strace/defs.h
Denys Vlasenko 75fe85c2ee Fix the case where we try to detach unattached processes
Before this change:
$ strace -D -p1
strace: -D and -p are mutually exclusive options
Process 1 detached  <==== WRONG! (and we try to SIGSTOP it!!!)

* defs.h: Change the meaning of TCB_ATTACHED: now it means "this tracee
is attached to us". Add TCB_STRACE_CHILD: "this tracee is our child".
* strace.c (kill_save_errno): Move up. No code changes.
(process_opt_p_list): Don't set TCB_ATTACHED on new tcb.
(startup_attach): Change how we work with TCB_ATTACHED.
Set TCB_STRACE_CHILD on -D.
(startup_child): Use kill_save_errno instead of kill.
Set TCB_ATTACHED and TCB_STRACE_CHILD on attached strace child.
If we are in -D case, don't set TCB_ATTACHED (we aren't attached yet).
(detach): do not do PTRACE_DETACH if TCB_ATTACHED is not set.
(cleanup): Check TCB_STRACE_CHILD instead of TCB_ATTACHED.
(trace): Likewise.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
2012-03-09 15:15:24 +01:00

637 lines
20 KiB
C

/*
* Copyright (c) 1991, 1992 Paul Kranenburg <pk@cs.few.eur.nl>
* Copyright (c) 1993 Branko Lankester <branko@hacktic.nl>
* Copyright (c) 1993, 1994, 1995, 1996 Rick Sladkey <jrs@world.std.com>
* 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.
*
* $Id$
*/
#ifdef HAVE_CONFIG_H
# include "config.h"
#endif
#ifdef MIPS
# include <sgidefs.h>
#endif
#include <features.h>
#ifdef _LARGEFILE64_SOURCE
/* This is the macro everything checks before using foo64 names. */
# ifndef _LFS64_LARGEFILE
# define _LFS64_LARGEFILE 1
# endif
#endif
/* Configuration section */
#ifndef MAX_QUALS
# if defined(MIPS)
# define MAX_QUALS 7000 /* maximum number of syscalls, signals, etc. */
# else
# define MAX_QUALS 2048 /* maximum number of syscalls, signals, etc. */
# endif
#endif
#ifndef DEFAULT_STRLEN
/* default maximum # of bytes printed in `printstr', change with -s switch */
# define DEFAULT_STRLEN 32
#endif
#ifndef DEFAULT_ACOLUMN
# define DEFAULT_ACOLUMN 40 /* default alignment column for results */
#endif
/* Maximum number of args to a syscall.
*
* Make sure that all entries in all syscallent.h files have nargs <= MAX_ARGS!
* linux/<ARCH>/syscallent.h: all have nargs <= 6.
*/
#ifndef MAX_ARGS
# define MAX_ARGS 6
#endif
#ifndef DEFAULT_SORTBY
# define DEFAULT_SORTBY "time" /* default sorting method for call profiling */
#endif
#include <sys/types.h>
#include <unistd.h>
#include <stdlib.h>
#include <stdio.h>
#include <ctype.h>
#include <string.h>
#include <time.h>
#include <sys/time.h>
#include <errno.h>
#ifdef HAVE_STDBOOL_H
# include <stdbool.h>
#endif
#ifdef STDC_HEADERS
# include <stddef.h>
#endif /* STDC_HEADERS */
#ifdef HAVE_SIGINFO_T
# include <signal.h>
#endif
#if defined(SPARC) || defined(SPARC64)
# define LINUXSPARC
#endif
#if defined(MIPS) && _MIPS_SIM == _MIPS_SIM_ABI32
# define LINUX_MIPSO32
#endif
#if defined(MIPS) && _MIPS_SIM == _MIPS_SIM_NABI32
# define LINUX_MIPSN32
# define LINUX_MIPS64
#endif
#if defined(MIPS) && _MIPS_SIM == _MIPS_SIM_ABI64
# define LINUX_MIPSN64
# define LINUX_MIPS64
#endif
#if (defined(LINUXSPARC) || defined(X86_64) || defined(ARM) || defined(AVR32)) && defined(__GLIBC__)
# include <sys/ptrace.h>
#else
/* Work around awkward prototype in ptrace.h. */
# define ptrace xptrace
# include <sys/ptrace.h>
# undef ptrace
# ifdef POWERPC
# define __KERNEL__
# include <asm/ptrace.h>
# undef __KERNEL__
# endif
extern long ptrace(int, int, char *, long);
#endif
#if !defined(__GLIBC__)
# define PTRACE_PEEKUSER PTRACE_PEEKUSR
# define PTRACE_POKEUSER PTRACE_POKEUSR
#endif
#if defined(X86_64) || defined(I386)
/* For struct pt_regs. x86 strace uses PTRACE_GETREGS.
