syscall.c

linux进程跟踪的工具和源代码

/*
 * Copyright (c) 1991, 1992 Paul Kranenburg <[e-mail]>
 * Copyright (c) 1993 Branko Lankester <[e-mail]>
 * Copyright (c) 1993, 1994, 1995, 1996 Rick Sladkey <[e-mail]>
 * Copyright (c) 1996-1999 Wichert Akkerman <[e-mail]>
 * Copyright (c) 1999 IBM Deutschland Entwicklung GmbH, IBM Corporation
 *                     Linux for s390 port by D.J. Barrow
 *                    <[e-mail],[e-mail]>
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. The name of the author may not be used to endorse or promote products
 *    derived from this software without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 *	$Id: syscall.c,v 1.30 2001/03/28 20:29:17 wichert Exp $
 */

#include "defs.h"

#include <signal.h>
#include <time.h>
#include <errno.h>
#include <sys/user.h>
#include <sys/syscall.h>
#include <sys/param.h>

#if HAVE_ASM_REG_H
#ifdef SPARC
#  define fpq kernel_fpq
#  define fq kernel_fq
#  define fpu kernel_fpu
#endif
#include <asm/reg.h>
#ifdef SPARC
#  undef fpq
#  undef fq
#  undef fpu 
#endif
#endif

#ifdef HAVE_SYS_REG_H
#include <sys/reg.h>
#ifndef PTRACE_PEEKUSR
# define PTRACE_PEEKUSR PTRACE_PEEKUSER
#endif
#elif defined(HAVE_LINUX_PTRACE_H)
#undef PTRACE_SYSCALL
#include <linux/ptrace.h>
#endif

#if defined(LINUX) && defined(IA64)
# include <asm/ptrace_offsets.h>
# include <asm/rse.h>
#endif

#ifndef SYS_ERRLIST_DECLARED
extern int sys_nerr;
extern char *sys_errlist[];
#endif /* SYS_ERRLIST_DECLARED */

#define NR_SYSCALL_BASE 0
#ifdef LINUX
#ifndef ERESTARTSYS
#define ERESTARTSYS	512
#endif
#ifndef ERESTARTNOINTR
#define ERESTARTNOINTR	513
#endif
#ifndef ERESTARTNOHAND
#define ERESTARTNOHAND	514	/* restart if no handler.. */
#endif
#ifndef ENOIOCTLCMD
#define ENOIOCTLCMD	515	/* No ioctl command */
#endif
#ifndef NSIG
#define NSIG 32
#endif
#ifdef ARM
#undef NSIG
#define NSIG 32
#undef NR_SYSCALL_BASE
#define NR_SYSCALL_BASE __NR_SYSCALL_BASE
#endif
#endif /* LINUX */

#include "syscall.h"

/* Define these shorthand notations to simplify the syscallent files. */
#define TF TRACE_FILE
#define TI TRACE_IPC
#define TN TRACE_NETWORK
#define TP TRACE_PROCESS
#define TS TRACE_SIGNAL

struct sysent sysent0[] = {
#include "syscallent.h"
};
int nsyscalls0 = sizeof sysent0 / sizeof sysent0[0];

#if SUPPORTED_PERSONALITIES >= 2
struct sysent sysent1[] = {
#include "syscallent1.h"
};
int nsyscalls1 = sizeof sysent1 / sizeof sysent1[0];
#endif /* SUPPORTED_PERSONALITIES >= 2 */

#if SUPPORTED_PERSONALITIES >= 3
struct sysent sysent2[] = {
#include "syscallent2.h"
};
int nsyscalls2 = sizeof sysent2 / sizeof sysent2[0];
#endif /* SUPPORTED_PERSONALITIES >= 3 */

struct sysent *sysent;
int nsyscalls;

/* Now undef them since short defines cause wicked namespace pollution. */
#undef TF
#undef TI
#undef TN
#undef TP
#undef TS

char *errnoent0[] = {
#include "errnoent.h"
};
int nerrnos0 = sizeof errnoent0 / sizeof errnoent0[0];

