对于一个应用冻屏(AppFreeze)问题,可通过AppFreeze日志和其增强日志(采样栈)确认是否为主线程阻塞类型:
对比告警和卡死两次事件的主线程堆栈是否一致。两次事件包括:THREAD_BLOCK_3S和THREAD_BLOCK_6S、THREAD_BLOCK_3S和APP_INPUT_BLOCK、LIFECYCLE_HALF_TIMEOUT和LIFECYCLE_TIMEOUT。若堆栈一致,通常情况下表明是主线程阻塞类型,建议结合增强日志进行更准确的判定。
若AppFreeze日志中仅有APP_INPUT_BLOCK事件的单次主线程堆栈,则无法判定为主线程阻塞类型。
对比AppFreeze日志中的主线程堆栈以及增强日志中的主线程采样栈是否一致,一致则表明是主线程阻塞类型。
此外可以参考堆栈信息章节中线程状态的介绍来判断线程在冻屏发生过程中是否在运行。
根因描述
应用的主线程业务等锁导致阻塞。
问题分析思路
若持有锁A的线程即主线程,或者持有锁A的线程等待的锁B被主线程或者其他等待锁A的线程持有,则该问题类型为死锁。
关键字
关注主线程堆栈栈顶共享库和符号是否匹配如下关键字。
共享库:
ld-musl-aarch64.so.1
符号:
案例分析
案例一:子线程和主线程AB锁互锁导致主线程阻塞
问题现象
当用户触发业务代码调用后,应用进入无响应状态,一段时间后应用退出,并生成冻屏日志文件。
问题分析
证据1:主线程堆栈关键栈帧
Tid:18728, Name:pfreezeanalysis state=S, utime=55, stime=15, priority=-54, nice=-10, clk=100 #00 pc 00000000001de3f0 /system/lib/ld-musl-aarch64.so.1(__timedwait_cp+156)(ede7850a729ebc7f50c03216608d9277) #01 pc 00000000001e4488 /system/lib/ld-musl-aarch64.so.1(__pthread_mutex_timedlock_inner+536)(ede7850a729ebc7f50c03216608d9277) #02 pc 000000000000a6c8 /data/storage/el1/bundle/libs/arm64/libentry.so(f5f54ab64da94326c537007dfa9fbf43b7db6ea1) #03 pc 000000000007010c /system/lib64/platformsdk/libace_napi.z.so(panda::JSValueRef ArkNativeFunctionCallBack<true>(panda::JsiRuntimeCallInfo*)+284)(c41c28aa2f0cdd937fc9063a8b806514) #04 pc 0000000000e8ad70 /system/lib64/module/arkcompiler/stub.an(RTStub_AsmInterpreterEntry+484) #05 pc 0000000000e8abc8 /system/lib64/module/arkcompiler/stub.an(RTStub_AsmInterpreterEntry+60)
#02栈帧表明等锁的业务方是libentry.so,结合代码继续分析。
证据2:故障栈帧所在的代码

代码行指向pthread_mutex_lock(&g_deadlockMutexSecond),即等待锁g_deadlockMutexSecond。继续分析g_deadlockMutexSecond的持锁情况。
证据3:DeadlockOtherThread的实现代码

DeadlockOtherThread是在主线程创建的子线程中调用的。子线程进入后先获取锁g_deadlockMutexSecond,然后等待1s再获取锁g_deadlockMutexFirst。创建子线程前主线程就已经获取g_deadlockMutexFirst,创建子线程后主线程又去获取g_deadlockMutexSecond。主线程和子线程发生死锁,主线程一直等锁阻塞。
案例二:两个子线程死锁,主线程无法等到子线程结束导致阻塞
问题现象
当用户触发业务代码调用后,应用进入无响应状态,一段时间后应用退出,并生成冻屏日志文件。
问题分析
证据1:主线程堆栈关键栈帧
Tid:30591, Name:pfreezeanalysis state=S, utime=21, stime=14, priority=-54, nice=-10, clk=100 #00 pc 00000000001e3578 /system/lib/ld-musl-aarch64.so.1(pthread_join+136)(ede7850a729ebc7f50c03216608d9277) #01 pc 000000000000aae0 /data/storage/el1/bundle/libs/arm64/libentry.so(f5f54ab64da94326c537007dfa9fbf43b7db6ea1) #02 pc 000000000007010c /system/lib64/platformsdk/libace_napi.z.so(panda::JSValueRef ArkNativeFunctionCallBack<true>(panda::JsiRuntimeCallInfo*)+284)(c41c28aa2f0cdd937fc9063a8b806514) #03 pc 0000000000e8ad70 /system/lib64/module/arkcompiler/stub.an(RTStub_AsmInterpreterEntry+484) #04 pc 0000000000e8abc8 /system/lib64/module/arkcompiler/stub.an(RTStub_AsmInterpreterEntry+60)