* PTRACE_GETREGS returns registers in the layout of this struct.
*/
# include <asm/ptrace.h>
#endif
#ifdef ALPHA
# define REG_R0 0
# define REG_A0 16
# define REG_A3 19
# define REG_FP 30
# define REG_PC 64
#endif /* ALPHA */
#ifdef MIPS
# define REG_V0 2
# define REG_A0 4
# define REG_A3 7
# define REG_SP 29
# define REG_EPC 64
#endif /* MIPS */
#ifdef HPPA
# define PT_GR20 (20*4)
# define PT_GR26 (26*4)
# define PT_GR28 (28*4)
# define PT_IAOQ0 (106*4)
# define PT_IAOQ1 (107*4)
#endif /* HPPA */
#ifdef SH64
/* SH64 Linux - this code assumes the following kernel API for system calls:
PC Offset 0
System Call Offset 16 (actually, (syscall no.) | (0x1n << 16),
where n = no. of parameters.
Other regs Offset 24+
On entry: R2-7 = parameters 1-6 (as many as necessary)
On return: R9 = result. */
/* Offset for peeks of registers */
# define REG_OFFSET (24)
# define REG_GENERAL(x) (8*(x)+REG_OFFSET)
# define REG_PC (0*8)
# define REG_SYSCALL (2*8)
#endif /* SH64 */
#define SUPPORTED_PERSONALITIES 1
#define DEFAULT_PERSONALITY 0
#ifdef LINUXSPARC
/* Indexes into the pt_regs.u_reg[] array -- UREG_XX from kernel are all off
* by 1 and use Ix instead of Ox. These work for both 32 and 64 bit Linux. */
# define U_REG_G1 0
# define U_REG_O0 7
# define U_REG_O1 8
# define PERSONALITY0_WORDSIZE 4
# define PERSONALITY1_WORDSIZE 4
# undef SUPPORTED_PERSONALITIES
# if defined(SPARC64)
# include <asm/psrcompat.h>
# define SUPPORTED_PERSONALITIES 3
# define PERSONALITY2_WORDSIZE 8
# else
# include <asm/psr.h>
# define SUPPORTED_PERSONALITIES 2
# endif /* SPARC64 */
#endif /* LINUXSPARC */
#ifdef X86_64
# undef SUPPORTED_PERSONALITIES
# define SUPPORTED_PERSONALITIES 2
# define PERSONALITY0_WORDSIZE 8
# define PERSONALITY1_WORDSIZE 4
#endif
#ifdef ARM
# undef SUPPORTED_PERSONALITIES
# define SUPPORTED_PERSONALITIES 2
# define PERSONALITY0_WORDSIZE 4
# define PERSONALITY1_WORDSIZE 4
#endif
#ifdef POWERPC64
# undef SUPPORTED_PERSONALITIES
# define SUPPORTED_PERSONALITIES 2
# define PERSONALITY0_WORDSIZE 8
# define PERSONALITY1_WORDSIZE 4
#endif
#if !HAVE_DECL_PTRACE_SETOPTIONS
# define PTRACE_SETOPTIONS 0x4200
#endif
#if !HAVE_DECL_PTRACE_GETEVENTMSG
# define PTRACE_GETEVENTMSG 0x4201
#endif
#if !HAVE_DECL_PTRACE_GETSIGINFO
# define PTRACE_GETSIGINFO 0x4202
#endif
#if !HAVE_DECL_PTRACE_O_TRACESYSGOOD
# define PTRACE_O_TRACESYSGOOD 0x00000001
#endif
#if !HAVE_DECL_PTRACE_O_TRACEFORK
# define PTRACE_O_TRACEFORK 0x00000002
#endif
#if !HAVE_DECL_PTRACE_O_TRACEVFORK
# define PTRACE_O_TRACEVFORK 0x00000004
#endif
#if !HAVE_DECL_PTRACE_O_TRACECLONE
# define PTRACE_O_TRACECLONE 0x00000008
#endif
#if !HAVE_DECL_PTRACE_O_TRACEEXEC
# define PTRACE_O_TRACEEXEC 0x00000010
#endif
#if !HAVE_DECL_PTRACE_O_TRACEEXIT