#if SUPPORTED_PERSONALITIES >= 2
char *errnoent1[] = {
#include "errnoent1.h"
};
int nerrnos1 = sizeof errnoent1 / sizeof errnoent1[0];
#endif /* SUPPORTED_PERSONALITIES >= 2 */

#if SUPPORTED_PERSONALITIES >= 3
char *errnoent2[] = {
#include "errnoent2.h"
};
int nerrnos2 = sizeof errnoent2 / sizeof errnoent2[0];
#endif /* SUPPORTED_PERSONALITIES >= 3 */

char **errnoent;
int nerrnos;

int current_personality;

int
set_personality(personality)
int personality;
{
	switch (personality) {
	case 0:
		errnoent = errnoent0;
		nerrnos = nerrnos0;
		sysent = sysent0;
		nsyscalls = nsyscalls0;
		ioctlent = ioctlent0;
		nioctlents = nioctlents0;
		signalent = signalent0;
		nsignals = nsignals0;
		break;

#if SUPPORTED_PERSONALITIES >= 2
	case 1:
		errnoent = errnoent1;
		nerrnos = nerrnos1;
		sysent = sysent1;
		nsyscalls = nsyscalls1;
		ioctlent = ioctlent1;
		nioctlents = nioctlents1;
		signalent = signalent1;
		nsignals = nsignals1;
		break;
#endif /* SUPPORTED_PERSONALITIES >= 2 */

#if SUPPORTED_PERSONALITIES >= 3
	case 2:
		errnoent = errnoent2;
		nerrnos = nerrnos2;
		sysent = sysent2;
		nsyscalls = nsyscalls2;
		ioctlent = ioctlent2;
		nioctlents = nioctlents2;
		signalent = signalent2;
		nsignals = nsignals2;
		break;
#endif /* SUPPORTED_PERSONALITIES >= 3 */

	default:
		return -1;
	}

	current_personality = personality;
	return 0;
}

int qual_flags[MAX_QUALS];

static int call_count[MAX_QUALS];
static int error_count[MAX_QUALS];
static struct timeval tv_count[MAX_QUALS];
static int sorted_count[MAX_QUALS];

static struct timeval shortest = { 1000000, 0 };

static int lookup_syscall(), lookup_signal(), lookup_fault(), lookup_desc();

static struct qual_options {
	int bitflag;
	char *option_name;
	int (*lookup)();
	char *argument_name;
} qual_options[] = {
	{ QUAL_TRACE,	"trace",	lookup_syscall,	"system call"	},
	{ QUAL_TRACE,	"t",		lookup_syscall,	"system call"	},
	{ QUAL_ABBREV,	"abbrev",	lookup_syscall,	"system call"	},
	{ QUAL_ABBREV,	"a",		lookup_syscall,	"system call"	},
	{ QUAL_VERBOSE,	"verbose",	lookup_syscall,	"system call"	},
	{ QUAL_VERBOSE,	"v",		lookup_syscall,	"system call"	},
	{ QUAL_RAW,	"raw",		lookup_syscall,	"system call"	},
	{ QUAL_RAW,	"x",		lookup_syscall,	"system call"	},
	{ QUAL_SIGNAL,	"signal",	lookup_signal,	"signal"	},
	{ QUAL_SIGNAL,	"signals",	lookup_signal,	"signal"	},
	{ QUAL_SIGNAL,	"s",		lookup_signal,	"signal"	},
	{ QUAL_FAULT,	"fault",	lookup_fault,	"fault"		},
	{ QUAL_FAULT,	"faults",	lookup_fault,	"fault"		},
	{ QUAL_FAULT,	"m",		lookup_fault,	"fault"		},
	{ QUAL_READ,	"read",		lookup_desc,	"descriptor"	},
	{ QUAL_READ,	"reads",	lookup_desc,	"descriptor"	},
	{ QUAL_READ,	"r",		lookup_desc,	"descriptor"	},
	{ QUAL_WRITE,	"write",	lookup_desc,	"descriptor"	},
	{ QUAL_WRITE,	"writes",	lookup_desc,	"descriptor"	},
	{ QUAL_WRITE,	"w",		lookup_desc,	"descriptor"	},
	{ 0,		NULL,		NULL,		NULL		},
};

