Line data    Source code

       1              : /**
       2              :  * Copyright (c) 2025 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              : #ifndef EXCEPTION_DUMPER_H
      11              : #define EXCEPTION_DUMPER_H
      12              : #include <string>
      13              : #include <mutex>
      14              : #include <deque>
      15              : #include "runtime/rt.h"
      16              : #include "adump_pub.h"
      17              : #include "dump_operator.h"
      18              : #include "dump_setting.h"
      19              : #include "adump_api.h"
      20              : #include "path.h"
      21              : 
      22              : namespace Adx {
      23              : class ExceptionDumper {
      24              : public:
      25           58 :     ExceptionDumper() = default;
      26              :     ~ExceptionDumper();
      27              :     int32_t ExceptionDumperInit(DumpType dumpType, const DumpConfig& dumpConfig);
      28              :     inline bool GetExceptionStatus() const;
      29              :     inline bool GetArgsExceptionStatus() const;
      30              :     inline bool GetCoredumpStatus() const;
      31              :     void SetDumpPath(const std::string& dumpPath);
      32              :     std::string CreateExtraDumpPath();
      33              :     std::string CreateDeviceDumpPath(uint32_t deviceId) const;
      34              :     const char* GetExtraDumpCPath() const;
      35              :     void AddDumpOperator(const OperatorInfo& opInfo);
      36              :     void AddDumpOperatorV2(const OperatorInfoV2& opInfo);
      37              :     int32_t DelDumpOperator(uint32_t deviceId, uint32_t streamId);
      38              :     int32_t DumpException(const rtExceptionInfo& exception);
      39              :     int32_t GetExceptionDumpPath(std::string& path);
      40              :     int32_t SaveExceptionInfo(
      41              :         const std::string& fileName, const std::string& userTag, const std::vector<TensorInfo>& tensors);
      42              :     void ExceptionModeDowngrade();
      43              :     bool IsRepeatEnableException(DumpType type, const DumpConfig& dumpConfig);
      44              :     bool IsEnabledExceptionDump() const;
      45              :     int32_t RegisterExceptionDumpCallback(ExceptionDumpCallback callback);
      46              :     int32_t UnregisterExceptionDumpCallback(ExceptionDumpCallback callback);
      47              : 
      48              : #ifdef __ADUMP_LLT
      49              :     void Reset();
      50              : #endif
      51              : 
      52              : private:
      53              :     std::string CreateDumpPath(Path& dumpPath) const;
      54              :     bool FindExceptionOperator(const rtExceptionInfo& exception, DumpOperator& excOp);
      55              :     int32_t DumpArgsException(const rtExceptionInfo& exception, const std::string& dumpPath);
      56              :     int32_t DumpArgsExceptionInner(const rtExceptionInfo& exception, const std::string& dumpPath);
      57              :     int32_t DumpArgsExceptionDefault(const rtExceptionInfo& exception, const std::string& dumpPath);
      58              :     int32_t DumpArgsExceptionFastRecovery(const rtExceptionInfo& exception) const;
      59              : 
      60              :     int32_t InvokeCallbacks(
      61              :         const rtExceptionInfo& exception, std::vector<ExceptionDumpInfo>& dumpInfos, ExceptionDumpMode& finalMode);
      62              :     void DumpCallbackData(
      63              :         const rtExceptionInfo& exception, const std::vector<ExceptionDumpInfo>& dumpInfos, const std::string& dumpPath);
      64              :     int32_t DumpNormalException(const rtExceptionInfo& exception, const std::string& dumpPath);
      65              :     int32_t DumpNormalExceptionDefault(const rtExceptionInfo& exception, const std::string& dumpPath);
      66              :     int32_t DumpDetailException(const rtExceptionInfo& exception, const std::string& dumpPath);
      67              :     int32_t DumpExceptionWithCallbacks(
      68              :         const rtExceptionInfo& exception, const std::string& dumpPath,
      69              :         int32_t (ExceptionDumper::*defaultDump)(const rtExceptionInfo&, const std::string&));
      70              :     // 在符号化(DumpErrorSymbols)之前，提前同步把 _host.o 落盘，供后续按错误 PC 解析源码行使用。best-effort。
      71              :     void DumpHostKernelBinBeforeSymbolize(const rtExceptionInfo& exception, const std::string& dumpPath) const;
      72              :     int32_t LoadTensorPluginLib();
      73              :     bool InitArgsExceptionMemory() const;
      74              :     std::string GetDumpSceneName() const;
      75              :     void Exit() const;
      76              :     bool coredumpStatus_{false};
      77              :     bool coredumpEnableComplete_{true};
      78              :     bool exceptionStatus_{false};
      79              :     bool argsExceptionStatus_{false};
      80              :     bool destructionFlag_{false};
      81              :     std::string dumpPath_;
      82              :     std::string extraDumpPath_;
      83              :     DumpSetting setting_;
      84              :     mutable std::mutex mutex_;
      85              :     std::deque<DumpOperator> agingOperators_;
      86              :     std::map<uint32_t, std::map<uint32_t, std::deque<DumpOperator>>> residentOperators_;
      87              :     std::vector<ExceptionDumpCallback> callbacks_;
      88              :     std::recursive_mutex callbackMutex_;
      89              : };
      90              : 
      91            8 : inline bool ExceptionDumper::GetExceptionStatus() const { return exceptionStatus_; }
      92              : 
      93           40 : inline bool ExceptionDumper::GetArgsExceptionStatus() const { return argsExceptionStatus_; }
      94              : 
      95           72 : inline bool ExceptionDumper::GetCoredumpStatus() const { return coredumpStatus_; }
      96              : } // namespace Adx
      97              : #endif // EXCEPTION_DUMPER_H