/* * Copyright (C) 2015 The Android Open Source Project * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ #ifndef SIMPLE_PERF_RECORD_H_ #define SIMPLE_PERF_RECORD_H_ #include <stdio.h> #include <sys/types.h> #include <memory> #include <queue> #include <string> #include <vector> #include <android-base/logging.h> #include "build_id.h" #include "perf_event.h" enum user_record_type { PERF_RECORD_USER_DEFINED_TYPE_START = 64, PERF_RECORD_ATTR = 64, PERF_RECORD_EVENT_TYPE, PERF_RECORD_TRACING_DATA, PERF_RECORD_BUILD_ID, PERF_RECORD_FINISHED_ROUND, SIMPLE_PERF_RECORD_TYPE_START = 32768, SIMPLE_PERF_RECORD_KERNEL_SYMBOL, // TODO: remove DsoRecord and SymbolRecord. SIMPLE_PERF_RECORD_DSO, SIMPLE_PERF_RECORD_SYMBOL, SIMPLE_PERF_RECORD_SPLIT, SIMPLE_PERF_RECORD_SPLIT_END, SIMPLE_PERF_RECORD_EVENT_ID, }; // perf_event_header uses u16 to store record size. However, that is not // enough for storing records like KERNEL_SYMBOL or TRACING_DATA. So define // a simpleperf_record_header struct to store record header for simpleperf // defined records (type > SIMPLE_PERF_RECORD_TYPE_START). struct simpleperf_record_header { uint32_t type; uint16_t size1; uint16_t size0; }; static_assert( sizeof(simpleperf_record_header) == sizeof(perf_event_header), "simpleperf_record_header should have the same size as perf_event_header"); struct PerfSampleIpType { uint64_t ip; }; struct PerfSampleTidType { uint32_t pid, tid; }; struct PerfSampleTimeType { uint64_t time; }; struct PerfSampleAddrType { uint64_t addr; }; struct PerfSampleIdType { uint64_t id; }; struct PerfSampleStreamIdType { uint64_t stream_id; }; struct PerfSampleCpuType { uint32_t cpu, res; }; struct PerfSamplePeriodType { uint64_t period; }; struct PerfSampleCallChainType { uint64_t ip_nr; const uint64_t* ips; }; struct PerfSampleRawType { uint32_t size; const char* data; }; struct BranchStackItemType { uint64_t from; uint64_t to; uint64_t flags; }; struct PerfSampleBranchStackType { uint64_t stack_nr; const BranchStackItemType* stack; }; struct PerfSampleRegsUserType { uint64_t abi; uint64_t reg_mask; uint64_t reg_nr; const uint64_t* regs; }; struct PerfSampleStackUserType { uint64_t size; const char* data; uint64_t dyn_size; }; struct RecordHeader { public: uint32_t type; uint16_t misc; uint32_t size; RecordHeader() : type(0), misc(0), size(0) {} explicit RecordHeader(const char* p) { auto pheader = reinterpret_cast<const perf_event_header*>(p); if (pheader->type < SIMPLE_PERF_RECORD_TYPE_START) { type = pheader->type; misc = pheader->misc; size = pheader->size; } else { auto sheader = reinterpret_cast<const simpleperf_record_header*>(p); type = sheader->type; misc = 0; size = (sheader->size1 << 16) | sheader->size0; } } void MoveToBinaryFormat(char*& p) const { if (type < SIMPLE_PERF_RECORD_TYPE_START) { auto pheader = reinterpret_cast<perf_event_header*>(p); pheader->type = type; pheader->misc = misc; CHECK_LT(size, 1u << 16); pheader->size = static_cast<uint16_t>(size); } else { auto sheader = reinterpret_cast<simpleperf_record_header*>(p); sheader->type = type; CHECK_EQ(misc, 0u); sheader->size1 = size >> 16; sheader->size0 = size & 0xffff; } p += sizeof(perf_event_header); } }; // SampleId is optional at the end of a record in binary format. Its content is // determined by sample_id_all and sample_type in perf_event_attr. To avoid the // complexity of referring to perf_event_attr each time, we copy sample_id_all // and sample_type inside the SampleId structure. struct SampleId { bool sample_id_all; uint64_t sample_type; PerfSampleTidType tid_data; // Valid if sample_id_all && PERF_SAMPLE_TID. PerfSampleTimeType time_data; // Valid if sample_id_all && PERF_SAMPLE_TIME. PerfSampleIdType id_data; // Valid if sample_id_all && PERF_SAMPLE_ID. PerfSampleStreamIdType stream_id_data; // Valid if sample_id_all && PERF_SAMPLE_STREAM_ID. PerfSampleCpuType cpu_data; // Valid if sample_id_all && PERF_SAMPLE_CPU. SampleId(); // Create the content of sample_id. It depends on the attr we use. size_t CreateContent(const perf_event_attr& attr, uint64_t event_id); // Parse sample_id from binary format in the buffer pointed by p. void ReadFromBinaryFormat(const perf_event_attr& attr, const char* p, const char* end); // Write the binary format of sample_id to the buffer pointed by p. void WriteToBinaryFormat(char*& p) const; void Dump(size_t indent) const; size_t Size() const; }; // Usually one record contains the following three parts in order in binary // format: // RecordHeader (at the head of a record, containing type and size info) // data depends on the record type // SampleId (optional part at the end of a record) // We hold the common parts (RecordHeader and SampleId) in the base class // Record, and hold the type specific data part in classes derived from Record. struct Record { RecordHeader header; SampleId sample_id; Record() : binary_(nullptr), own_binary_(false) {} explicit Record(const char* p) : header(p), binary_(p), own_binary_(false) {} Record(Record&& other); virtual ~Record() { if (own_binary_) { delete[] binary_; } } void OwnBinary() { own_binary_ = true; } uint32_t type() const { return header.type; } uint16_t misc() const { return header.misc; } uint32_t size() const { return header.size; } static uint32_t header_size() { return sizeof(perf_event_header); } bool InKernel() const { return (header.misc & PERF_RECORD_MISC_CPUMODE_MASK) == PERF_RECORD_MISC_KERNEL; } void SetTypeAndMisc(uint32_t type, uint16_t misc) { header.type = type; header.misc = misc; } void SetSize(uint32_t size) { header.size = size; } void Dump(size_t indent = 0) const; const char* Binary() const { return binary_; } virtual uint64_t Timestamp() const; virtual uint32_t Cpu() const; virtual uint64_t Id() const; protected: void UpdateBinary(const char* new_binary); virtual void DumpData(size_t) const = 0; const char* binary_; bool own_binary_; DISALLOW_COPY_AND_ASSIGN(Record); }; struct MmapRecord : public Record { struct MmapRecordDataType { uint32_t pid, tid; uint64_t addr; uint64_t len; uint64_t pgoff; }; const MmapRecordDataType* data; const char* filename; MmapRecord(const perf_event_attr& attr, const char* p); MmapRecord(const perf_event_attr& attr, bool in_kernel, uint32_t pid, uint32_t tid, uint64_t addr, uint64_t len, uint64_t pgoff, const std::string& filename, uint64_t event_id, uint64_t time = 0); void SetDataAndFilename(const MmapRecordDataType& data, const std::string& filename); protected: void DumpData(size_t indent) const override; }; struct Mmap2Record : public Record { struct Mmap2RecordDataType { uint32_t pid, tid; uint64_t addr; uint64_t len; uint64_t pgoff; uint32_t maj; uint32_t min; uint64_t ino; uint64_t ino_generation; uint32_t prot, flags; }; const Mmap2RecordDataType* data; const char* filename; Mmap2Record(const perf_event_attr& attr, const char* p); void SetDataAndFilename(const Mmap2RecordDataType& data, const std::string& filename); protected: void DumpData(size_t indent) const override; }; struct CommRecord : public Record { struct CommRecordDataType { uint32_t pid, tid; }; const CommRecordDataType* data; const char* comm; CommRecord(const perf_event_attr& attr, const char* p); CommRecord(const perf_event_attr& attr, uint32_t pid, uint32_t tid, const std::string& comm, uint64_t event_id, uint64_t time); protected: void DumpData(size_t indent) const override; }; struct ExitOrForkRecord : public Record { struct ExitOrForkRecordDataType { uint32_t pid, ppid; uint32_t