binutils-gdb/gdb/procfs.c
Stu Grossman fb63d460a7 * procfs.c: Handle process exits more elegantly by trapping on
entry to _exit.  Also, cleanup procinfo list when process dies of
it's own accord (as opposed to being killed).
1993-10-17 09:12:58 +00:00

3572 lines
86 KiB
C
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

/* Machine independent support for SVR4 /proc (process file system) for GDB.
Copyright 1991, 1992 Free Software Foundation, Inc.
Written by Fred Fish at Cygnus Support.
This file is part of GDB.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
/* N O T E S
For information on the details of using /proc consult section proc(4)
in the UNIX System V Release 4 System Administrator's Reference Manual.
The general register and floating point register sets are manipulated by
separate ioctl's. This file makes the assumption that if FP0_REGNUM is
defined, then support for the floating point register set is desired,
regardless of whether or not the actual target has floating point hardware.
*/
#include "defs.h"
#include <time.h>
#include <sys/procfs.h>
#include <fcntl.h>
#include <errno.h>
#include <string.h>
#include <stropts.h>
#include <poll.h>
#include "inferior.h"
#include "target.h"
#include "command.h"
#include "gdbcore.h"
#define MAX_SYSCALLS 256 /* Maximum number of syscalls for table */
#ifndef PROC_NAME_FMT
#define PROC_NAME_FMT "/proc/%05d"
#endif
extern struct target_ops procfs_ops; /* Forward declaration */
#if 1 /* FIXME: Gross and ugly hack to resolve coredep.c global */
CORE_ADDR kernel_u_addr;
#endif
#ifdef BROKEN_SIGINFO_H /* Workaround broken SGS <sys/siginfo.h> */
#undef si_pid
#define si_pid _data._proc.pid
#undef si_uid
#define si_uid _data._proc._pdata._kill.uid
#endif /* BROKEN_SIGINFO_H */
/* All access to the inferior, either one started by gdb or one that has
been attached to, is controlled by an instance of a procinfo structure,
defined below. Since gdb currently only handles one inferior at a time,
the procinfo structure for the inferior is statically allocated and
only one exists at any given time. There is a separate procinfo
structure for use by the "info proc" command, so that we can print
useful information about any random process without interfering with
the inferior's procinfo information. */
struct procinfo {
struct procinfo *next;
int pid; /* Process ID of inferior */
int fd; /* File descriptor for /proc entry */
char *pathname; /* Pathname to /proc entry */
int had_event; /* poll/select says something happened */
int was_stopped; /* Nonzero if was stopped prior to attach */
int nopass_next_sigstop; /* Don't pass a sigstop on next resume */
prrun_t prrun; /* Control state when it is run */
prstatus_t prstatus; /* Current process status info */
gregset_t gregset; /* General register set */
fpregset_t fpregset; /* Floating point register set */
fltset_t fltset; /* Current traced hardware fault set */
sigset_t trace; /* Current traced signal set */
sysset_t exitset; /* Current traced system call exit set */
sysset_t entryset; /* Current traced system call entry set */
fltset_t saved_fltset; /* Saved traced hardware fault set */
sigset_t saved_trace; /* Saved traced signal set */
sigset_t saved_sighold; /* Saved held signal set */
sysset_t saved_exitset; /* Saved traced system call exit set */
sysset_t saved_entryset; /* Saved traced system call entry set */
};
/* List of inferior process information */
static struct procinfo *procinfo_list = NULL;
static struct pollfd *poll_list; /* pollfds used for waiting on /proc */
static int num_poll_list = 0; /* Number of entries in poll_list */
static int last_resume_pid = -1; /* Last pid used with procfs_resume */
/* Much of the information used in the /proc interface, particularly for
printing status information, is kept as tables of structures of the
following form. These tables can be used to map numeric values to
their symbolic names and to a string that describes their specific use. */
struct trans {
int value; /* The numeric value */
char *name; /* The equivalent symbolic value */
char *desc; /* Short description of value */
};
/* Translate bits in the pr_flags member of the prstatus structure, into the
names and desc information. */
static struct trans pr_flag_table[] =
{
#if defined (PR_STOPPED)
PR_STOPPED, "PR_STOPPED", "Process is stopped",
#endif
#if defined (PR_ISTOP)
PR_ISTOP, "PR_ISTOP", "Stopped on an event of interest",
#endif
#if defined (PR_DSTOP)
PR_DSTOP, "PR_DSTOP", "A stop directive is in effect",
#endif
#if defined (PR_ASLEEP)
PR_ASLEEP, "PR_ASLEEP", "Sleeping in an interruptible system call",
#endif
#if defined (PR_FORK)
PR_FORK, "PR_FORK", "Inherit-on-fork is in effect",
#endif
#if defined (PR_RLC)
PR_RLC, "PR_RLC", "Run-on-last-close is in effect",
#endif
#if defined (PR_PTRACE)
PR_PTRACE, "PR_PTRACE", "Process is being controlled by ptrace",
#endif
#if defined (PR_PCINVAL)
PR_PCINVAL, "PR_PCINVAL", "PC refers to an invalid virtual address",
#endif
#if defined (PR_ISSYS)
PR_ISSYS, "PR_ISSYS", "Is a system process",
#endif
#if defined (PR_STEP)
PR_STEP, "PR_STEP", "Process has single step pending",
#endif
#if defined (PR_KLC)
PR_KLC, "PR_KLC", "Kill-on-last-close is in effect",
#endif
#if defined (PR_ASYNC)
PR_ASYNC, "PR_ASYNC", "Asynchronous stop is in effect",
#endif
#if defined (PR_PCOMPAT)
PR_PCOMPAT, "PR_PCOMPAT", "Ptrace compatibility mode in effect",
#endif
0, NULL, NULL
};
/* Translate values in the pr_why field of the prstatus struct. */
static struct trans pr_why_table[] =
{
#if defined (PR_REQUESTED)