#01栈帧表明等锁的业务方是libentry.so,结合代码继续分析。
证据2:故障栈帧所在的代码

代码行指向pthread_join(t1, nullptr),即等待子线程结束。继续分析子线程的实现。
证据3:两个子线程的实现代码

线程A进入后先获取锁g_deadlockMutexFirst,然后等待1ms再获取锁g_deadlockMutexSecond。线程B则反过来,先获取锁g_deadlockMutexSecond再等待1ms获取锁g_deadlockMutexFirst。这样两个子线程进入后都获取到了对方想要获取的锁,从而造成死锁。两个子线程都无法结束,导致主线程等待阻塞。
预防建议
优化锁的使用策略,减少锁竞争,建议使用无锁数据结构。
根因描述
应用主线程业务向系统服务或其他进程发起调用同步Binder接口,对端长时间未返回,导致阻塞。
问题分析思路
关键字
关注主线程堆栈共享库和符号是否匹配如下关键字。
共享库:
libipc_common.z.so
符号:
WriteBinder
案例分析
案例:系统音频服务未及时响应导致主线程阻塞
问题现象
当用户触发业务代码调用后,应用进入无响应状态,一段时间后应用退出,并生成冻屏日志文件。
问题分析
证据1:主线程堆栈关键栈帧
Tid:41777, Name:pfreezeanalysis state=S, utime=118, stime=49, priority=-14, nice=-10, clk=100 #00 pc 0000000000197768 /system/lib/ld-musl-aarch64.so.1(ioctl+172)(8c2ca4e308a1de9ed1e6b4ccc14c7443) #01 pc 000000000000f0ac /system/lib64/platformsdk/libipc_common.z.so(OHOS::BinderConnector::WriteBinder(unsigned long, void*)+120)(21def93073c96c4161bfa61b948c2957) #02 pc 0000000000074934 /system/lib64/platformsdk/libipc_single.z.so(OHOS::BinderInvoker::TransactWithDriver(bool)+284)(5a12a4b7e05d71d18ea20b9f1e1aed8a) #03 pc 0000000000073154 /system/lib64/platformsdk/libipc_single.z.so(OHOS::BinderInvoker::WaitForCompletion(OHOS::MessageParcel*)+124)(5a12a4b7e05d71d18ea20b9f1e1aed8a) #04 pc 0000000000072624 /system/lib64/platformsdk/libipc_single.z.so(OHOS::BinderInvoker::SendRequest(int, unsigned int, OHOS::MessageParcel&, OHOS::MessageParcel&, OHOS::MessageOption&)+620)(5a12a4b7e05d71d18ea20b9f1e1aed8a) #05 pc 0000000000049d24 /system/lib64/platformsdk/libipc_single.z.so(OHOS::IPCObjectProxy::SendRequestInner(bool, unsigned int, OHOS::MessageParcel&, OHOS::MessageParcel&, OHOS::MessageOption&)+272)(5a12a4b7e05d71d18ea20b9f1e1aed8a) #06 pc 000000000004a73c /system/lib64/platformsdk/libipc_single.z.so(OHOS::IPCObjectProxy::SendRequest(unsigned int, OHOS::MessageParcel&, OHOS::MessageParcel&, OHOS::MessageOption&)+216)(5a12a4b7e05d71d18ea20b9f1e1aed8a) #07 pc 0000000000040dd4 /system/lib64/libaudio_policy_client_idl_interface.z.so(OHOS::AudioStandard::AudioPolicyProxy::GetAudioScene(int&)+320)(b95b8a9a8cca6ee70d60967fc1751cc5) #08 pc 000000000008cfec /system/lib64/platformsdk/libaudio_client.z.so(OHOS::AudioStandard::AudioSystemManager::GetAudioScene() const+44)(0e4839c7e3f20d5152f89c053c31d414) #09 pc 00000000000e0c44 /system/lib64/module/multimedia/libaudio.z.so(OHOS::AudioStandard::NapiAudioManager::GetAudioSceneSync(napi_env__*, napi_callback_info__*) (.cfi)+228)(3201d994d8d5f9e8fdf1c9aa5ed39509) #10 pc 0000000000067c90 /system/lib64/platformsdk/libace_napi.z.so(panda::JSValueRef