Generate a dump when each of the specified GC memory thresholds are e…

…xceeded (#194)

* Refactor procdump/profiler

* initial checkin

* Feature complete

* Update README with new switch and example

* Fix memory leak

* Fixes

* PR feedback

* PR feedback

.devcontainer/devcontainer.json

@@ -5,5 +5,14 @@
 	"build": {
 		// If you want to use a different docker file (such as Dockerfile_Rocky) you can specify it here
 		"dockerfile": "Dockerfile_Ubuntu"
+	},
+	"customizations": {
+		"vscode": {
+			"extensions": [
+				"ms-vscode.cpptools",
+				"ms-vscode.cpptools-extension-pack",
+				"ms-vscode.makefile-tools"
+			]
+		}
 	}
 }

Makefile

@@ -54,7 +54,7 @@ install:
 	cp procdump.1 $(DESTDIR)$(MANDIR)
 
 $(OBJDIR)/ProcDumpProfiler.so: $(PROFSRCDIR)/ClassFactory.cpp $(PROFSRCDIR)/ProcDumpProfiler.cpp $(PROFSRCDIR)/dllmain.cpp $(PROFSRCDIR)/corprof_i.cpp $(PROFSRCDIR)/easylogging++.cc | $(OBJDIR)
-	$(PROFCLANG) -o $@ $(PROFCXXFLAGS) -I $(PROFINCDIR) $^
+	$(PROFCLANG) -o $@ $(PROFCXXFLAGS) -I $(PROFINCDIR) -I $(INCDIR) $^
 	ld -r -b binary -o $(OBJDIR)/ProcDumpProfiler.o $(OBJDIR)/ProcDumpProfiler.so
 
 $(OBJDIR)/%.o: $(SRCDIR)/%.c | $(OBJDIR)

