itgmania212121/stepmania/src/archutils/Unix/CrashHandler.cpp

#include "global.h"

#include <cstdio>
#include <cstdlib>
#include <cstdarg>
#include <unistd.h>
#include <cerrno>
#include <limits.h>
#include <fcntl.h>
#include <csignal>

#include "RageLog.h" /* for RageLog::GetAdditionalLog, etc, only */
#include "RageThreads.h"
#include "Backtrace.h"

#include <sys/types.h>
#include <sys/wait.h>

#include "CrashHandler.h"
#include "CrashHandlerInternal.h"

#if defined(DARWIN)
# include "../Darwin/DarwinThreadHelpers.h"
#endif

extern uint64_t GetInvalidThreadId();
extern const char *g_pCrashHandlerArgv0;

static void safe_print(int fd, ...)
{
	va_list ap;
	va_start(ap, fd);

	while(1)
	{
		const char *p = va_arg(ap, const char *);
		if(p == NULL)
			break;
		write(fd, p, strlen(p));
	}
	va_end(ap);
}

static const char *itoa(unsigned n)
{
	static char ret[32];
	char *p = ret;
	for( int div = 1000000000; div > 0; div /= 10 )
	{
		*p++ = (n / div) + '0';
		n %= div;
	}
	*p = 0;
	p = ret;
	while( p[0] == '0' && p[1] )
		++p;
	return p;
}

#if defined(LINUX)
static void GetExecutableName( char *buf, int bufsize )
{
	/* readlink(/proc/pid/exe).  This is more reliable than argv[0]. */
	char proc_fn[1024] = "/proc/";
	strcat( proc_fn, itoa( getpid() ) );
	strcat( proc_fn, "/exe" );

	int got = readlink( proc_fn, buf, bufsize-1 );
	if( got == -1 )
	{
		safe_print( fileno(stderr), "Crash handler readlink ", proc_fn, " failed: ", strerror(errno), "\n", NULL );

		strncpy( buf, g_pCrashHandlerArgv0, bufsize );
		buf[bufsize-1] = 0;
	}


	buf[got] = 0;
}
#else
static void GetExecutableName( char *buf, int bufsize )
{
	strncpy( buf, g_pCrashHandlerArgv0, bufsize );
	buf[bufsize-1] = 0;
}
#endif

static void NORETURN spawn_child_process( int from_parent )
{
	/* We need to re-exec ourself, to get a clean process.  Close all
	 * FDs except for 0-2 and to_child, and then assign to_child to 3. */
	for(int fd = 3; fd < 1024; ++fd)
		if(fd != from_parent) close(fd);
       
	if(from_parent != 3)
	{
		dup2(from_parent, 3);
		close(from_parent);
	}

	char path[1024];
	GetExecutableName( path, sizeof(path) );

	/* Use execve; it's the lowest-level of the exec calls.  The others may allocate. */
	char *argv[3] = { path, CHILD_MAGIC_PARAMETER, NULL };
	char *envp[1] = { NULL };
	execve( path, argv, envp );

	/* If we got here, the exec failed.  We can't call strerror. */
	// safe_print(fileno(stderr), "Crash handler execl(", path, ") failed: ", strerror(errno), "\n", NULL);
	safe_print(fileno(stderr), "Crash handler execl(", path, ") failed: ", itoa( errno ), "\n", NULL);
	_exit(1);
}

/* write(), but retry a couple times on EINTR. */
static int retried_write( int fd, const void *buf, size_t count )
{
	int tries = 3, ret;
	do
	{
		ret = write( fd, buf, count );
	}
	while( ret == -1 && errno == EINTR && tries-- );
	
	return ret;
}

static bool parent_write(int to_child, const void *p, size_t size)
{
	int ret = retried_write(to_child, p, size);
	if( ret == -1 )
	{
		safe_print(fileno(stderr), "Unexpected write() result (", strerror(errno), ")\n", NULL);
		return false;
	}

	if(size_t(ret) != size)
	{
		safe_print(fileno(stderr), "Unexpected write() result (", itoa(ret), ")\n", NULL);
		return false;
	}

	return true;
}

static void parent_process( int to_child, const CrashData *crash )
{
	/* 1. Write the CrashData. */
	if( !parent_write(to_child, crash, sizeof(CrashData)) )
		return;
	
	/* 2. Write info. */
	const char *p = RageLog::GetInfo();
	int size = strlen(p)+1;
	if( !parent_write(to_child, &size, sizeof(size)) )
		return;
	if( !parent_write(to_child, p, size) )
		return;

	/* 3. Write AdditionalLog. */
	p = RageLog::GetAdditionalLog();
	size = strlen(p)+1;
	if( !parent_write(to_child, &size, sizeof(size)) )
		return;
	if( !parent_write(to_child, p, size) )
		return;
	
	/* 4. Write RecentLogs. */
	int cnt = 0;
	const char *ps[1024];
	while( cnt < 1024 && (ps[cnt] = RageLog::GetRecentLog( cnt )) != NULL )
		++cnt;

	parent_write(to_child, &cnt, sizeof(cnt));
	for( int i = 0; i < cnt; ++i )
	{
		size = strlen(ps[i])+1;
		if( !parent_write(to_child, &size, sizeof(size)) )
			return;
		if( !parent_write(to_child, ps[i], size) )
			return;
	}