ArkNativeFunctionCallBack<true>(panda::JsiRuntimeCallInfo*)+288)(76bc1ed675edcc5f429a976d6c9b955d) #11 pc 0000000000e179f4 /system/lib64/module/arkcompiler/stub.an(RTStub_PushCallArgsAndDispatchNative+40) #12 pc 000000000046b900 /system/lib64/module/arkcompiler/stub.an(BCStub_HandleCallthis0Imm8V8StwCopy+392) #13 at triggerBinderInterfaceBlockSync (entry|entry|1.0.0|src/main/ets/pages/page_second/page_third_appfreeze/appfreeze_threadblock.ts:147:43)
堆栈的#13栈帧表明应用侧业务函数是triggerBinderInterfaceBlockSync(),#09栈帧表明业务调用的接口是系统音频服务提供的GetAudioSceneSync()接口。由于系统音频服务未及时响应返回,主线程阻塞在同步Binder通信上。由于对端进程是系统音频服务,不属于故障应用的业务,建议反馈给系统侧优化。若对端进程是故障应用的业务,可以按照下述步骤继续分析。
证据2:BinderCatcher信息
41777:41777 to 784:801 code 1b wait:6.446522656 s frz_state:1, ns:-1:-1 to -1:-1, debug:41777:41777 to 784:801, active_code:0, active_thread=0, pending_async_proc=0
BinderCatcher信息表明主线程等待对端进程响应的时间已超过6s,对端进程的进程号是784,IPC线程号是801。
证据3:对端进程IPC线程的堆栈关键栈帧
Tid:801, Name:OS_IPC_0_801 state=S, utime=14, stime=9, priority=20, nice=0, clk=100 #00 pc 0000000000197768 /system/lib/ld-musl-aarch64.so.1(ioctl+172)(8c2ca4e308a1de9ed1e6b4ccc14c7443) #01 pc 000000000000f0ac /system/lib64/platformsdk/libipc_common.z.so(OHOS::BinderConnector::WriteBinder(unsigned long, void*)+120)(21def93073c96c4161bfa61b948c2957) #02 pc 0000000000074934 /system/lib64/platformsdk/libipc_single.z.so(OHOS::BinderInvoker::TransactWithDriver(bool)+284)(5a12a4b7e05d71d18ea20b9f1e1aed8a) #03 pc 0000000000074d24 /system/lib64/platformsdk/libipc_single.z.so(OHOS::BinderInvoker::StartWorkLoop()+100)(5a12a4b7e05d71d18ea20b9f1e1aed8a) #04 pc 0000000000077958 /system/lib64/platformsdk/libipc_single.z.so(OHOS::BinderInvoker::JoinThread(bool)+80)(5a12a4b7e05d71d18ea20b9f1e1aed8a) #05 pc 000000000006c248 /system/lib64/platformsdk/libipc_single.z.so(OHOS::IPCWorkThread::ThreadHandler(void*)+1056)(5a12a4b7e05d71d18ea20b9f1e1aed8a) #06 pc 00000000001de740 /system/lib/ld-musl-aarch64.so.1(8c2ca4e308a1de9ed1e6b4ccc14c7443) #07 pc 00000000000b59ec /system/lib/ld-musl-aarch64.so.1(8c2ca4e308a1de9ed1e6b4ccc14c7443)
IPC线程的堆栈表明服务未响应的根因也是同步Binder接口调用阻塞,则按本章节的分析思路继续分析。
预防建议
将同步Binder调用移至子线程,或使用异步接口。
根因描述
应用主线程业务向系统服务或其他进程发起调用同步Binder接口,对端Binder线程耗尽,导致调用长时间未返回并处于阻塞状态。
问题分析思路
关键字
关注主线程堆栈共享库和符号是否匹配如下关键字。
共享库:
libipc_common.z.so
符号:
WriteBinder
案例分析
案例:系统音频服务Binder线程耗尽未及时响应导致主线程阻塞
问题现象
当用户触发业务代码调用后,应用进入无响应状态,一段时间后应用退出,并生成冻屏日志文件。
问题分析
证据1:主线程堆栈关键栈帧
