_dfx_compatibility_002_0.bin):2026-09-07-14:19:25.305.958 [hcomm_endpoint_c_adpt.cc:409] [3374811][HcommEndpointDestroy] endpoint not found, endpointHandle[0x0]
566: [ERROR] GE(3374811,hixlCS_N2_dfx_compatibility_002_0.bin):2026-09-07-14:19:25.305.968 [endpoint.cc:136][HIXL] 3374811 Finalize: ErrorNo: 503900 Call api:HcommEndpointDestroy failed, ret:6, ep_handle:(nil)

Special notes for this issue/备注 (Optional / 选填)

likedislike
LLeewis成员
21 天前 关联了看板:HCOMM
Leewis成员
21 天前 评论:

/assign @gcw_NcEfY7mt

likedislike
CANN-robotCANN-robot成员
21 天前 将 gcw_NcEfY7mt 设为负责人
Leewis成员
21 天前 评论:

针对 A3 芯片的 SOC 版本映射表进行修正, 相关PR已修改上库;https://gitcode.com/cann/hcomm/pull/5117

likedislike
LLeewis成员
21 天前 issue类型由 任务 改变为 缺陷
LLeewis成员
21 天前 issue状态由 待办的 改变为 已完成
LLeewis成员
21 天前 关闭了 issue
CANN-robotCANN-robot成员
21 天前 添加了label:resolved
LLeewis成员
12 天前 移除了看板:HCOMM