static int
lookup_syscall(s)
char *s;
{
	int i;

	for (i = 0; i < nsyscalls; i++) {
		if (strcmp(s, sysent[i].sys_name) == 0)
			return i;
	}
	return -1;
}

static int
lookup_signal(s)
char *s;
{
	int i;
	char buf[32];

	if (s && *s && isdigit((unsigned char)*s))
		return atoi(s);
	strcpy(buf, s);
	s = buf;
	for (i = 0; s[i]; i++)
		s[i] = toupper((unsigned char)(s[i]));
	if (strncmp(s, "SIG", 3) == 0)
		s += 3;
	for (i = 0; i <= NSIG; i++) {
		if (strcmp(s, signame(i) + 3) == 0)
			return i;
	}
	return -1;
}

static int
lookup_fault(s)
char *s;
{
	return -1;
}

static int
lookup_desc(s)
char *s;
{
	if (s && *s && isdigit((unsigned char)*s))
		return atoi(s);
	return -1;
}

static int
lookup_class(s)
char *s;
{
	if (strcmp(s, "file") == 0)
		return TRACE_FILE;
	if (strcmp(s, "ipc") == 0)
		return TRACE_IPC;
	if (strcmp(s, "network") == 0)
		return TRACE_NETWORK;
	if (strcmp(s, "process") == 0)
		return TRACE_PROCESS;
	if (strcmp(s, "signal") == 0)
		return TRACE_SIGNAL;
	return -1;
}

void
qualify(s)
char *s;
{
	struct qual_options *opt;
	int not;
	char *p;
	int i, n;

	opt = &qual_options[0];
	for (i = 0; (p = qual_options[i].option_name); i++) {
		n = strlen(p);
		if (strncmp(s, p, n) == 0 && s[n] == '=') {
			opt = &qual_options[i];
			s += n + 1;
			break;
		}
	}
	not = 0;
	if (*s == '!') {
		not = 1;
		s++;
	}
	if (strcmp(s, "none") == 0) {
		not = 1 - not;
		s = "all";
	}
	if (strcmp(s, "all") == 0) {
		for (i = 0; i < MAX_QUALS; i++) {
			if (not)
				qual_flags[i] &= ~opt->bitflag;
			else
				qual_flags[i] |= opt->bitflag;
		}
		return;
	}
	for (i = 0; i < MAX_QUALS; i++) {
		if (not)
			qual_flags[i] |= opt->bitflag;
		else
			qual_flags[i] &= ~opt->bitflag;
	}
	for (p = strtok(s, ","); p; p = strtok(NULL, ",")) {
		if (opt->bitflag == QUAL_TRACE && (n = lookup_class(p)) > 0) {
			for (i = 0; i < MAX_QUALS; i++) {
				if (sysent[i].sys_flags & n) {
					if (not)
						qual_flags[i] &= ~opt->bitflag;
					else
						qual_flags[i] |= opt->bitflag;
				}
			}
			continue;
		}
		if ((n = (*opt->lookup)(p)) < 0) {
			fprintf(stderr, "strace: invalid %s `%s'\n",
				opt->argument_name, p);
			exit(1);
		}
		if (not)
			qual_flags[n] &= ~opt->bitflag;
		else
			qual_flags[n] |= opt->bitflag;
	}
	return;
}