Tid:17052, Name:pfreezeanalysis state=S, utime=203, stime=76, priority=-14, nice=-10, clk=100 #00 pc 000000000019514c /system/lib/ld-musl-aarch64.so.1(ioctl+188)(9dcef83da65dee0250e2b6dbef45e66a) #01 pc 000000000000f030 /system/lib64/platformsdk/libipc_common.z.so(OHOS::BinderConnector::WriteBinder(unsigned long, void*)+120)(c3d820e943670c03802f3c6a6f6c02f7) #02 pc 0000000000071d64 /system/lib64/platformsdk/libipc_single.z.so(OHOS::BinderInvoker::TransactWithDriver(bool)+284)(725e65c2222991d0c1d55e2cf0cc2738) #03 pc 0000000000070764 /system/lib64/platformsdk/libipc_single.z.so(OHOS::BinderInvoker::WaitForCompletion(OHOS::MessageParcel*)+124)(725e65c2222991d0c1d55e2cf0cc2738) #04 pc 000000000006fc34 /system/lib64/platformsdk/libipc_single.z.so(OHOS::BinderInvoker::SendRequest(int, unsigned int, OHOS::MessageParcel&, OHOS::MessageParcel&, OHOS::MessageOption&)+620)(725e65c2222991d0c1d55e2cf0cc2738) #05 pc 0000000000048d14 /system/lib64/platformsdk/libipc_single.z.so(OHOS::IPCObjectProxy::SendRequestInner(bool, unsigned int, OHOS::MessageParcel&, OHOS::MessageParcel&, OHOS::MessageOption&)+272)(725e65c2222991d0c1d55e2cf0cc2738) #06 pc 000000000004972c /system/lib64/platformsdk/libipc_single.z.so(OHOS::IPCObjectProxy::SendRequest(unsigned int, OHOS::MessageParcel&, OHOS::MessageParcel&, OHOS::MessageOption&)+216)(725e65c2222991d0c1d55e2cf0cc2738) #07 pc 000000000003ee60 /system/lib64/libaudio_policy_client_idl_interface.z.so(OHOS::AudioStandard::AudioPolicyProxy::GetAudioScene(int&)+320)(16a5f13d157f85d0ce2574baf96e445a) #08 pc 00000000000896c8 /system/lib64/platformsdk/libaudio_client.z.so(OHOS::AudioStandard::AudioSystemManager::GetAudioScene() const+44)(fba17e35e3611dc6050105edd9d5bf7c) #09 pc 00000000000d3314 /system/lib64/module/multimedia/libaudio.z.so(OHOS::AudioStandard::NapiAudioManager::GetAudioSceneSync(napi_env__*, napi_callback_info__*) (.cfi)+228)(35129c60e8427727fa0ca4566d0f6524) #10 pc 0000000000066120 /system/lib64/platformsdk/libace_napi.z.so(panda::JSValueRef ArkNativeFunctionCallBack<true>(panda::JsiRuntimeCallInfo*)+288)(44ba8b408e7efd513bdc33ebeab8b671) #11 pc 0000000000e1537c /system/lib64/module/arkcompiler/stub.an(RTStub_PushCallArgsAndDispatchNative+40) #12 pc 0000000000466904 /system/lib64/module/arkcompiler/stub.an(BCStub_HandleCallthis0Imm8V8StwCopy+392) #13 at triggerBinderInterfaceBlockSync (entry|entry|1.0.0|src/main/ets/pages/page_second/page_third_appfreeze/appfreeze_threadblock.ts:216:43)
堆栈的#13栈帧表明应用侧业务函数是triggerBinderInterfaceBlockSync(),#09栈帧表明业务调用的接口是系统音频服务提供的GetAudioSceneSync()接口。由于系统音频服务未及时响应返回,主线程阻塞在同步Binder通信上。由于对端进程是系统音频服务,不属于故障应用的业务,建议反馈给系统侧优化。若对端进程是故障应用的业务,可以按照下述步骤继续分析。
证据2:BinderCatcher信息
17052:17052 to 772:0 code 1b wait:6.74904063 s frz_state:1, ns:-1:-1 to -1:-1, debug:17052:17052 to 772:0, active_code:0, active_thread=0, pending_async_proc=0