# define PTRACE_O_TRACEEXIT 0x00000040
#endif
#if !HAVE_DECL_PTRACE_EVENT_FORK
# define PTRACE_EVENT_FORK 1
#endif
#if !HAVE_DECL_PTRACE_EVENT_VFORK
# define PTRACE_EVENT_VFORK 2
#endif
#if !HAVE_DECL_PTRACE_EVENT_CLONE
# define PTRACE_EVENT_CLONE 3
#endif
#if !HAVE_DECL_PTRACE_EVENT_EXEC
# define PTRACE_EVENT_EXEC 4
#endif
#if !HAVE_DECL_PTRACE_EVENT_VFORK_DONE
# define PTRACE_EVENT_VFORK_DONE 5
#endif
#if !HAVE_DECL_PTRACE_EVENT_EXIT
# define PTRACE_EVENT_EXIT 6
#endif
/* Experimental code using PTRACE_SEIZE can be enabled here: */
//# define USE_SEIZE 1
#ifdef USE_SEIZE
# undef PTRACE_SEIZE
# define PTRACE_SEIZE 0x4206
# undef PTRACE_INTERRUPT
# define PTRACE_INTERRUPT 0x4207
# undef PTRACE_LISTEN
# define PTRACE_LISTEN 0x4208
# undef PTRACE_SEIZE_DEVEL
# define PTRACE_SEIZE_DEVEL 0x80000000
# undef PTRACE_EVENT_STOP
# define PTRACE_EVENT_STOP 7
# define PTRACE_EVENT_STOP1 128
#endif
#if !defined __GNUC__
# define __attribute__(x) /*nothing*/
#endif
#define ARRAY_SIZE(a) (sizeof(a) / sizeof((a)[0]))
#if defined(I386)
extern struct pt_regs i386_regs;
#endif
/* Trace Control Block */
struct tcb {
int flags; /* See below for TCB_ values */
int pid; /* Process Id of this entry */
int u_nargs; /* System call argument count */
int u_error; /* Error code */
long scno; /* System call number */
long u_arg[MAX_ARGS]; /* System call arguments */
#if defined(LINUX_MIPSN32)
long long ext_arg[MAX_ARGS]; /* System call arguments */
#endif
long u_rval; /* (first) return value */
#ifdef HAVE_LONG_LONG
long long u_lrval; /* long long return value */
#endif
int ptrace_errno;
#if SUPPORTED_PERSONALITIES > 1
int currpers; /* Personality at the time of scno update */
#endif
int curcol; /* Output column for this process */
FILE *outf; /* Output file for this process */
const char *auxstr; /* Auxiliary info from syscall (see RVAL_STR) */
struct timeval stime; /* System time usage as of last process wait */
struct timeval dtime; /* Delta for system time usage */
struct timeval etime; /* Syscall entry time */
/* Support for tracing forked processes */
long baddr; /* `Breakpoint' address */
long inst[2]; /* Instructions on above */
};
/* TCB flags */
#define TCB_INUSE 00001 /* This table entry is in use */
/* We have attached to this process, but did not see it stopping yet.
* (If this bit is not set, we either didn't attach yet,
* or we did attach to it, already saw it stopping at least once,
* did some init work on it and cleared this bit. TODO: maybe it makes sense
* to split these two states?)
*/
#define TCB_STARTUP 00002
#define TCB_IGNORE_ONE_SIGSTOP 00004 /* Next SIGSTOP is to be ignored */
/*
* Are we in system call entry or in syscall exit?
*
* This bit is set after all syscall entry processing is done.