    /* 5. Write CHECKPOINTs. */
    static char buf[1024*32];
    Checkpoints::GetLogs( buf, sizeof(buf), "\n" );
    size = strlen(buf)+1;
    if( !parent_write(to_child, &size, sizeof(size)) )
	return;
    if( !parent_write(to_child, buf, size) )
	return;
	
    /* 6. Write the crashed thread's name. */
    p = RageThread::GetCurThreadName();
    size = strlen(p)+1;
    if( !parent_write(to_child, &size, sizeof(size)) )
	return;
    if( !parent_write(to_child, p, size) )
	return;
}


/* The parent process is the crashed process.  It'll send data to the
 * child, who will do stuff with it.  The parent then waits for the
 * child to quit, and exits. 
 *
 * We can do whatever fancy things we want in the child process.  However,
 * let's not open any windows until we at least try to shut down OpenGL,
 * since it may cause problems.  We don't want to try to shut down OpenGL
 * until we've sent all of our data, since it might explode.
 *
 * So, first fork off the error reporting child, send data to it, shut down
 * OpenGL, close the socket and wait for the child to shut down.
 *
 * The child reads the data from the parent, waits for the socket to close
 * (EOF), and it's then free to open windows and stuff.
 *
 * XXX: make sure the parent dying doesn't take out the child
 */

/* The x86 backtrace() in glibc doesn't make any effort at all to decode
 * signal trampolines.  The result is that it doesn't properly show the
 * function that actually caused the signal--which is the most important
 * one!  So, we have to do it all ourself. */
const char *SignalName( int signo )
{
#define X(a) case a: return #a;
	switch( signo )
	{
	case SIGALRM: return "Alarm";
	case SIGBUS: return "Bus error";
	case SIGFPE: return "Floating point exception";
	X(SIGHUP)
	case SIGILL: return "Illegal instruction";
	X(SIGINT)
	case SIGPIPE: return "Broken pipe";
	case SIGABRT: return "Aborted";
	X(SIGQUIT)
	case SIGSEGV: return "Segmentation fault";
	X(SIGTRAP) X(SIGTERM) X(SIGVTALRM) X(SIGXCPU) X(SIGXFSZ)
#if defined(HAVE_DECL_SIGPWR) && HAVE_DECL_SIGPWR
	X(SIGPWR)
#endif
	default:
	{
		static char buf[128];
		strcpy( buf, "Unknown signal " );
		strcat( buf, itoa(signo) );
		return buf;
	}
	}
#undef X
}

static void RunCrashHandler( const CrashData *crash )
{
	if( g_pCrashHandlerArgv0 == NULL )
	{
		safe_print(fileno(stderr), "Crash handler failed: CrashHandlerHandleArgs was not called\n", NULL);
		_exit(1);
	}
	
	/* Block SIGPIPE, so we get EPIPE. */
	struct sigaction sa;
	memset( &sa, 0, sizeof(sa) );
	sa.sa_handler = SIG_IGN;
	if( sigaction( SIGPIPE, &sa, NULL ) != 0 )
	{
		safe_print(fileno(stderr), "sigaction() failed: %s", strerror(errno), NULL);
		/* non-fatal */
	}
	       
	static int received = 0;
	static int active = 0;

	if( active )
	{
		/* We've received a second signal.  This may mean that another thread
		 * crashed before we stopped it, or it may mean that the crash handler
		 * crashed. */
		switch( crash->type )
		{
		case CrashData::SIGNAL:
			safe_print( fileno(stderr), "Fatal signal (", SignalName(crash->signal), ")", NULL );
			break;

		case CrashData::FORCE_CRASH:
			safe_print( fileno(stderr), "Crash handler failed: \"", crash->reason, "\"", NULL );
			break;

		default:
			safe_print( fileno(stderr), "Unexpected RunCrashHandler call (", itoa(crash->type), ")", NULL );
			break;
		}

		if( active == 1 )
			safe_print( fileno(stderr), " while still in the crash handler\n", NULL);
		else if( active == 2 )
			safe_print( fileno(stderr), " while in the crash handler child\n", NULL);

		_exit(1);
	}
	active = 1;
	received = getpid();

	/* Stop other threads.  XXX: This prints a spurious ptrace error if any threads
	 * are already suspended, which happens in ForceCrashHandlerDeadlock(). */
	RageThread::HaltAllThreads();
	
	/* We need to be very careful, since we're under crash conditions.  Let's fork
	 * a process and exec ourself to get a clean environment to work in. */
	int fds[2];
	if(pipe(fds) != 0)
	{
		safe_print(fileno(stderr), "Crash handler pipe() failed: ", strerror(errno), "\n", NULL);
		exit(1);
	}

	pid_t childpid = fork();
	if( childpid == -1 )
	{
		safe_print(fileno(stderr), "Crash handler fork() failed: ", strerror(errno), "\n", NULL);
		_exit(1);
	}

	if( childpid == 0 )
	{
		active = 2;
		close(fds[1]);
		spawn_child_process(fds[0]);
	}
	else
	{
		close(fds[0]);
		parent_process( fds[1], crash );
		close( fds[1] );

		int status = 0;
		waitpid( childpid, &status, 0 );