BinderCatcher信息表明主线程等待对端进程响应的时间已超过6s,对端进程的进程号是772,IPC线程号是0,表明对端进程Binder线程耗尽,无法分配出IPC线程。
预防建议
根因描述
应用主线程在进行同步耗时I/O操作时被阻塞。
问题分析思路
关键字
关注主线程堆栈栈顶共享库和符号是否匹配如下关键字。
共享库:
ld-musl-aarch64.so.1
符号:
案例分析
案例:文件读取操作耗时导致主线程阻塞
问题现象
当用户触发业务代码调用后,应用进入无响应状态,一段时间后应用退出,并生成冻屏日志文件。
问题分析
证据:主线程堆栈关键栈帧
Tid:19502, Name:pfreezeanalysis state=R, utime=189, stime=635, priority=-14, nice=-10, clk=100 #00 pc 00000000001ead94 /system/lib/ld-musl-aarch64.so.1(pread+76)(8c2ca4e308a1de9ed1e6b4ccc14c7443) #01 pc 00000000000121e0 /system/lib/libprocess.so(open+556) #02 pc 000000000001e108 /system/lib64/platformsdk/libuv.so(uv__fs_work.llvm.255697381303559595+204)(dd4ce8a088606e6444f0dceac8540868) #03 pc 0000000000020998 /system/lib64/platformsdk/libuv.so(uv_fs_open+176)(dd4ce8a088606e6444f0dceac8540868) #04 pc 0000000000109738 /system/lib64/module/file/libfs.z.so(OHOS::FileManagement::ModuleFileIO::OpenFileByPath(std::__h::basic_string<char, std::__h::char_traits<char>, std::__h::allocator<char>> const&, unsigned int)+148)(34d72fdf9d62dd4b9396c4c393f51d9f) #05 pc 0000000000107988 /system/lib64/module/file/libfs.z.so(OHOS::FileManagement::ModuleFileIO::Open::Sync(napi_env__*, napi_callback_info__*) (.cfi)+788)(34d72fdf9d62dd4b9396c4c393f51d9f) #06 pc 0000000000067c90 /system/lib64/platformsdk/libace_napi.z.so(panda::JSValueRef ArkNativeFunctionCallBack<true>(panda::JsiRuntimeCallInfo*)+288)(76bc1ed675edcc5f429a976d6c9b955d) #07 pc 0000000000e179f4 /system/lib64/module/arkcompiler/stub.an(RTStub_PushCallArgsAndDispatchNative+40) #08 pc 000000000046c00c /system/lib64/module/arkcompiler/stub.an(BCStub_HandleCallthis2Imm8V8V8V8StwCopy+432) #09 at write (entry|entry|1.0.0|src/main/ets/pages/page_second/page_third_appfreeze/appfreeze_threadblock_io.ts:182:31) #10 at anonymous (entry|entry|1.0.0|src/main/ets/pages/page_second/page_third_appfreeze/appfreeze_threadblock_io.ts:155:26)
栈顶表明主线程阻塞在读取操作上,#09栈帧表明I/O操作的业务在write函数中。
预防建议
将耗时I/O操作移至子线程执行,避免主线程被阻塞。
根因描述
应用主线程业务触发长时间GC操作阻塞。
问题分析思路
关键字
关注主线程堆栈共享库和符号是否匹配如下关键字。
共享库:
libark_jsruntime.so
符号:
FullGC
案例分析
案例:FullGC耗时导致主线程阻塞
问题现象
当用户触发业务代码调用后,应用进入无响应状态,一段时间后应用退出,并生成冻屏日志文件。
问题分析
证据:主线程堆栈关键栈帧
Tid:22887, Name:pfreezeanalysis state=R, utime=628, stime=55, priority=-14, nice=-10, clk=100 #00 pc 0000000000015890 /system/lib/libprocess.so(trap_handler+1232) #01 pc 00000000001edecc /system/lib/ld-musl-aarch64.so.1(signal_chain_handler+864)(8c2ca4e308a1de9ed1e6b4ccc14c7443) #02 pc 0000000000001d30 [shmm] #03 pc 0000000000001480 [shmm](__kernel_rt_sigreturn+0) #04 pc 000000000020e9a8 