PR_REQUESTED, "PR_REQUESTED", "Directed to stop via PIOCSTOP/PIOCWSTOP",
#endif
#if defined (PR_SIGNALLED)
PR_SIGNALLED, "PR_SIGNALLED", "Receipt of a traced signal",
#endif
#if defined (PR_FAULTED)
PR_FAULTED, "PR_FAULTED", "Incurred a traced hardware fault",
#endif
#if defined (PR_SYSENTRY)
PR_SYSENTRY, "PR_SYSENTRY", "Entry to a traced system call",
#endif
#if defined (PR_SYSEXIT)
PR_SYSEXIT, "PR_SYSEXIT", "Exit from a traced system call",
#endif
#if defined (PR_JOBCONTROL)
PR_JOBCONTROL, "PR_JOBCONTROL", "Default job control stop signal action",
#endif
#if defined (PR_SUSPENDED)
PR_SUSPENDED, "PR_SUSPENDED", "Process suspended",
#endif
0, NULL, NULL
};
/* Hardware fault translation table. */
static struct trans faults_table[] =
{
#if defined (FLTILL)
FLTILL, "FLTILL", "Illegal instruction",
#endif
#if defined (FLTPRIV)
FLTPRIV, "FLTPRIV", "Privileged instruction",
#endif
#if defined (FLTBPT)
FLTBPT, "FLTBPT", "Breakpoint trap",
#endif
#if defined (FLTTRACE)
FLTTRACE, "FLTTRACE", "Trace trap",
#endif
#if defined (FLTACCESS)
FLTACCESS, "FLTACCESS", "Memory access fault",
#endif
#if defined (FLTBOUNDS)
FLTBOUNDS, "FLTBOUNDS", "Memory bounds violation",
#endif
#if defined (FLTIOVF)
FLTIOVF, "FLTIOVF", "Integer overflow",
#endif
#if defined (FLTIZDIV)
FLTIZDIV, "FLTIZDIV", "Integer zero divide",
#endif
#if defined (FLTFPE)
FLTFPE, "FLTFPE", "Floating-point exception",
#endif
#if defined (FLTSTACK)
FLTSTACK, "FLTSTACK", "Unrecoverable stack fault",
#endif
#if defined (FLTPAGE)
FLTPAGE, "FLTPAGE", "Recoverable page fault",
#endif
0, NULL, NULL
};
/* Translation table for signal generation information. See UNIX System
V Release 4 Programmer's Reference Manual, siginfo(5). */
static struct sigcode {
int signo;
int code;
char *codename;
char *desc;
} siginfo_table[] = {
#if defined (SIGILL) && defined (ILL_ILLOPC)
SIGILL, ILL_ILLOPC, "ILL_ILLOPC", "Illegal opcode",
#endif
#if defined (SIGILL) && defined (ILL_ILLOPN)
SIGILL, ILL_ILLOPN, "ILL_ILLOPN", "Illegal operand",
#endif
#if defined (SIGILL) && defined (ILL_ILLADR)
SIGILL, ILL_ILLADR, "ILL_ILLADR", "Illegal addressing mode",
#endif
#if defined (SIGILL) && defined (ILL_ILLTRP)
SIGILL, ILL_ILLTRP, "ILL_ILLTRP", "Illegal trap",
#endif
#if defined (SIGILL) && defined (ILL_PRVOPC)
SIGILL, ILL_PRVOPC, "ILL_PRVOPC", "Privileged opcode",
#endif
#if defined (SIGILL) && defined (ILL_PRVREG)
SIGILL, ILL_PRVREG, "ILL_PRVREG", "Privileged register",
#endif
#if defined (SIGILL) && defined (ILL_COPROC)
SIGILL, ILL_COPROC, "ILL_COPROC", "Coprocessor error",
#endif
#if defined (SIGILL) && defined (ILL_BADSTK)
SIGILL, ILL_BADSTK, "ILL_BADSTK", "Internal stack error",
#endif
#if defined (SIGFPE) && defined (FPE_INTDIV)
SIGFPE, FPE_INTDIV, "FPE_INTDIV", "Integer divide by zero",
#endif
#if defined (SIGFPE) && defined (FPE_INTOVF)
SIGFPE, FPE_INTOVF, "FPE_INTOVF", "Integer overflow",
#endif
#if defined (SIGFPE) && defined (FPE_FLTDIV)
SIGFPE, FPE_FLTDIV, "FPE_FLTDIV", "Floating point divide by zero",
#endif
#if defined (SIGFPE) && defined (FPE_FLTOVF)
SIGFPE, FPE_FLTOVF, "FPE_FLTOVF", "Floating point overflow",
#endif
#if defined (SIGFPE) && defined (FPE_FLTUND)
SIGFPE, FPE_FLTUND, "FPE_FLTUND", "Floating point underflow",
#endif
#if defined (SIGFPE) && defined (FPE_FLTRES)
SIGFPE, FPE_FLTRES, "FPE_FLTRES", "Floating point inexact result",
#endif
#if defined (SIGFPE) && defined (FPE_FLTINV)
SIGFPE, FPE_FLTINV, "FPE_FLTINV", "Invalid floating point operation",
#endif
#if defined (SIGFPE) && defined (FPE_FLTSUB)
SIGFPE, FPE_FLTSUB, "FPE_FLTSUB", "Subscript out of range",
#endif
#if defined (SIGSEGV) && defined (SEGV_MAPERR)
SIGSEGV, SEGV_MAPERR, "SEGV_MAPERR", "Address not mapped to object",
#endif
#if defined (SIGSEGV) && defined (SEGV_ACCERR)
SIGSEGV, SEGV_ACCERR, "SEGV_ACCERR", "Invalid permissions for object",
#endif
#if defined (SIGBUS) && defined (BUS_ADRALN)
SIGBUS, BUS_ADRALN, "BUS_ADRALN", "Invalid address alignment",
#endif
#if defined (SIGBUS) && defined (BUS_ADRERR)
SIGBUS, BUS_ADRERR, "BUS_ADRERR", "Non-existent physical address",
#endif
#if defined (SIGBUS) && defined (BUS_OBJERR)
SIGBUS, BUS_OBJERR, "BUS_OBJERR", "Object specific hardware error",
#endif
#if defined (SIGTRAP) && defined (TRAP_BRKPT)
SIGTRAP, TRAP_BRKPT, "TRAP_BRKPT", "Process breakpoint",
#endif
#if defined (SIGTRAP) && defined (TRAP_TRACE)
SIGTRAP, TRAP_TRACE, "TRAP_TRACE", "Process trace trap",
#endif
#if defined (SIGCLD) && defined (CLD_EXITED)
SIGCLD, CLD_EXITED, "CLD_EXITED", "Child has exited",
#endif
#if defined (SIGCLD) && defined (CLD_KILLED)
SIGCLD, CLD_KILLED, "CLD_KILLED", "Child was killed",
#endif
#if defined (SIGCLD) && defined (CLD_DUMPED)
SIGCLD, CLD_DUMPED, "CLD_DUMPED", "Child has terminated abnormally",
#endif
#if defined (SIGCLD) && defined (CLD_TRAPPED)
SIGCLD, CLD_TRAPPED, "CLD_TRAPPED", "Traced child has trapped",
#endif
#if defined (SIGCLD) && defined (CLD_STOPPED)
SIGCLD, CLD_STOPPED, "CLD_STOPPED", "Child has stopped",
#endif
#if defined (SIGCLD) && defined (CLD_CONTINUED)
SIGCLD, CLD_CONTINUED, "CLD_CONTINUED", "Stopped child had continued",
#endif
#if defined (SIGPOLL) && defined (POLL_IN)
SIGPOLL, POLL_IN, "POLL_IN", "Input input available",
#endif
#if defined (SIGPOLL) && defined (POLL_OUT)
SIGPOLL, POLL_OUT, "POLL_OUT", "Output buffers available",
#endif
#if defined (SIGPOLL) && defined (POLL_MSG)
SIGPOLL, POLL_MSG, "POLL_MSG", "Input message available",
#endif
#if defined (SIGPOLL) && defined (POLL_ERR)
SIGPOLL, POLL_ERR, "POLL_ERR", "I/O error",
#endif
#if defined (SIGPOLL) && defined (POLL_PRI)