* Therefore, this bit will be set when next ptrace stop occurs,
* which should be syscall exit stop. Other stops which are possible
* directly after syscall entry (death, ptrace event stop)
* are simpler and handled without calling trace_syscall(), therefore
* the places where TCB_INSYSCALL can be set but we aren't in syscall stop
* are limited to trace(), this condition is never observed in trace_syscall()
* and below.
* The bit is cleared after all syscall exit processing is done.
* User-generated SIGTRAPs and post-execve SIGTRAP make it necessary
* to be very careful and NOT set TCB_INSYSCALL bit when they are encountered.
* TCB_WAITEXECVE bit is used for this purpose (see below).
*
* Use entering(tcp) / exiting(tcp) to check this bit to make code more readable.
*/
#define TCB_INSYSCALL 00010
#define TCB_ATTACHED 00020 /* It is attached already */
/* Are we PROG from "strace PROG [ARGS]" invocation? */
#define TCB_STRACE_CHILD 00040
#define TCB_BPTSET 00100 /* "Breakpoint" set after fork(2) */
#define TCB_REPRINT 01000 /* We should reprint this syscall on exit */
#define TCB_FILTERED 02000 /* This system call has been filtered out */
/* x86 does not need TCB_WAITEXECVE.
* It can detect SIGTRAP by looking at eax/rax.
* See "not a syscall entry (eax = %ld)\n" message
* in syscall_fixup_on_sysenter().
*/
#if defined(ALPHA) || defined(AVR32) || defined(SPARC) || defined(SPARC64) \
|| defined(POWERPC) || defined(IA64) || defined(HPPA) \
|| defined(SH) || defined(SH64) || defined(S390) || defined(S390X) \
|| defined(ARM) || defined(MIPS) || defined(BFIN) || defined(TILE)
/* This tracee has entered into execve syscall. Expect post-execve SIGTRAP
* to happen. (When it is detected, tracee is continued and this bit is cleared.)
*/
# define TCB_WAITEXECVE 04000
#endif
#include <sys/syscall.h>
/* qualifier flags */
#define QUAL_TRACE 0001 /* this system call should be traced */
#define QUAL_ABBREV 0002 /* abbreviate the structures of this syscall */
#define QUAL_VERBOSE 0004 /* decode the structures of this syscall */
#define QUAL_RAW 0010 /* print all args in hex for this syscall */
#define QUAL_SIGNAL 0020 /* report events with this signal */
#define QUAL_FAULT 0040 /* report events with this fault */
#define QUAL_READ 0100 /* dump data read on this file descriptor */
#define QUAL_WRITE 0200 /* dump data written to this file descriptor */
#define entering(tcp) (!((tcp)->flags & TCB_INSYSCALL))
#define exiting(tcp) ((tcp)->flags & TCB_INSYSCALL)
#define syserror(tcp) ((tcp)->u_error != 0)
#define verbose(tcp) (qual_flags[(tcp)->scno] & QUAL_VERBOSE)
#define abbrev(tcp) (qual_flags[(tcp)->scno] & QUAL_ABBREV)
#define filtered(tcp) ((tcp)->flags & TCB_FILTERED)
struct xlat {
int val;
const char *str;
};
extern const struct xlat open_mode_flags[];
extern const struct xlat addrfams[];
extern const struct xlat struct_user_offsets[];
extern const struct xlat open_access_modes[];
/* Format of syscall return values */
#define RVAL_DECIMAL 000 /* decimal format */
#define RVAL_HEX 001 /* hex format */
#define RVAL_OCTAL 002 /* octal format */
#define RVAL_UDECIMAL 003 /* unsigned decimal format */
#define RVAL_LDECIMAL 004 /* long decimal format */
#define RVAL_LHEX 005 /* long hex format */
#define RVAL_LOCTAL 006 /* long octal format */
#define RVAL_LUDECIMAL 007 /* long unsigned decimal format */
#define RVAL_MASK 007 /* mask for these values */
#define RVAL_STR 010 /* Print `auxstr' field after return val */
#define RVAL_NONE 020 /* Print nothing */
#ifndef offsetof
# define offsetof(type, member) (((char *) &(((type *) NULL)->member)) - \
((char *) (type *) NULL))
#endif
/* get offset of member within a user struct */
#define uoff(member) offsetof(struct user, member)
#define TRACE_FILE 001 /* Trace file-related syscalls. */
#define TRACE_IPC 002 /* Trace IPC-related syscalls. */
#define TRACE_NETWORK 004 /* Trace network-related syscalls. */
#define TRACE_PROCESS 010 /* Trace process-related syscalls. */
#define TRACE_SIGNAL 020 /* Trace signal-related syscalls. */
#define TRACE_DESC 040 /* Trace file descriptor-related syscalls. */
#define SYSCALL_NEVER_FAILS 0100 /* Syscall is always successful. */
typedef enum {
CFLAG_NONE = 0,
CFLAG_ONLY_STATS,
CFLAG_BOTH
} cflag_t;
extern int *qual_flags;
extern int debug, followfork;
extern unsigned int ptrace_setoptions;
extern int dtime, xflag, qflag;
extern cflag_t cflag;
extern int max_strlen;
/*
* Which tcb has incomplete line being printed right now?