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::FullGCMarkRootVisitor::VisitRoot(panda::ecmascript::Root, panda::ecmascript::ObjectSlot)+548)(4025597d50430dc8ac5e5f0b44678df8) #05 pc 00000000008ec928 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::FullGCMarkRootVisitor::VisitRangeRoot(panda::ecmascript::Root, panda::ecmascript::ObjectSlot, panda::ecmascript::ObjectSlot)+64)(4025597d50430dc8ac5e5f0b44678df8) #06 pc 0000000000459134 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::EcmaVM::IterateSTWRoots(panda::ecmascript::RootVisitor&)+384)(4025597d50430dc8ac5e5f0b44678df8) #07 pc 0000000000458a6c /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::EcmaVM::Iterate(panda::ecmascript::RootVisitor&)+24)(4025597d50430dc8ac5e5f0b44678df8) #08 pc 00000000008ec000 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::FullGC::Mark()+248)(4025597d50430dc8ac5e5f0b44678df8) #09 pc 00000000008eb750 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::FullGC::RunPhases()+1660)(4025597d50430dc8ac5e5f0b44678df8) #10 pc 00000000001ef154 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::Heap::CollectGarbageImpl(panda::ecmascript::TriggerGCType, panda::ecmascript::GCReason)+1212)(4025597d50430dc8ac5e5f0b44678df8) #11 pc 00000000003854b0 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::Heap::CollectGarbage(panda::ecmascript::TriggerGCType, panda::ecmascript::GCReason)+136)(4025597d50430dc8ac5e5f0b44678df8) #12 pc 00000000003853b8 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::Heap::CheckAndTriggerOldGC(unsigned long)+312)(4025597d50430dc8ac5e5f0b44678df8) #13 pc 00000000001eae98 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::ObjectFactory::NewTaggedArray(unsigned int, panda::ecmascript::JSTaggedValue)+5012)(4025597d50430dc8ac5e5f0b44678df8) #14 pc 000000000026bc2c /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::JSObject::TryOptimizeAsFastElements(panda::ecmascript::JSThread const*, panda::ecmascript::JSHandle<panda::ecmascript::JSObject>)+168)(4025597d50430dc8ac5e5f0b44678df8) #15 pc 00000000004eeb9c /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::builtins::BuiltinsArray::Fill(panda::ecmascript::EcmaRuntimeCallInfo*)+1028)(4025597d50430dc8ac5e5f0b44678df8) #16 pc 0000000000e179f4 /system/lib64/module/arkcompiler/stub.an(RTStub_PushCallArgsAndDispatchNative+40) #17 pc 0000000000b25df8 /system/lib64/module/arkcompiler/stub.an(BuiltinStub_ArrayFillStwCopy+472) #18 pc 000000000046bce4 /system/lib64/module/arkcompiler/stub.an(BCStub_HandleCallthis1Imm8V8V8StwCopy+528) #19 at generateMemoryPressure (entry|entry|1.0.0|src/main/ets/pages/page_second/page_third_appfreeze/appfreeze_threadblock.ts:1865:56)