SIGPOLL, POLL_PRI, "POLL_PRI", "High priority input available",
#endif
#if defined (SIGPOLL) && defined (POLL_HUP)
SIGPOLL, POLL_HUP, "POLL_HUP", "Device disconnected",
#endif
0, 0, NULL, NULL
};
static char *syscall_table[MAX_SYSCALLS];
/* Prototypes for local functions */
static void
set_proc_siginfo PARAMS ((struct procinfo *, int));
static void
init_syscall_table PARAMS ((void));
static char *
syscallname PARAMS ((int));
static char *
signalname PARAMS ((int));
static char *
errnoname PARAMS ((int));
static int
proc_address_to_fd PARAMS ((struct procinfo *, CORE_ADDR, int));
static int
open_proc_file PARAMS ((int, struct procinfo *, int));
static void
close_proc_file PARAMS ((struct procinfo *));
static void
unconditionally_kill_inferior PARAMS ((struct procinfo *));
static NORETURN void
proc_init_failed PARAMS ((struct procinfo *, char *));
static void
info_proc PARAMS ((char *, int));
static void
info_proc_flags PARAMS ((struct procinfo *, int));
static void
info_proc_stop PARAMS ((struct procinfo *, int));
static void
info_proc_siginfo PARAMS ((struct procinfo *, int));
static void
info_proc_syscalls PARAMS ((struct procinfo *, int));
static void
info_proc_mappings PARAMS ((struct procinfo *, int));
static void
info_proc_signals PARAMS ((struct procinfo *, int));
static void
info_proc_faults PARAMS ((struct procinfo *, int));
static char *
mappingflags PARAMS ((long));
static char *
lookupname PARAMS ((struct trans *, unsigned int, char *));
static char *
lookupdesc PARAMS ((struct trans *, unsigned int));
static int
do_attach PARAMS ((int pid));
static void
do_detach PARAMS ((int siggnal));
static void
procfs_create_inferior PARAMS ((char *, char *, char **));
static void
procfs_notice_signals PARAMS ((int pid));
static struct procinfo *
find_procinfo PARAMS ((pid_t pid, int okfail));
/* External function prototypes that can't be easily included in any
header file because the args are typedefs in system include files. */
extern void
supply_gregset PARAMS ((gregset_t *));
extern void
fill_gregset PARAMS ((gregset_t *, int));
extern void
supply_fpregset PARAMS ((fpregset_t *));
extern void
fill_fpregset PARAMS ((fpregset_t *, int));
/*
LOCAL FUNCTION
find_procinfo -- convert a process id to a struct procinfo
SYNOPSIS
static struct procinfo * find_procinfo (pid_t pid, int okfail);
DESCRIPTION
Given a process id, look it up in the procinfo chain. Returns
a struct procinfo *. If can't find pid, then call error(),
unless okfail is set, in which case, return NULL;
*/
static struct procinfo *
find_procinfo (pid, okfail)
pid_t pid;
int okfail;
{
struct procinfo *procinfo;
for (procinfo = procinfo_list; procinfo; procinfo = procinfo->next)
if (procinfo->pid == pid)
return procinfo;
if (okfail)
return NULL;
error ("procfs (find_procinfo): Couldn't locate pid %d", pid);
}
/*
LOCAL MACRO
current_procinfo -- convert inferior_pid to a struct procinfo
SYNOPSIS
static struct procinfo * current_procinfo;
DESCRIPTION
Looks up inferior_pid in the procinfo chain. Always returns a
struct procinfo *. If process can't be found, we error() out.
*/
#define current_procinfo find_procinfo (inferior_pid, 0)
/*
LOCAL FUNCTION
add_fd -- Add the fd to the poll/select list
SYNOPSIS
static void add_fd (struct procinfo *);
DESCRIPTION
Add the fd of the supplied procinfo to the list of fds used for
poll/select operations.
*/
static void
add_fd (pi)
struct procinfo *pi;
{
if (num_poll_list <= 0)
poll_list = (struct pollfd *) xmalloc (sizeof (struct pollfd));
else
poll_list = (struct pollfd *) xrealloc (poll_list,
(num_poll_list + 1)
* sizeof (struct pollfd));
poll_list[num_poll_list].fd = pi->fd;
poll_list[num_poll_list].events = POLLPRI;
num_poll_list++;
}
static void
remove_fd (pi)
struct procinfo *pi;
{
int i;
for (i = 0; i < num_poll_list; i++)
{
if (poll_list[i].fd == pi->fd)
{
if (i != num_poll_list - 1)
memcpy (poll_list, poll_list + i + 1,
(num_poll_list - i - 1) * sizeof (struct pollfd));
num_poll_list--;
if (num_poll_list == 0)
free (poll_list);
else
poll_list = (struct pollfd *) xrealloc (poll_list,
num_poll_list
* sizeof (struct pollfd));
return;
}
}
}
#define LOSING_POLL unixware_sux
static struct procinfo *
wait_fd ()
{
struct procinfo *pi;
int num_fds;
int i;
if (attach_flag)
set_sigint_trap (); /* Causes SIGINT to be passed on to the
attached process. */
#ifndef LOSING_POLL
num_fds = poll (poll_list, num_poll_list, -1);
#else
pi = current_procinfo;
if (ioctl (pi->fd, PIOCWSTOP, &pi->prstatus) < 0)
{
print_sys_errmsg (pi->pathname, errno);
error ("PIOCWSTOP failed");
}
pi->had_event = 1;
#endif
if (attach_flag)
clear_sigint_trap();
#ifndef LOSING_POLL
if (num_fds <= 0)
{
print_sys_errmsg ("poll failed\n", errno);
error ("Poll failed, returned %d", num_fds);
}
for (i = 0; i < num_poll_list && num_fds > 0; i++)
{
if ((poll_list[i].revents & (POLLPRI|POLLERR|POLLHUP|POLLNVAL)) == 0)
continue;
for (pi = procinfo_list; pi; pi = pi->next)
{
if (poll_list[i].fd == pi->fd)
{
if (ioctl (pi->fd, PIOCSTATUS, &pi->prstatus) < 0)
{
print_sys_errmsg (pi->pathname, errno);
error ("PIOCSTATUS failed");
}
num_fds--;
pi->had_event = 1;
break;
}
}
if (!pi)
error ("procfs_wait: Couldn't find procinfo for fd %d\n",
poll_list[i].fd);
}
#endif /* LOSING_POLL */
return pi;
}
/*
LOCAL FUNCTION
lookupdesc -- translate a value to a summary desc string
SYNOPSIS
static char *lookupdesc (struct trans *transp, unsigned int val);
DESCRIPTION
Given a pointer to a translation table and a value to be translated,
lookup the desc string and return it.