* NULL if last line has been completed ('\n'-terminated).
* printleader(tcp) sets it. Clearing is open-coded.
*/
extern struct tcb *printing_tcp;
enum bitness_t { BITNESS_CURRENT = 0, BITNESS_32 };
void error_msg(const char *fmt, ...) __attribute__ ((format(printf, 1, 2)));
void perror_msg(const char *fmt, ...) __attribute__ ((format(printf, 1, 2)));
void error_msg_and_die(const char *fmt, ...) __attribute__ ((noreturn, format(printf, 1, 2)));
void perror_msg_and_die(const char *fmt, ...) __attribute__ ((noreturn, format(printf, 1, 2)));
void die_out_of_memory(void) __attribute__ ((noreturn));
extern void set_personality(int personality);
extern const char *xlookup(const struct xlat *, int);
extern struct tcb *alloc_tcb(int, int);
extern void droptcb(struct tcb *);
#define alloctcb(pid) alloc_tcb((pid), 1)
extern void set_sortby(const char *);
extern void set_overhead(int);
extern void qualify(const char *);
extern int ptrace_restart(int request, struct tcb *tcp, int sig);
extern int trace_syscall(struct tcb *);
extern void count_syscall(struct tcb *, struct timeval *);
extern void printxval(const struct xlat *, int, const char *);
extern int printargs(struct tcb *);
extern int printargs_lu(struct tcb *);
extern int printargs_ld(struct tcb *);
extern void addflags(const struct xlat *, int);
extern int printflags(const struct xlat *, int, const char *);
extern const char *sprintflags(const char *, const struct xlat *, int);
extern int umoven(struct tcb *, long, int, char *);
extern int umovestr(struct tcb *, long, int, char *);
extern int upeek(struct tcb *, long, long *);
extern void dumpiov(struct tcb *, int, long);
extern void dumpstr(struct tcb *, long, int);
extern void printstr(struct tcb *, long, int);
extern void printnum(struct tcb *, long, const char *);
extern void printnum_int(struct tcb *, long, const char *);
extern void printpath(struct tcb *, long);
extern void printpathn(struct tcb *, long, int);
#define TIMESPEC_TEXT_BUFSIZE (sizeof(long)*3 * 2 + sizeof("{%u, %u}"))
#define TIMEVAL_TEXT_BUFSIZE TIMESPEC_TEXT_BUFSIZE
extern void printtv_bitness(struct tcb *, long, enum bitness_t, int);
extern char *sprinttv(char *, struct tcb *, long, enum bitness_t, int special);
extern void print_timespec(struct tcb *, long);
extern void sprint_timespec(char *, struct tcb *, long);
#ifdef HAVE_SIGINFO_T
extern void printsiginfo(siginfo_t *, int);
#endif
extern const char *getfdpath(struct tcb *, int);
extern void printfd(struct tcb *, int);
extern void printsock(struct tcb *, long, int);
extern void print_sock_optmgmt(struct tcb *, long, int);
extern void printrusage(struct tcb *, long);
#ifdef ALPHA
extern void printrusage32(struct tcb *, long);
#endif
extern void printuid(const char *, unsigned long);
extern int clearbpt(struct tcb *);
/*
* On Linux, "setbpt" is a misnomer: we don't set a breakpoint
* (IOW: no poking in user's text segment),
* instead we change fork/vfork/clone into clone(CLONE_PTRACE).