tid, ptid; uint64_t time; }; const ExitOrForkRecordDataType* data; ExitOrForkRecord(const perf_event_attr& attr, const char* p); ExitOrForkRecord() : data(nullptr) {} protected: void DumpData(size_t indent) const override; }; struct ExitRecord : public ExitOrForkRecord { ExitRecord(const perf_event_attr& attr, const char* p) : ExitOrForkRecord(attr, p) {} }; struct ForkRecord : public ExitOrForkRecord { ForkRecord(const perf_event_attr& attr, const char* p) : ExitOrForkRecord(attr, p) {} ForkRecord(const perf_event_attr& attr, uint32_t pid, uint32_t tid, uint32_t ppid, uint32_t ptid, uint64_t event_id); }; struct LostRecord : public Record { uint64_t id; uint64_t lost; LostRecord(const perf_event_attr& attr, const char* p); protected: void DumpData(size_t indent) const override; }; struct SampleRecord : public Record { uint64_t sample_type; // sample_type is a bit mask determining which fields // below are valid. PerfSampleIpType ip_data; // Valid if PERF_SAMPLE_IP. PerfSampleTidType tid_data; // Valid if PERF_SAMPLE_TID. PerfSampleTimeType time_data; // Valid if PERF_SAMPLE_TIME. PerfSampleAddrType addr_data; // Valid if PERF_SAMPLE_ADDR. PerfSampleIdType id_data; // Valid if PERF_SAMPLE_ID. PerfSampleStreamIdType stream_id_data; // Valid if PERF_SAMPLE_STREAM_ID. PerfSampleCpuType cpu_data; // Valid if PERF_SAMPLE_CPU. PerfSamplePeriodType period_data; // Valid if PERF_SAMPLE_PERIOD. PerfSampleCallChainType callchain_data; // Valid if PERF_SAMPLE_CALLCHAIN. PerfSampleRawType raw_data; // Valid if PERF_SAMPLE_RAW. PerfSampleBranchStackType branch_stack_data; // Valid if PERF_SAMPLE_BRANCH_STACK. PerfSampleRegsUserType regs_user_data; // Valid if PERF_SAMPLE_REGS_USER. PerfSampleStackUserType stack_user_data; // Valid if PERF_SAMPLE_STACK_USER. SampleRecord(const perf_event_attr& attr, const char* p); SampleRecord(const perf_event_attr& attr, uint64_t id, uint64_t ip, uint32_t pid, uint32_t tid, uint64_t time, uint32_t cpu, uint64_t period, const std::vector<uint64_t>& ips); void ReplaceRegAndStackWithCallChain(const std::vector<uint64_t>& ips); uint64_t Timestamp() const override; uint32_t Cpu() const override; uint64_t Id() const override; uint64_t GetValidStackSize() const { // If stack_user_data.dyn_size == 0, it may be because the kernel misses // the patch to update dyn_size, like in N9 (See b/22612370). So assume // all stack data is valid if dyn_size == 0. if (stack_user_data.dyn_size == 0) { return stack_user_data.size; } return stack_user_data.dyn_size; } protected: void DumpData(size_t indent) const override; }; // BuildIdRecord is defined in user-space, stored in BuildId feature section in // record file. struct BuildIdRecord : public Record { uint32_t pid; BuildId build_id; const char* filename; explicit BuildIdRecord(const char* p); BuildIdRecord(bool in_kernel, pid_t pid, const BuildId& build_id, const std::string& filename); protected: void DumpData(size_t indent) const override; }; struct KernelSymbolRecord : public Record { uint32_t kallsyms_size; const char* kallsyms; explicit KernelSymbolRecord(const char* p); explicit KernelSymbolRecord(const std::string& kallsyms); protected: void DumpData(size_t indent) const override; }; struct DsoRecord : public Record { uint64_t dso_type; uint64_t dso_id; uint64_t min_vaddr; const char* dso_name; explicit DsoRecord(const char* p); DsoRecord(uint64_t dso_type, uint64_t dso_id, const std::string& dso_name, uint64_t min_vaddr); protected: void DumpData(size_t indent) const override; }; struct SymbolRecord : public Record { uint64_t addr; uint64_t len; uint64_t dso_id; const char* name; explicit SymbolRecord(const char* p); SymbolRecord(uint64_t addr, uint64_t len, const std::string& name, uint64_t