README.md

@@ -25,6 +25,7 @@ procdump [-n Count]
         [-s Seconds]
         [-c|-cl CPU_Usage]
         [-m|-ml Commit_Usage1[,Commit_Usage2,...]]
+        [-gcm Memory_Usage1[,Memory_Usage2...]]
         [-tc Thread_Threshold]
         [-fc FileDescriptor_Threshold]
         [-sig Signal_Number]
@@ -42,8 +43,9 @@ Options:
    -s      Consecutive seconds before dump is written (default is 10).
    -c      CPU threshold above which to create a dump of the process.
    -cl     CPU threshold below which to create a dump of the process.
-   -m      Memory commit thresholds (MB) above which to create dumps.
-   -ml     Memory commit thresholds (MB) below which to create dumps.
+   -m      Memory commit threshold(s) (MB) above which to create dumps.
+   -ml     Memory commit threshold(s) (MB) below which to create dumps.
+   -gcm    [.NET] GC memory threshold(s) (MB) above which to create dumps.
    -tc     Thread count threshold above which to create a dump of the process.
    -fc     File descriptor count threshold above which to create a dump of the process.
    -sig    Signal number to intercept to create a dump of the process.
@@ -90,6 +92,10 @@ The following will create a core dump when memory usage is >= 100 MB followed by
 ```
 sudo procdump -m 100,200 1234
 ```
+The following will create a core dump when .NET memory usage is >= 100 MB followed by another dump when memory usage is >= 200MB.
+```
+sudo procdump -gcm 100,200 1234
+```
 The following will create a core dump in the `/tmp` directory immediately.
 ```
 sudo procdump 1234 /tmp

include/Includes.h

@@ -1,3 +1,4 @@
+#include "ProfilerCommon.h"
 #include "CoreDumpWriter.h"
 #include "Events.h"
 #include "GenHelpers.h"

include/Monitor.h

@@ -30,6 +30,9 @@ bool IsQuit(struct ProcDumpConfiguration *self);
 int SetQuit(struct ProcDumpConfiguration *self, int quit);
 bool ContinueMonitoring(struct ProcDumpConfiguration *self);
 bool BeginMonitoring(struct ProcDumpConfiguration *self);
+bool MonitorDotNet(struct ProcDumpConfiguration *self);
+char* GetThresholds(struct ProcDumpConfiguration *self);
+char* GetClientData(struct ProcDumpConfiguration *self, char* fullDumpPath);
 
 // Monitor worker threads
 void *CommitMonitoringThread(void *thread_args /* struct ProcDumpConfiguration* */);
@@ -38,7 +41,7 @@ void *ThreadCountMonitoringThread(void *thread_args /* struct ProcDumpConfigurat
 void *FileDescriptorCountMonitoringThread(void *thread_args /* struct ProcDumpConfiguration* */);
 void *SignalMonitoringThread(void *thread_args /* struct ProcDumpConfiguration* */);
 void *TimerThread(void *thread_args /* struct ProcDumpConfiguration* */);
-void *ExceptionMonitoringThread(void *thread_args /* struct ProcDumpConfiguration* */);
+void *DotNetMonitoringThread(void *thread_args /* struct ProcDumpConfiguration* */);
 void *ProcessMonitor(void *thread_args /* struct ProcDumpConfiguration* */);
 void *WaitForProfilerCompletion(void *thread_args /* struct ProcDumpConfiguration* */);
 

include/ProcDumpConfiguration.h

@@ -43,17 +43,6 @@
 // Structs
 // -------------------
 
-enum TriggerType
-{
-    Processor,
-    Commit,
-    Timer,
-    Signal,
-    ThreadCount,
-    FileDescriptorCount,
-    Exception
-};
-
 struct TriggerThread
 {
     pthread_t thread;
@@ -96,8 +85,10 @@ struct ProcDumpConfiguration {
     int CpuThreshold;               // -c
     bool bCpuTriggerBelowValue;     // -cl
     int* MemoryThreshold;           // -m
+    int MemoryThresholdCount;
     int MemoryCurrentThreshold;
     bool bMemoryTriggerBelowValue;  // -m or -ml
+    bool bMonitoringGCMemory;       // -gcm
     int ThresholdSeconds;           // -s
     bool bTimerThreshold;           // -s
     int NumberOfDumpsToCollect;     // -n

include/ProfilerCommon.h

@@ -0,0 +1,24 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License
+
+//--------------------------------------------------------------------
+//
+// Profiler Common
+//
+//--------------------------------------------------------------------
+#ifndef PROFILERCOMMON_H
+#define PROFILERCOMMON_H
+
+enum TriggerType
+{
+    Processor,
+    Commit,
+    Timer,
+    Signal,
+    ThreadCount,
+    FileDescriptorCount,
+    Exception,
+    GCThreshold
+};
+
+#endif // PROFILERCOMMON_H

include/ProfilerHelpers.h

@@ -14,8 +14,8 @@
 #define PROFILER_FILE_NAME  "procdumpprofiler.so"
 #define PROFILER_GUID       "{cf0d821e-299b-5307-a3d8-b283c03916dd}"
 
-int InjectProfiler(pid_t pid, char* filter, char* fullDumpPath);
-int LoadProfiler(pid_t pid, char* filter, char* fullDumpPath);
+int InjectProfiler(pid_t pid, char* clientData);
+int LoadProfiler(pid_t pid, char* clientData);
 int ExtractProfiler();
 char* GetEncodedExceptionFilter(char* exceptionFilterCmdLine, unsigned int numDumps);
 

profiler/inc/ProcDumpProfiler.h

@@ -19,6 +19,7 @@
 #include "cor.h"
 #include "corprof.h"
 #include "profilerstring.h"
+#include "ProfilerCommon.h"
 #include "easylogging++.h"
 
 #define DETACH_TIMEOUT  30000
@@ -103,16 +104,20 @@ class CorProfiler : public ICorProfilerCallback8
     ICorProfilerInfo* corProfilerInfo;
     ICorProfilerInfo8* corProfilerInfo8;
     std::vector<struct ExceptionMonitorEntry> exceptionMonitorList;
+    std::vector<uint64_t> gcMemoryThresholdMonitorList;
     pthread_t healthThread;
     std::string processName;
     std::string fullDumpPath;
     pthread_mutex_t endDumpCondition;
+    enum TriggerType triggerType;
+    int currentThresholdIndex;
+    bool gen2Collection;
 
     String GetExceptionName(ObjectID objectId);
     String GetExceptionMessage(ObjectID objectId);
-    bool ParseClientData(char* fw);
+    bool ParseClientData(char* clientData);
     WCHAR* GetUint16(char* buffer);
-    std::string GetDumpName(uint16_t dumpCount,std::string exceptionName);
+    std::string GetDumpName(uint16_t dumpCount, std::string name);
     std::string GetProcessName();
     bool GenerateCoreClrDump(char* socketName, char* dumpFileName);
     bool IsCoreClrProcess(pid_t pid, char** socketName);
@@ -122,6 +127,8 @@ class CorProfiler : public ICorProfilerCallback8
     int send_all(int socket, void* buffer, size_t length);
     int recv_all(int socket, void* buffer, size_t length);
     bool WildcardSearch(WCHAR*, WCHAR*);
+    u_int64_t GetGCHeapSize();
+    bool WriteDumpHelper(std::string dumpName);
 
 public:
     CorProfiler();