*/
static char *
lookupdesc (transp, val)
struct trans *transp;
unsigned int val;
{
char *desc;
for (desc = NULL; transp -> name != NULL; transp++)
{
if (transp -> value == val)
{
desc = transp -> desc;
break;
}
}
/* Didn't find a translation for the specified value, set a default one. */
if (desc == NULL)
{
desc = "Unknown";
}
return (desc);
}
/*
LOCAL FUNCTION
lookupname -- translate a value to symbolic name
SYNOPSIS
static char *lookupname (struct trans *transp, unsigned int val,
char *prefix);
DESCRIPTION
Given a pointer to a translation table, a value to be translated,
and a default prefix to return if the value can't be translated,
match the value with one of the translation table entries and
return a pointer to the symbolic name.
If no match is found it just returns the value as a printable string,
with the given prefix. The previous such value, if any, is freed
at this time.
*/
static char *
lookupname (transp, val, prefix)
struct trans *transp;
unsigned int val;
char *prefix;
{
static char *locbuf;
char *name;
for (name = NULL; transp -> name != NULL; transp++)
{
if (transp -> value == val)
{
name = transp -> name;
break;
}
}
/* Didn't find a translation for the specified value, build a default
one using the specified prefix and return it. The lifetime of
the value is only until the next one is needed. */
if (name == NULL)
{
if (locbuf != NULL)
{
free (locbuf);
}
locbuf = xmalloc (strlen (prefix) + 16);
sprintf (locbuf, "%s %u", prefix, val);
name = locbuf;
}
return (name);
}
static char *
sigcodename (sip)
siginfo_t *sip;
{
struct sigcode *scp;
char *name = NULL;
static char locbuf[32];
for (scp = siginfo_table; scp -> codename != NULL; scp++)
{
if ((scp -> signo == sip -> si_signo) &&
(scp -> code == sip -> si_code))
{
name = scp -> codename;
break;
}
}
if (name == NULL)
{
sprintf (locbuf, "sigcode %u", sip -> si_signo);
name = locbuf;
}
return (name);
}
static char *
sigcodedesc (sip)
siginfo_t *sip;
{
struct sigcode *scp;
char *desc = NULL;
for (scp = siginfo_table; scp -> codename != NULL; scp++)
{
if ((scp -> signo == sip -> si_signo) &&
(scp -> code == sip -> si_code))
{
desc = scp -> desc;
break;
}
}
if (desc == NULL)
{
desc = "Unrecognized signal or trap use";
}
return (desc);
}
/*
LOCAL FUNCTION
syscallname - translate a system call number into a system call name
SYNOPSIS
char *syscallname (int syscallnum)
DESCRIPTION
Given a system call number, translate it into the printable name
of a system call, or into "syscall <num>" if it is an unknown
number.
*/
static char *
syscallname (syscallnum)
int syscallnum;
{
static char locbuf[32];
char *rtnval;
if (syscallnum >= 0 && syscallnum < MAX_SYSCALLS)
{
rtnval = syscall_table[syscallnum];
}
else
{
sprintf (locbuf, "syscall %u", syscallnum);
rtnval = locbuf;
}
return (rtnval);
}
/*
LOCAL FUNCTION
init_syscall_table - initialize syscall translation table
SYNOPSIS
void init_syscall_table (void)
DESCRIPTION
Dynamically initialize the translation table to convert system
call numbers into printable system call names. Done once per
gdb run, on initialization.
NOTES
This is awfully ugly, but preprocessor tricks to make it prettier
tend to be nonportable.
*/
static void
init_syscall_table ()
{
#if defined (SYS_exit)
syscall_table[SYS_exit] = "exit";
#endif
#if defined (SYS_fork)
syscall_table[SYS_fork] = "fork";
#endif
#if defined (SYS_read)
syscall_table[SYS_read] = "read";
#endif
#if defined (SYS_write)
syscall_table[SYS_write] = "write";
#endif
#if defined (SYS_open)
syscall_table[SYS_open] = "open";
#endif
#if defined (SYS_close)
syscall_table[SYS_close] = "close";
#endif
#if defined (SYS_wait)
syscall_table[SYS_wait] = "wait";
#endif
#if defined (SYS_creat)
syscall_table[SYS_creat] = "creat";
#endif
#if defined (SYS_link)
syscall_table[SYS_link] = "link";
#endif
#if defined (SYS_unlink)
syscall_table[SYS_unlink] = "unlink";
#endif
#if defined (SYS_exec)
syscall_table[SYS_exec] = "exec";
#endif
#if defined (SYS_execv)
syscall_table[SYS_execv] = "execv";
#endif
#if defined (SYS_execve)
syscall_table[SYS_execve] = "execve";
#endif
#if defined (SYS_chdir)
syscall_table[SYS_chdir] = "chdir";
#endif
#if defined (SYS_time)
syscall_table[SYS_time] = "time";
#endif
#if defined (SYS_mknod)
syscall_table[SYS_mknod] = "mknod";
#endif
#if defined (SYS_chmod)
syscall_table[SYS_chmod] = "chmod";
#endif
#if defined (SYS_chown)
syscall_table[SYS_chown] = "chown";
#endif
#if defined (SYS_brk)
syscall_table[SYS_brk] = "brk";
#endif
#if defined (SYS_stat)
syscall_table[SYS_stat] = "stat";
#endif
#if defined (SYS_lseek)
syscall_table[SYS_lseek] = "lseek";
#endif
#if defined (SYS_getpid)
syscall_table[SYS_getpid] = "getpid";
#endif
#if defined (SYS_mount)
syscall_table[SYS_mount] = "mount";
#endif
#if defined (SYS_umount)
syscall_table[SYS_umount] = "umount";
#endif
#if defined (SYS_setuid)
syscall_table[SYS_setuid] = "setuid";
#endif
#if defined (SYS_getuid)
syscall_table[SYS_getuid] = "getuid";
#endif
#if defined (SYS_stime)
syscall_table[SYS_stime] = "stime";
#endif
#if defined (SYS_ptrace)
syscall_table[SYS_ptrace] = "ptrace";
#endif
#if defined (SYS_alarm)
syscall_table[SYS_alarm] = "alarm";
#endif
#if defined (SYS_fstat)
syscall_table[SYS_fstat] = "fstat";
#endif
#if defined (SYS_pause)
syscall_table[SYS_pause] = "pause";
#endif
#if defined (SYS_utime)
syscall_table[SYS_utime] = "utime";
#endif
#if defined (SYS_stty)
syscall_table[SYS_stty] = "stty";
#endif
#if defined (SYS_gtty)
syscall_table[SYS_gtty] = "gtty";
#endif
#if defined (SYS_access)
syscall_table[SYS_access] = "access";
#endif
#if defined (SYS_nice)
syscall_table[SYS_nice] = "nice";
#endif
#if defined (SYS_statfs)