* On newer kernels, we use PTRACE_O_TRACECLONE/TRACE[V]FORK instead.
*/
extern int setbpt(struct tcb *);
extern void printcall(struct tcb *);
extern const char *signame(int);
extern void print_sigset(struct tcb *, long, int);
extern void printsignal(int);
extern void printleader(struct tcb *);
extern void tabto(void);
extern void call_summary(FILE *);
extern void tprint_iov(struct tcb *, unsigned long, unsigned long, int decode_iov);
extern void tprint_open_modes(mode_t);
extern const char *sprint_open_modes(mode_t);
extern int is_restart_error(struct tcb *);
extern int pathtrace_select(const char *);
extern int pathtrace_match(struct tcb *);
extern int change_syscall(struct tcb *, int);
extern int internal_fork(struct tcb *);
extern int internal_exec(struct tcb *);
extern const struct ioctlent *ioctl_lookup(long);
extern const struct ioctlent *ioctl_next_match(const struct ioctlent *);
extern int ioctl_decode(struct tcb *, long, long);
extern int term_ioctl(struct tcb *, long, long);
extern int sock_ioctl(struct tcb *, long, long);
extern int proc_ioctl(struct tcb *, int, int);
extern int rtc_ioctl(struct tcb *, long, long);
extern int scsi_ioctl(struct tcb *, long, long);
extern int block_ioctl(struct tcb *, long, long);
extern int tv_nz(struct timeval *);
extern int tv_cmp(struct timeval *, struct timeval *);
extern double tv_float(struct timeval *);
extern void tv_add(struct timeval *, struct timeval *, struct timeval *);
extern void tv_sub(struct timeval *, struct timeval *, struct timeval *);
extern void tv_mul(struct timeval *, struct timeval *, int);
extern void tv_div(struct timeval *, struct timeval *, int);
#if !defined HAVE_STPCPY
/* Some libc have stpcpy, some don't. Sigh...
* Roll our private implementation...
*/
#undef stpcpy
#define stpcpy strace_stpcpy
extern char *stpcpy(char *dst, const char *src);
#endif
#if defined(SPARC) || defined(SPARC64) || defined(IA64) || defined(SH)
extern long getrval2(struct tcb *);
#endif
#define umove(pid, addr, objp) \
umoven((pid), (addr), sizeof *(objp), (char *) (objp))
#define printtv(tcp, addr) \
printtv_bitness((tcp), (addr), BITNESS_CURRENT, 0)
#define printtv_special(tcp, addr) \
printtv_bitness((tcp), (addr), BITNESS_CURRENT, 1)
extern void tprintf(const char *fmt, ...) __attribute__ ((format (printf, 1, 2)));
extern void tprints(const char *str);
#ifndef HAVE_STRERROR
const char *strerror(int);
#endif
#ifndef HAVE_STRSIGNAL
const char *strsignal(int);
#endif
extern int current_personality;
extern const int personality_wordsize[];
struct sysent {
unsigned nargs;
int sys_flags;
int (*sys_func)();
const char *sys_name;
};
struct ioctlent {
const char *doth;
const char *symbol;
unsigned long code;
};
extern const struct sysent *sysent;
extern unsigned nsyscalls;
extern const char *const *errnoent;
extern unsigned nerrnos;
extern const struct ioctlent *ioctlent;
extern unsigned nioctlents;
extern const char *const *signalent;
extern unsigned nsignals;
#define SCNO_IN_RANGE(scno) \
((unsigned long)(scno) < nsyscalls && sysent[scno].sys_func)
#if HAVE_LONG_LONG
/* _l refers to the lower numbered u_arg,
* _h refers to the higher numbered u_arg
*/
#if HAVE_LITTLE_ENDIAN_LONG_LONG
#define LONG_LONG(_l,_h) \
((long long)((unsigned long long)(unsigned)(_l) | ((unsigned long long)(_h)<<32)))
#else
#define LONG_LONG(_l,_h) \
((long long)((unsigned long long)(unsigned)(_h) | ((unsigned long long)(_l)<<32)))
#endif
extern int printllval(struct tcb *, const char *, int);
#endif
#ifdef IA64
extern long ia32;
#endif
extern int not_failing_only;
extern int show_fd_path;
extern int tracing_paths;