关键字FullGC表明主线程阻塞在GC上,#19栈帧表明GC前的业务是generateMemoryPressure。
预防建议
减少内存分配,优化数据结构,避免频繁GC。
根因描述
应用主线程OOM触发长时间抓取内存快照导致阻塞。
问题分析思路
关键字
关注主线程堆栈共享库和符号是否匹配如下关键字。
共享库:
libark_jsruntime.so
符号:
DumpHeapSnapshot
案例分析
案例:OOM抓取内存快照耗时导致主线程阻塞
问题现象
当用户触发业务代码调用后,应用进入无响应状态,一段时间后应用退出,并生成冻屏日志文件。
问题分析
证据:主线程堆栈关键栈帧
Tid:12892, Name:pfreezeanalysis state=R, utime=648, stime=86, priority=-14, nice=-10, clk=100 #00 pc 0000000000a45b30 /system/lib64/platformsdk/libark_jsruntime.so(4025597d50430dc8ac5e5f0b44678df8) #01 pc 0000000000a98880 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::EntryIdMap::FindId(unsigned long)+48)(4025597d50430dc8ac5e5f0b44678df8) #02 pc 0000000000aaa860 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::HeapSnapshot::GenerateNode(panda::ecmascript::JSTaggedValue, unsigned long, bool, bool, bool)+620)(4025597d50430dc8ac5e5f0b44678df8) #03 pc 0000000000aa2644 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::HeapSnapshot::FillEdges(bool)+852)(4025597d50430dc8ac5e5f0b44678df8) #04 pc 0000000000aa2138 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::HeapSnapshot::BuildUp(bool)+40)(4025597d50430dc8ac5e5f0b44678df8) #05 pc 0000000000a9aa30 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::HeapProfiler::MakeHeapSnapshot(panda::ecmascript::HeapProfiler::SampleType, panda::ecmascript::DumpSnapShotOption const&, bool)+360)(4025597d50430dc8ac5e5f0b44678df8) #06 pc 0000000000a9a1c8 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::HeapProfiler::DoDump(panda::ecmascript::Stream*, panda::ecmascript::Progress*, panda::ecmascript::DumpSnapShotOption const&)+100)(4025597d50430dc8ac5e5f0b44678df8) #07 pc 0000000000a9b630 /system/lib64/platformsdk/libark_jsruntime.so(panda::ecmascript::HeapProfiler::DumpHeapSnapshot(panda::ecmascript::Stream*, panda::ecmascript::DumpSnapShotOption const&, panda::ecmascript::Progress*, std::__h::function<void (unsigned char)>)+972)(4025597d50430dc8ac5e5f0b44678df8) #08 pc 0000000000a88120 /system/lib64/platformsdk/libark_jsruntime.so(panda::DFXJSNApi::DumpHeapSnapshot(panda::ecmascript::EcmaVM const*, panda::ecmascript::Stream*, panda::ecmascript::DumpSnapShotOption const&, panda::ecmascript::Progress*, std::__h::function<void (unsigned char)>)+136)(4025597d50430dc8ac5e5f0b44678df8) #09 pc 0000000000a87fec /system/lib64/platformsdk/libark_jsruntime.so(panda::DFXJSNApi::DumpHeapSnapshot(panda::ecmascript::EcmaVM const*, std::__h::basic_string<char, std::__h::char_traits<char>, std::__h::allocator<char>> const&, panda::ecmascript::DumpSnapShotOption const&, std::__h::function<void (unsigned