syscall_table[SYS_statfs] = "statfs";
#endif
#if defined (SYS_sync)
syscall_table[SYS_sync] = "sync";
#endif
#if defined (SYS_kill)
syscall_table[SYS_kill] = "kill";
#endif
#if defined (SYS_fstatfs)
syscall_table[SYS_fstatfs] = "fstatfs";
#endif
#if defined (SYS_pgrpsys)
syscall_table[SYS_pgrpsys] = "pgrpsys";
#endif
#if defined (SYS_xenix)
syscall_table[SYS_xenix] = "xenix";
#endif
#if defined (SYS_dup)
syscall_table[SYS_dup] = "dup";
#endif
#if defined (SYS_pipe)
syscall_table[SYS_pipe] = "pipe";
#endif
#if defined (SYS_times)
syscall_table[SYS_times] = "times";
#endif
#if defined (SYS_profil)
syscall_table[SYS_profil] = "profil";
#endif
#if defined (SYS_plock)
syscall_table[SYS_plock] = "plock";
#endif
#if defined (SYS_setgid)
syscall_table[SYS_setgid] = "setgid";
#endif
#if defined (SYS_getgid)
syscall_table[SYS_getgid] = "getgid";
#endif
#if defined (SYS_signal)
syscall_table[SYS_signal] = "signal";
#endif
#if defined (SYS_msgsys)
syscall_table[SYS_msgsys] = "msgsys";
#endif
#if defined (SYS_sys3b)
syscall_table[SYS_sys3b] = "sys3b";
#endif
#if defined (SYS_acct)
syscall_table[SYS_acct] = "acct";
#endif
#if defined (SYS_shmsys)
syscall_table[SYS_shmsys] = "shmsys";
#endif
#if defined (SYS_semsys)
syscall_table[SYS_semsys] = "semsys";
#endif
#if defined (SYS_ioctl)
syscall_table[SYS_ioctl] = "ioctl";
#endif
#if defined (SYS_uadmin)
syscall_table[SYS_uadmin] = "uadmin";
#endif
#if defined (SYS_utssys)
syscall_table[SYS_utssys] = "utssys";
#endif
#if defined (SYS_fsync)
syscall_table[SYS_fsync] = "fsync";
#endif
#if defined (SYS_umask)
syscall_table[SYS_umask] = "umask";
#endif
#if defined (SYS_chroot)
syscall_table[SYS_chroot] = "chroot";
#endif
#if defined (SYS_fcntl)
syscall_table[SYS_fcntl] = "fcntl";
#endif
#if defined (SYS_ulimit)
syscall_table[SYS_ulimit] = "ulimit";
#endif
#if defined (SYS_rfsys)
syscall_table[SYS_rfsys] = "rfsys";
#endif
#if defined (SYS_rmdir)
syscall_table[SYS_rmdir] = "rmdir";
#endif
#if defined (SYS_mkdir)
syscall_table[SYS_mkdir] = "mkdir";
#endif
#if defined (SYS_getdents)
syscall_table[SYS_getdents] = "getdents";
#endif
#if defined (SYS_sysfs)
syscall_table[SYS_sysfs] = "sysfs";
#endif
#if defined (SYS_getmsg)
syscall_table[SYS_getmsg] = "getmsg";
#endif
#if defined (SYS_putmsg)
syscall_table[SYS_putmsg] = "putmsg";
#endif
#if defined (SYS_poll)
syscall_table[SYS_poll] = "poll";
#endif
#if defined (SYS_lstat)
syscall_table[SYS_lstat] = "lstat";
#endif
#if defined (SYS_symlink)
syscall_table[SYS_symlink] = "symlink";
#endif
#if defined (SYS_readlink)
syscall_table[SYS_readlink] = "readlink";
#endif
#if defined (SYS_setgroups)
syscall_table[SYS_setgroups] = "setgroups";
#endif
#if defined (SYS_getgroups)
syscall_table[SYS_getgroups] = "getgroups";
#endif
#if defined (SYS_fchmod)
syscall_table[SYS_fchmod] = "fchmod";
#endif
#if defined (SYS_fchown)
syscall_table[SYS_fchown] = "fchown";
#endif
#if defined (SYS_sigprocmask)
syscall_table[SYS_sigprocmask] = "sigprocmask";
#endif
#if defined (SYS_sigsuspend)
syscall_table[SYS_sigsuspend] = "sigsuspend";
#endif
#if defined (SYS_sigaltstack)
syscall_table[SYS_sigaltstack] = "sigaltstack";
#endif
#if defined (SYS_sigaction)
syscall_table[SYS_sigaction] = "sigaction";
#endif
#if defined (SYS_sigpending)
syscall_table[SYS_sigpending] = "sigpending";
#endif
#if defined (SYS_context)
syscall_table[SYS_context] = "context";
#endif
#if defined (SYS_evsys)
syscall_table[SYS_evsys] = "evsys";
#endif
#if defined (SYS_evtrapret)
syscall_table[SYS_evtrapret] = "evtrapret";
#endif
#if defined (SYS_statvfs)
syscall_table[SYS_statvfs] = "statvfs";
#endif
#if defined (SYS_fstatvfs)
syscall_table[SYS_fstatvfs] = "fstatvfs";
#endif
#if defined (SYS_nfssys)
syscall_table[SYS_nfssys] = "nfssys";
#endif
#if defined (SYS_waitsys)
syscall_table[SYS_waitsys] = "waitsys";
#endif
#if defined (SYS_sigsendsys)
syscall_table[SYS_sigsendsys] = "sigsendsys";
#endif
#if defined (SYS_hrtsys)
syscall_table[SYS_hrtsys] = "hrtsys";
#endif
#if defined (SYS_acancel)
syscall_table[SYS_acancel] = "acancel";
#endif
#if defined (SYS_async)
syscall_table[SYS_async] = "async";
#endif
#if defined (SYS_priocntlsys)
syscall_table[SYS_priocntlsys] = "priocntlsys";
#endif
#if defined (SYS_pathconf)
syscall_table[SYS_pathconf] = "pathconf";
#endif
#if defined (SYS_mincore)
syscall_table[SYS_mincore] = "mincore";
#endif
#if defined (SYS_mmap)
syscall_table[SYS_mmap] = "mmap";
#endif
#if defined (SYS_mprotect)
syscall_table[SYS_mprotect] = "mprotect";
#endif
#if defined (SYS_munmap)
syscall_table[SYS_munmap] = "munmap";
#endif
#if defined (SYS_fpathconf)
syscall_table[SYS_fpathconf] = "fpathconf";
#endif
#if defined (SYS_vfork)
syscall_table[SYS_vfork] = "vfork";
#endif
#if defined (SYS_fchdir)
syscall_table[SYS_fchdir] = "fchdir";
#endif
#if defined (SYS_readv)
syscall_table[SYS_readv] = "readv";
#endif
#if defined (SYS_writev)
syscall_table[SYS_writev] = "writev";
#endif
#if defined (SYS_xstat)
syscall_table[SYS_xstat] = "xstat";
#endif
#if defined (SYS_lxstat)
syscall_table[SYS_lxstat] = "lxstat";
#endif
#if defined (SYS_fxstat)
syscall_table[SYS_fxstat] = "fxstat";
#endif
#if defined (SYS_xmknod)
syscall_table[SYS_xmknod] = "xmknod";
#endif
#if defined (SYS_clocal)
syscall_table[SYS_clocal] = "clocal";
#endif
#if defined (SYS_setrlimit)
syscall_table[SYS_setrlimit] = "setrlimit";
#endif
#if defined (SYS_getrlimit)
syscall_table[SYS_getrlimit] = "getrlimit";
#endif
#if defined (SYS_lchown)
syscall_table[SYS_lchown] = "lchown";
#endif
#if defined (SYS_memcntl)
syscall_table[SYS_memcntl] = "memcntl";
#endif
#if defined (SYS_getpmsg)
syscall_table[SYS_getpmsg] = "getpmsg";