static void
dumpio(tcp)
struct tcb *tcp;
{
	if (syserror(tcp))
		return;
	if (tcp->u_arg[0] < 0 || tcp->u_arg[0] >= MAX_QUALS)
		return;
	switch (tcp->scno + NR_SYSCALL_BASE) {
	case SYS_read:
#ifdef SYS_recv
	case SYS_recv:
#endif
#ifdef SYS_recvfrom
	case SYS_recvfrom:
#endif
		if (qual_flags[tcp->u_arg[0]] & QUAL_READ)
			dumpstr(tcp, tcp->u_arg[1], tcp->u_rval);
		break;
	case SYS_write:
#ifdef SYS_send
	case SYS_send:
#endif
#ifdef SYS_sendto
	case SYS_sendto:
#endif
		if (qual_flags[tcp->u_arg[0]] & QUAL_WRITE)
			dumpstr(tcp, tcp->u_arg[1], tcp->u_arg[2]);
		break;
	}
}

#ifndef FREEBSD
enum subcall_style { shift_style, deref_style, mask_style, door_style };
#else /* FREEBSD */
enum subcall_style { shift_style, deref_style, mask_style, door_style, table_style };

struct subcall {
  int call;
  int nsubcalls;
  int subcalls[5];
};

const struct subcall subcalls_table[] = {
  { SYS_shmsys, 5, { SYS_shmat, SYS_shmctl, SYS_shmdt, SYS_shmget, SYS_shmctl } },
#ifdef SYS_semconfig
  { SYS_semsys, 4, { SYS___semctl, SYS_semget, SYS_semop, SYS_semconfig } },
#else
  { SYS_semsys, 3, { SYS___semctl, SYS_semget, SYS_semop } },
#endif
  { SYS_msgsys, 4, { SYS_msgctl, SYS_msgget, SYS_msgsnd, SYS_msgrcv } },
};
#endif /* FREEBSD */

#if !(defined(LINUX) && ( defined(ALPHA) || defined(IA64) || defined(MIPS) ))

const int socket_map [] = {
	       /* SYS_SOCKET      */ 97,
	       /* SYS_BIND        */ 104,
	       /* SYS_CONNECT     */ 98,
	       /* SYS_LISTEN      */ 106,
	       /* SYS_ACCEPT      */ 99,
	       /* SYS_GETSOCKNAME */ 150,
	       /* SYS_GETPEERNAME */ 141,
	       /* SYS_SOCKETPAIR  */ 135,
	       /* SYS_SEND        */ 101,
	       /* SYS_RECV        */ 102,
	       /* SYS_SENDTO      */ 133,
	       /* SYS_RECVFROM    */ 125,
	       /* SYS_SHUTDOWN    */ 134,
	       /* SYS_SETSOCKOPT  */ 105,
	       /* SYS_GETSOCKOPT  */ 118,
	       /* SYS_SENDMSG     */ 114,
	       /* SYS_RECVMSG     */ 113
};

void
sparc_socket_decode (tcp)
struct tcb *tcp;
{
	volatile long addr;
	volatile int i, n;

	if (tcp->u_arg [0] < 1 || tcp->u_arg [0] > sizeof(socket_map)/sizeof(int)+1){
		return;
	}
	tcp->scno = socket_map [tcp->u_arg [0]-1];
	n = tcp->u_nargs = sysent [tcp->scno].nargs;
	addr = tcp->u_arg [1];
	for (i = 0; i < n; i++){
	        int arg;
		if (umoven (tcp, addr, sizeof (arg), (void *) &arg) < 0)
			arg = 0;
		tcp->u_arg [i] = arg;
		addr += sizeof (arg);
	}
}

static void
decode_subcall(tcp, subcall, nsubcalls, style)
struct tcb *tcp;
int subcall;
int nsubcalls;
enum subcall_style style;
{
	int i, addr, mask, arg;