dso_id); protected: void DumpData(size_t indent) const override; }; struct TracingDataRecord : public Record { uint32_t data_size; const char* data; explicit TracingDataRecord(const char* p); explicit TracingDataRecord(const std::vector<char>& tracing_data); protected: void DumpData(size_t indent) const override; }; struct EventIdRecord : public Record { uint64_t count; struct EventIdData { uint64_t attr_id; uint64_t event_id; } const* data; explicit EventIdRecord(const char* p); explicit EventIdRecord(const std::vector<uint64_t>& data); protected: void DumpData(size_t indent) const override; }; // UnknownRecord is used for unknown record types, it makes sure all unknown // records are not changed when modifying perf.data. struct UnknownRecord : public Record { const char* data; explicit UnknownRecord(const char* p); protected: void DumpData(size_t indent) const override; }; // Read record from the buffer pointed by [p]. But the record doesn't own // the buffer. std::unique_ptr<Record> ReadRecordFromBuffer(const perf_event_attr& attr, uint32_t type, const char* p); // Read record from the buffer pointed by [p]. And the record owns the buffer. std::unique_ptr<Record> ReadRecordFromOwnedBuffer(const perf_event_attr& attr, uint32_t type, const char* p); // Read records from the buffer pointed by [buf]. None of the records own // the buffer. std::vector<std::unique_ptr<Record>> ReadRecordsFromBuffer( const perf_event_attr& attr, const char* buf, size_t buf_size); // Read one record from the buffer pointed by [p]. But the record doesn't // own the buffer. std::unique_ptr<Record> ReadRecordFromBuffer(const perf_event_attr& attr, const char* p); // RecordCache is a cache used when receiving records from the kernel. // It sorts received records based on type and timestamp, and pops records // in sorted order. Records from the kernel need to be sorted because // records may come from different cpus at the same time, and it is affected // by the order in which we collect records from different cpus. // RecordCache pushes records and pops sorted record online. It uses two checks // to help ensure that records are popped in order. Each time we pop a record A, // it is the earliest record among all records in the cache. In addition, we // have checks for min_cache_size and min_time_diff. For min_cache_size check, // we check if the cache size >= min_cache_size, which is based on the // assumption that if we have received (min_cache_size - 1) records after // record A, we are not likely to receive a record earlier than A. For // min_time_diff check, we check if record A is generated min_time_diff ns // earlier than the latest record, which is based on the assumption that if we // have received a record for time t, we are not likely to receive a record for // time (t - min_time_diff) or earlier. class RecordCache { public: explicit RecordCache(bool has_timestamp, size_t min_cache_size = 1000u, uint64_t min_time_diff_in_ns = 1000000u); ~RecordCache(); void Push(std::unique_ptr<Record> record); void Push(std::vector<std::unique_ptr<Record>> records); std::unique_ptr<Record> Pop(); std::vector<std::unique_ptr<Record>> PopAll(); std::unique_ptr<Record> ForcedPop(); private: struct RecordWithSeq { uint32_t seq; Record* record; RecordWithSeq(uint32_t seq, Record* record) : seq(seq), record(record) {} bool IsHappensBefore(const RecordWithSeq& other) const; }; struct RecordComparator { bool operator()(const RecordWithSeq& r1, const RecordWithSeq& r2); }; bool has_timestamp_; size_t min_cache_size_; uint64_t min_time_diff_in_ns_; uint64_t last_time_; uint32_t cur_seq_; std::priority_queue<RecordWithSeq, std::vector<RecordWithSeq>, RecordComparator> queue_; }; #endif // SIMPLE_PERF_RECORD_H_