#endif
#if defined (SYS_putpmsg)
syscall_table[SYS_putpmsg] = "putpmsg";
#endif
#if defined (SYS_rename)
syscall_table[SYS_rename] = "rename";
#endif
#if defined (SYS_uname)
syscall_table[SYS_uname] = "uname";
#endif
#if defined (SYS_setegid)
syscall_table[SYS_setegid] = "setegid";
#endif
#if defined (SYS_sysconfig)
syscall_table[SYS_sysconfig] = "sysconfig";
#endif
#if defined (SYS_adjtime)
syscall_table[SYS_adjtime] = "adjtime";
#endif
#if defined (SYS_systeminfo)
syscall_table[SYS_systeminfo] = "systeminfo";
#endif
#if defined (SYS_seteuid)
syscall_table[SYS_seteuid] = "seteuid";
#endif
#if defined (SYS_sproc)
syscall_table[SYS_sproc] = "sproc";
#endif
}
/*
GLOBAL FUNCTION
ptrace -- override library version to force errors for /proc version
SYNOPSIS
int ptrace (int request, int pid, PTRACE_ARG3_TYPE arg3, int arg4)
DESCRIPTION
When gdb is configured to use /proc, it should not be calling
or otherwise attempting to use ptrace. In order to catch errors
where use of /proc is configured, but some routine is still calling
ptrace, we provide a local version of a function with that name
that does nothing but issue an error message.
*/
int
ptrace (request, pid, arg3, arg4)
int request;
int pid;
PTRACE_ARG3_TYPE arg3;
int arg4;
{
error ("internal error - there is a call to ptrace() somewhere");
/*NOTREACHED*/
}
/*
LOCAL FUNCTION
procfs_kill_inferior - kill any currently inferior
SYNOPSIS
void procfs_kill_inferior (void)
DESCRIPTION
Kill any current inferior.
NOTES
Kills even attached inferiors. Presumably the user has already
been prompted that the inferior is an attached one rather than
one started by gdb. (FIXME?)
*/
static void
procfs_kill_inferior ()
{
target_mourn_inferior ();
}
/*
LOCAL FUNCTION
unconditionally_kill_inferior - terminate the inferior
SYNOPSIS
static void unconditionally_kill_inferior (struct procinfo *)
DESCRIPTION
Kill the specified inferior.
NOTE
A possibly useful enhancement would be to first try sending
the inferior a terminate signal, politely asking it to commit
suicide, before we murder it (we could call that
politely_kill_inferior()).
*/
static void
unconditionally_kill_inferior (pi)
struct procinfo *pi;
{
int signo;
int ppid;
ppid = pi->prstatus.pr_ppid;
signo = SIGKILL;
ioctl (pi->fd, PIOCKILL, &signo);
close_proc_file (pi);
/* Only wait() for our direct children. Our grandchildren zombies are killed
by the death of their parents. */
if (ppid == getpid())
wait ((int *) 0);
}
/*
LOCAL FUNCTION
procfs_xfer_memory -- copy data to or from inferior memory space
SYNOPSIS
int procfs_xfer_memory (CORE_ADDR memaddr, char *myaddr, int len,
int dowrite, struct target_ops target)
DESCRIPTION
Copy LEN bytes to/from inferior's memory starting at MEMADDR
from/to debugger memory starting at MYADDR. Copy from inferior
if DOWRITE is zero or to inferior if DOWRITE is nonzero.
Returns the length copied, which is either the LEN argument or
zero. This xfer function does not do partial moves, since procfs_ops
doesn't allow memory operations to cross below us in the target stack
anyway.
NOTES
The /proc interface makes this an almost trivial task.
*/
static int
procfs_xfer_memory (memaddr, myaddr, len, dowrite, target)
CORE_ADDR memaddr;
char *myaddr;
int len;
int dowrite;
struct target_ops *target; /* ignored */
{
int nbytes = 0;
struct procinfo *pi;
pi = current_procinfo;
if (lseek(pi->fd, (off_t) memaddr, 0) == (off_t) memaddr)
{
if (dowrite)
{
nbytes = write (pi->fd, myaddr, len);
}
else
{
nbytes = read (pi->fd, myaddr, len);
}
if (nbytes < 0)
{
nbytes = 0;
}
}
return (nbytes);
}
/*
LOCAL FUNCTION
procfs_store_registers -- copy register values back to inferior
SYNOPSIS
void procfs_store_registers (int regno)
DESCRIPTION
Store our current register values back into the inferior. If
REGNO is -1 then store all the register, otherwise store just
the value specified by REGNO.
NOTES
If we are storing only a single register, we first have to get all
the current values from the process, overwrite the desired register
in the gregset with the one we want from gdb's registers, and then
send the whole set back to the process. For writing all the
registers, all we have to do is generate the gregset and send it to
the process.
Also note that the process has to be stopped on an event of interest
for this to work, which basically means that it has to have been
run under the control of one of the other /proc ioctl calls and not
ptrace. Since we don't use ptrace anyway, we don't worry about this
fine point, but it is worth noting for future reference.
Gdb is confused about what this function is supposed to return.
Some versions return a value, others return nothing. Some are
declared to return a value and actually return nothing. Gdb ignores
anything returned. (FIXME)
*/
static void
procfs_store_registers (regno)
int regno;
{
struct procinfo *pi;
pi = current_procinfo;
if (regno != -1)
{
ioctl (pi->fd, PIOCGREG, &pi->gregset);
}
fill_gregset (&pi->gregset, regno);
ioctl (pi->fd, PIOCSREG, &pi->gregset);
#if defined (FP0_REGNUM)
/* Now repeat everything using the floating point register set, if the
target has floating point hardware. Since we ignore the returned value,
we'll never know whether it worked or not anyway. */
if (regno != -1)
{
ioctl (pi->fd, PIOCGFPREG, &pi->fpregset);
}
fill_fpregset (&pi->fpregset, regno);
ioctl (pi->fd, PIOCSFPREG, &pi->fpregset);
#endif /* FP0_REGNUM */
}
/*
LOCAL FUNCTION
create_procinfo - initialize access to a /proc entry
SYNOPSIS
void create_procinfo (int pid)
DESCRIPTION
Allocate a procinfo structure, open the /proc file and then sets up
the set of signals and faults that are to be traced.