		/* We need to resume threads before continuing, or we may deadlock on _exit(). */
		/* XXX: race condition:  If two threads are deadlocked on a pair of mutexes, there's
		 * a chance that they'll both try to lock the other's mutex at the same time.  If
		 * that happens, then they'll both time out the lock at about the same time.  One
		 * will trigger the deadlock crash first, and the other will be suspended.  However,
		 * once we resume it here, it'll continue, and * trigger the deadlock crash again.
		 * It can result in unrecoverable deadlocks. */
		RageThread::ResumeAllThreads();

		if( WIFSIGNALED(status) )
			safe_print( fileno(stderr), "Crash handler child exited with signal ", itoa(WTERMSIG(status)), "\n", NULL);
	}
}

void ForceCrashHandler( const char *reason )
{
	CrashData crash;
	memset( &crash, 0, sizeof(crash) );

	crash.type = CrashData::FORCE_CRASH;
	strncpy( crash.reason, reason, sizeof(crash.reason) );
	crash.reason[ sizeof(crash.reason)-1 ] = 0;

	GetBacktrace( crash.BacktracePointers[0], BACKTRACE_MAX_SIZE, NULL );

	RunCrashHandler( &crash );
}

void ForceCrashHandlerDeadlock( CString reason, uint64_t iID )
{
	CrashData crash;
	memset( &crash, 0, sizeof(crash) );

	crash.type = CrashData::FORCE_CRASH;

	GetBacktrace( crash.BacktracePointers[0], BACKTRACE_MAX_SIZE, NULL );

	if( iID == GetInvalidThreadId() )
	{
		/* Backtrace all threads. */
		int iCnt = 1;
		for( int i = 0; RageThread::EnumThreadIDs(i, iID); ++i )
		{
			if( iID == GetInvalidThreadId() || iID == RageThread::GetCurrentThreadID() )
				continue;

			BacktraceContext ctx;
			if( GetThreadBacktraceContext( iID, &ctx ) )
				GetBacktrace( crash.BacktracePointers[iCnt], BACKTRACE_MAX_SIZE, &ctx );
			strncpy( crash.m_ThreadName[iCnt], RageThread::GetThreadNameByID(iID), sizeof(crash.m_ThreadName[0])-1 );

			++iCnt;

			if( iCnt == CrashData::MAX_BACKTRACE_THREADS )
				break;
		}
	}
	else
	{
		BacktraceContext ctx;
		if( !GetThreadBacktraceContext( iID, &ctx ) )
			reason += "; GetThreadBacktraceContext failed";
		else
			GetBacktrace( crash.BacktracePointers[1], BACKTRACE_MAX_SIZE, &ctx );
		strncpy( crash.m_ThreadName[1], RageThread::GetThreadNameByID(iID), sizeof(crash.m_ThreadName[0])-1 );
	}

	strncpy( crash.m_ThreadName[0], RageThread::GetCurThreadName(), sizeof(crash.m_ThreadName[0])-1 );

	strncpy( crash.reason, reason, min(sizeof(crash.reason) - 1, reason.length()) );
	crash.reason[ sizeof(crash.reason)-1 ] = 0;

	RunCrashHandler( &crash );

	_exit(1);
}

/* XXX test for recursive crashes here (eg. GetBacktrace crashing) */

void CrashSignalHandler( int signal, siginfo_t *si, const ucontext_t *uc )
{
	CrashData crash;
	memset( &crash, 0, sizeof(crash) );

	crash.type = CrashData::SIGNAL;
	crash.signal = signal;
	crash.si = *si;

	BacktraceContext ctx;
	GetSignalBacktraceContext( &ctx, uc );
	GetBacktrace( crash.BacktracePointers[0], BACKTRACE_MAX_SIZE, &ctx );

	strncpy( crash.m_ThreadName[0], RageThread::GetCurThreadName(), sizeof(crash.m_ThreadName[0])-1 );

	RunCrashHandler( &crash );
}

void InitializeCrashHandler()
{
	InitializeBacktrace();
}

/*
 * (c) 2003-2004 Glenn Maynard
 * All rights reserved.
 * 
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the
 * "Software"), to deal in the Software without restriction, including
 * without limitation the rights to use, copy, modify, merge, publish,
 * distribute, and/or sell copies of the Software, and to permit persons to
 * whom the Software is furnished to do so, provided that the above
 * copyright notice(s) and this permission notice appear in all copies of
 * the Software and that both the above copyright notice(s) and this
 * permission notice appear in supporting documentation.
 * 
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF
 * THIRD PARTY RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS
 * INCLUDED IN THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT
 * OR CONSEQUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS
 * OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR
 * OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
 * PERFORMANCE OF THIS SOFTWARE.
 */