char)> const&)+120)(4025597d50430dc8ac5e5f0b44678df8) #10 pc 000000000006d620 /system/lib64/platformsdk/libace_napi.z.so(ArkNativeEngine::DumpHeapSnapshot(std::__h::basic_string<char, std::__h::char_traits<char>, std::__h::allocator<char>> const&, bool, DumpFormat, bool, bool, bool, bool)+192)(76bc1ed675edcc5f429a976d6c9b955d) #11 pc 000000000000f4b4 /system/lib64/module/libhidebug.z.so(OHOS::HiviewDFX::DumpHeapSnapshot(std::__h::basic_string<char, std::__h::char_traits<char>, std::__h::allocator<char>> const&, napi_env__*, bool)+376)(91aaf2c7bd1d5fb2dfa77f48a3ffc31d) #12 pc 000000000000facc /system/lib64/module/libhidebug.z.so(OHOS::HiviewDFX::DumpJsHeapData(napi_env__*, napi_callback_info__*)+788)(91aaf2c7bd1d5fb2dfa77f48a3ffc31d) #13 pc 0000000000067c90 /system/lib64/platformsdk/libace_napi.z.so(panda::JSValueRef ArkNativeFunctionCallBack<true>(panda::JsiRuntimeCallInfo*)+288)(76bc1ed675edcc5f429a976d6c9b955d) #14 pc 0000000000e179f4 /system/lib64/module/arkcompiler/stub.an(RTStub_PushCallArgsAndDispatchNative+40) #15 pc 000000000046bc74 /system/lib64/module/arkcompiler/stub.an(BCStub_HandleCallthis1Imm8V8V8StwCopy+416)
堆栈中DumpHeapSnapshot关键字表明应用正在生成堆快照,快照生成耗时导致主线程阻塞。
预防建议
优化内存使用和管理,避免内存超限发生OOM。
根因描述
应用主线程执行耗时操作导致阻塞。
问题分析思路
关键字
关注主线程堆栈中的应用侧业务栈帧:
#13 at triggerBinderInterfaceBlockSync (entry|entry|1.0.0|src/main/ets/pages/page_second/page_third_appfreeze/appfreeze_threadblock.ts:147:43)
#01 pc 000000000007a638 /data/storage/el1/bundle/libs/arm64/libentry.so(TriggerLongTimeOp(napi_env__*, napi_callback_info__*)+44)(f9377ecb68078186ec50cb9740a6a4c259058746)
案例分析
案例:业务函数耗时导致主线程阻塞
问题现象
当用户触发业务代码调用后,应用进入无响应状态,一段时间后应用退出,并生成冻屏日志文件。
问题分析
证据1:主线程堆栈关键栈帧
Tid:23173, Name:pfreezeanalysis state=S, utime=61, stime=45, priority=-14, nice=-10, clk=100 #00 pc 00000000001ebb00 /system/lib/ld-musl-aarch64.so.1(sleep+64)(8c2ca4e308a1de9ed1e6b4ccc14c7443) #01 pc 000000000007a638 /data/storage/el1/bundle/libs/arm64/libentry.so(TriggerLongTimeOp(napi_env__*, napi_callback_info__*)+44)(f9377ecb68078186ec50cb9740a6a4c259058746) #02 pc 0000000000067c90 /system/lib64/platformsdk/libace_napi.z.so(panda::JSValueRef ArkNativeFunctionCallBack<true>(panda::JsiRuntimeCallInfo*)+288)(76bc1ed675edcc5f429a976d6c9b955d) #03 pc 0000000000e179f4 /system/lib64/module/arkcompiler/stub.an(RTStub_PushCallArgsAndDispatchNative+40) #04 pc 000000000046b900 /system/lib64/module/arkcompiler/stub.an(BCStub_HandleCallthis0Imm8V8StwCopy+392)
#01栈帧表明存在耗时操作的业务方是libentry.so(TriggerLongTimeOp),结合代码继续分析。
证据2:故障栈帧所在的代码

TriggerLongTimeOp方法中的while循环耗时导致主线程阻塞。
预防建议
应用侧可以优化的方向:
如果应用在后台,可能由于系统资源供给不足导致,可以忽略本次应用冻屏故障。