NOTES
If proc_init_failed ever gets called, control returns to the command
processing loop via the standard error handling code.
*/
static void
create_procinfo (pid)
int pid;
{
struct procinfo *pi;
if (find_procinfo (pid, 1))
return; /* All done! It already exists */
pi = (struct procinfo *) xmalloc (sizeof (struct procinfo));
if (!open_proc_file (pid, pi, O_RDWR))
proc_init_failed (pi, "can't open process file");
/* Add new process to process info list */
pi->next = procinfo_list;
procinfo_list = pi;
add_fd (pi); /* Add to list for poll/select */
memset ((char *) &pi->prrun, 0, sizeof (pi->prrun));
prfillset (&pi->prrun.pr_trace);
procfs_notice_signals (pid);
prfillset (&pi->prrun.pr_fault);
prdelset (&pi->prrun.pr_fault, FLTPAGE);
if (ioctl (pi->fd, PIOCWSTOP, &pi->prstatus) < 0)
proc_init_failed (pi, "PIOCWSTOP failed");
if (ioctl (pi->fd, PIOCSFAULT, &pi->prrun.pr_fault) < 0)
proc_init_failed (pi, "PIOCSFAULT failed");
}
/*
LOCAL FUNCTION
procfs_init_inferior - initialize target vector and access to a
/proc entry
SYNOPSIS
void procfs_init_inferior (int pid)
DESCRIPTION
When gdb starts an inferior, this function is called in the parent
process immediately after the fork. It waits for the child to stop
on the return from the exec system call (the child itself takes care
of ensuring that this is set up), then sets up the set of signals
and faults that are to be traced.
NOTES
If proc_init_failed ever gets called, control returns to the command
processing loop via the standard error handling code.
*/
static void
procfs_init_inferior (pid)
int pid;
{
push_target (&procfs_ops);
create_procinfo (pid);
add_thread (pid); /* Setup initial thread */
}
/*
GLOBAL FUNCTION
procfs_notice_signals
SYNOPSIS
static void procfs_notice_signals (int pid);
DESCRIPTION
When the user changes the state of gdb's signal handling via the
"handle" command, this function gets called to see if any change
in the /proc interface is required. It is also called internally
by other /proc interface functions to initialize the state of
the traced signal set.
One thing it does is that signals for which the state is "nostop",
"noprint", and "pass", have their trace bits reset in the pr_trace
field, so that they are no longer traced. This allows them to be
delivered directly to the inferior without the debugger ever being
involved.
*/
static void
procfs_notice_signals (pid)
int pid;
{
int signo;
struct procinfo *pi;
pi = find_procinfo (pid, 0);
for (signo = 0; signo < NSIG; signo++)
{
if (signal_stop_state (signo) == 0 &&
signal_print_state (signo) == 0 &&
signal_pass_state (signo) == 1)
{
prdelset (&pi->prrun.pr_trace, signo);
}
else
{
praddset (&pi->prrun.pr_trace, signo);
}
}
if (ioctl (pi->fd, PIOCSTRACE, &pi->prrun.pr_trace))
{
print_sys_errmsg ("PIOCSTRACE failed", errno);
}
}
/*
LOCAL FUNCTION
proc_set_exec_trap -- arrange for exec'd child to halt at startup
SYNOPSIS
void proc_set_exec_trap (void)
DESCRIPTION
This function is called in the child process when starting up
an inferior, prior to doing the exec of the actual inferior.
It sets the child process's exitset to make exit from the exec
system call an event of interest to stop on, and then simply
returns. The child does the exec, the system call returns, and
the child stops at the first instruction, ready for the gdb
parent process to take control of it.
NOTE
We need to use all local variables since the child may be sharing
it's data space with the parent, if vfork was used rather than
fork.
Also note that we want to turn off the inherit-on-fork flag in
the child process so that any grand-children start with all
tracing flags cleared.
*/
static void
proc_set_exec_trap ()
{
sysset_t exitset;
sysset_t entryset;
auto char procname[32];
int fd;
sprintf (procname, PROC_NAME_FMT, getpid ());
if ((fd = open (procname, O_RDWR)) < 0)
{
perror (procname);
fflush (stderr);
_exit (127);
}
premptyset (&exitset);
premptyset (&entryset);
/* GW: Rationale...
Not all systems with /proc have all the exec* syscalls with the same
names. On the SGI, for example, there is no SYS_exec, but there
*is* a SYS_execv. So, we try to account for that. */
#ifdef SYS_exec
praddset (&exitset, SYS_exec);
#endif
#ifdef SYS_execve
praddset (&exitset, SYS_execve);
#endif
#ifdef SYS_execv
praddset (&exitset, SYS_execv);
#endif
if (ioctl (fd, PIOCSEXIT, &exitset) < 0)
{
perror (procname);
fflush (stderr);
_exit (127);
}
praddset (&entryset, SYS_exit);
if (ioctl (fd, PIOCSENTRY, &entryset) < 0)
{
perror (procname);
fflush (stderr);
_exit (126);
}
/* Turn off inherit-on-fork flag so that all grand-children of gdb
start with tracing flags cleared. */
#if defined (PIOCRESET) /* New method */
{
long pr_flags;
pr_flags = PR_FORK;
ioctl (fd, PIOCRESET, &pr_flags);
}
#else
#if defined (PIOCRFORK) /* Original method */
ioctl (fd, PIOCRFORK, NULL);
#endif
#endif
/* Turn on run-on-last-close flag so that this process will not hang
if GDB goes away for some reason. */
#if defined (PIOCSET) /* New method */
{
long pr_flags;
pr_flags = PR_RLC;
(void) ioctl (fd, PIOCSET, &pr_flags);
}
#else
#if defined (PIOCSRLC) /* Original method */
(void) ioctl (fd, PIOCSRLC, 0);
#endif
#endif
}
/*
GLOBAL FUNCTION
proc_iterate_over_mappings -- call function for every mapped space
SYNOPSIS
int proc_iterate_over_mappings (int (*func)())
DESCRIPTION
Given a pointer to a function, call that function for every
mapped address space, passing it an open file descriptor for
the file corresponding to that mapped address space (if any)
and the base address of the mapped space. Quit when we hit
the end of the mappings or the function returns nonzero.
*/
int
proc_iterate_over_mappings (func)
int (*func) PARAMS ((int, CORE_ADDR));
{
int nmap;
int fd;
int funcstat = 0;
struct prmap *prmaps;
struct prmap *prmap;
struct procinfo *pi;
pi = current_procinfo;
if (ioctl (pi->fd, PIOCNMAP, &nmap) == 0)
{
prmaps = (struct prmap *) alloca ((nmap + 1) * sizeof (*prmaps));
if (ioctl (pi->fd, PIOCMAP, prmaps) == 0)
{
for (prmap = prmaps; prmap -> pr_size && funcstat == 0; ++prmap)
{
fd = proc_address_to_fd (pi, (CORE_ADDR) prmap -> pr_vaddr, 0);
funcstat = (*func) (fd, (CORE_ADDR) prmap -> pr_vaddr);
close (fd);
}
}
}
return (funcstat);
}
#if 0 /* Currently unused */
/*
GLOBAL FUNCTION
proc_base_address -- find base address for segment containing address
SYNOPSIS
CORE_ADDR proc_base_address (CORE_ADDR addr)
DESCRIPTION
Given an address of a location in the inferior, find and return
the base address of the mapped segment containing that address.