	switch (style) {
	case shift_style:
		if (tcp->u_arg[0] < 0 || tcp->u_arg[0] >= nsubcalls)
			return;
		tcp->scno = subcall + tcp->u_arg[0];
		if (sysent[tcp->scno].nargs != -1)
			tcp->u_nargs = sysent[tcp->scno].nargs;
		else
			tcp->u_nargs--;
		for (i = 0; i < tcp->u_nargs; i++)
			tcp->u_arg[i] = tcp->u_arg[i + 1];
		break;
	case deref_style:
		if (tcp->u_arg[0] < 0 || tcp->u_arg[0] >= nsubcalls)
			return;
		tcp->scno = subcall + tcp->u_arg[0];
		addr = tcp->u_arg[1];
		for (i = 0; i < sysent[tcp->scno].nargs; i++) {
			if (umove(tcp, addr, &arg) < 0)
				arg = 0;
			tcp->u_arg[i] = arg;
			addr += sizeof(arg);
		}
		tcp->u_nargs = sysent[tcp->scno].nargs;
		break;
	case mask_style:
		mask = (tcp->u_arg[0] >> 8) & 0xff;
		for (i = 0; mask; i++)
			mask >>= 1;
		if (i >= nsubcalls)
			return;
		tcp->u_arg[0] &= 0xff;
		tcp->scno = subcall + i;
		if (sysent[tcp->scno].nargs != -1)
			tcp->u_nargs = sysent[tcp->scno].nargs;
		break;
	case door_style:
		/*
		 * Oh, yuck.  The call code is the *sixth* argument.
		 * (don't you mean the *last* argument? - JH)
		 */
		if (tcp->u_arg[5] < 0 || tcp->u_arg[5] >= nsubcalls)
			return;
		tcp->scno = subcall + tcp->u_arg[5];
		if (sysent[tcp->scno].nargs != -1)
			tcp->u_nargs = sysent[tcp->scno].nargs;
		else
			tcp->u_nargs--;
		break;
#ifdef FREEBSD
	case table_style:
		for (i = 0; i < sizeof(subcalls_table) / sizeof(struct subcall); i++)
			if (subcalls_table[i].call == tcp->scno) break;
		if (i < sizeof(subcalls_table) / sizeof(struct subcall) &&
		    tcp->u_arg[0] >= 0 && tcp->u_arg[0] < subcalls_table[i].nsubcalls) {
			tcp->scno = subcalls_table[i].subcalls[tcp->u_arg[0]];
			for (i = 0; i < tcp->u_nargs; i++)
				tcp->u_arg[i] = tcp->u_arg[i + 1];
		}
		break;
#endif /* FREEBSD */
	}
}
#endif

struct tcb *tcp_last = NULL;

static int
internal_syscall(tcp)
struct tcb *tcp;
{
	/*
	 * We must always trace a few critical system calls in order to
	 * correctly support following forks in the presence of tracing
	 * qualifiers.
	 */
	switch (tcp->scno + NR_SYSCALL_BASE) {
#ifdef SYS_fork
	case SYS_fork:
#endif
#ifdef SYS_vfork
	case SYS_vfork:
#endif
		internal_fork(tcp);
		break;
#ifdef SYS_clone
	case SYS_clone:
		internal_clone(tcp);
		break;
#endif
#ifdef SYS_execv
	case SYS_execv:
#endif
#ifdef SYS_execve
	case SYS_execve:
#endif
		internal_exec(tcp);
		break;

#ifdef SYS_wait
	case SYS_wait:
#endif
#ifdef SYS_wait4
	case SYS_wait4:
#endif
#ifdef SYS_waitpid
	case SYS_waitpid:
#endif
#ifdef SYS_waitsys
	case SYS_waitsys:
#endif
		internal_wait(tcp);
		break;

#ifdef SYS_exit
	case SYS_exit:
#endif
		internal_exit(tcp);
		break;
	}
	return 0;
}


#ifdef LINUX
#if defined (I386)
	static long eax;
#elif defined (IA64)
	long r8, r10, psr;
	long ia32 = 0;
#elif defined (POWERPC)
	static long result,flags;