This is used for example, by the shared library support code,
where we have the pc value for some location in the shared library
where we are stopped, and need to know the base address of the
segment containing that address.
*/
CORE_ADDR
proc_base_address (addr)
CORE_ADDR addr;
{
int nmap;
struct prmap *prmaps;
struct prmap *prmap;
CORE_ADDR baseaddr = 0;
struct procinfo *pi;
pi = current_procinfo;
if (ioctl (pi->fd, PIOCNMAP, &nmap) == 0)
{
prmaps = (struct prmap *) alloca ((nmap + 1) * sizeof (*prmaps));
if (ioctl (pi->fd, PIOCMAP, prmaps) == 0)
{
for (prmap = prmaps; prmap -> pr_size; ++prmap)
{
if ((prmap -> pr_vaddr <= (caddr_t) addr) &&
(prmap -> pr_vaddr + prmap -> pr_size > (caddr_t) addr))
{
baseaddr = (CORE_ADDR) prmap -> pr_vaddr;
break;
}
}
}
}
return (baseaddr);
}
#endif /* 0 */
/*
LOCAL FUNCTION
proc_address_to_fd -- return open fd for file mapped to address
SYNOPSIS
int proc_address_to_fd (struct procinfo *pi, CORE_ADDR addr, complain)
DESCRIPTION
Given an address in the current inferior's address space, use the
/proc interface to find an open file descriptor for the file that
this address was mapped in from. Return -1 if there is no current
inferior. Print a warning message if there is an inferior but
the address corresponds to no file (IE a bogus address).
*/
static int
proc_address_to_fd (pi, addr, complain)
struct procinfo *pi;
CORE_ADDR addr;
int complain;
{
int fd = -1;
if ((fd = ioctl (pi->fd, PIOCOPENM, (caddr_t *) &addr)) < 0)
{
if (complain)
{
print_sys_errmsg (pi->pathname, errno);
warning ("can't find mapped file for address 0x%x", addr);
}
}
return (fd);
}
/* Attach to process PID, then initialize for debugging it
and wait for the trace-trap that results from attaching. */
static void
procfs_attach (args, from_tty)
char *args;
int from_tty;
{
char *exec_file;
int pid;
if (!args)
error_no_arg ("process-id to attach");
pid = atoi (args);
if (pid == getpid()) /* Trying to masturbate? */
error ("I refuse to debug myself!");
if (from_tty)
{
exec_file = (char *) get_exec_file (0);
if (exec_file)
printf ("Attaching to program `%s', %s\n", exec_file, target_pid_to_str (pid));
else
printf ("Attaching to %s\n", target_pid_to_str (pid));
fflush (stdout);
}
do_attach (pid);
inferior_pid = pid;
push_target (&procfs_ops);
}
/* Take a program previously attached to and detaches it.
The program resumes execution and will no longer stop
on signals, etc. We'd better not have left any breakpoints
in the program or it'll die when it hits one. For this
to work, it may be necessary for the process to have been
previously attached. It *might* work if the program was
started via the normal ptrace (PTRACE_TRACEME). */
static void
procfs_detach (args, from_tty)
char *args;
int from_tty;
{
int siggnal = 0;
if (from_tty)
{
char *exec_file = get_exec_file (0);
if (exec_file == 0)
exec_file = "";
printf ("Detaching from program: %s %s\n",
exec_file, target_pid_to_str (inferior_pid));
fflush (stdout);
}
if (args)
siggnal = atoi (args);
do_detach (siggnal);
inferior_pid = 0;
unpush_target (&procfs_ops); /* Pop out of handling an inferior */
}
/* Get ready to modify the registers array. On machines which store
individual registers, this doesn't need to do anything. On machines
which store all the registers in one fell swoop, this makes sure
that registers contains all the registers from the program being
debugged. */
static void
procfs_prepare_to_store ()
{
#ifdef CHILD_PREPARE_TO_STORE
CHILD_PREPARE_TO_STORE ();
#endif
}
/* Print status information about what we're accessing. */
static void
procfs_files_info (ignore)
struct target_ops *ignore;
{
printf ("\tUsing the running image of %s %s via /proc.\n",
attach_flag? "attached": "child", target_pid_to_str (inferior_pid));
}
/* ARGSUSED */
static void
procfs_open (arg, from_tty)
char *arg;
int from_tty;
{
error ("Use the \"run\" command to start a Unix child process.");
}
/*
LOCAL FUNCTION
do_attach -- attach to an already existing process
SYNOPSIS
int do_attach (int pid)
DESCRIPTION
Attach to an already existing process with the specified process
id. If the process is not already stopped, query whether to
stop it or not.
NOTES
The option of stopping at attach time is specific to the /proc
versions of gdb. Versions using ptrace force the attachee
to stop. (I have changed this version to do so, too. All you
have to do is "continue" to make it go on. -- gnu@cygnus.com)
*/
static int
do_attach (pid)
int pid;
{
int result;
struct procinfo *pi;
pi = (struct procinfo *) xmalloc (sizeof (struct procinfo));
if (!open_proc_file (pid, pi, O_RDWR))
{
free (pi);
perror_with_name (pi->pathname);
/* NOTREACHED */
}
/* Add new process to process info list */
pi->next = procinfo_list;
procinfo_list = pi;
add_fd (pi); /* Add to list for poll/select */
/* Get current status of process and if it is not already stopped,
then stop it. Remember whether or not it was stopped when we first
examined it. */
if (ioctl (pi->fd, PIOCSTATUS, &pi->prstatus) < 0)
{
print_sys_errmsg (pi->pathname, errno);
close_proc_file (pi);
error ("PIOCSTATUS failed");
}
if (pi->prstatus.pr_flags & (PR_STOPPED | PR_ISTOP))
{
pi->was_stopped = 1;
}
else
{
pi->was_stopped = 0;
if (1 || query ("Process is currently running, stop it? "))
{
/* Make it run again when we close it. */
#if defined (PIOCSET) /* New method */
{
long pr_flags;
pr_flags = PR_RLC;
result = ioctl (pi->fd, PIOCSET, &pr_flags);
}
#else
#if defined (PIOCSRLC) /* Original method */
result = ioctl (pi->fd, PIOCSRLC, 0);
#endif
#endif
if (result < 0)
{
print_sys_errmsg (pi->pathname, errno);
close_proc_file (pi);
error ("PIOCSRLC or PIOCSET failed");
}
if (ioctl (pi->fd, PIOCSTOP, &pi->prstatus) < 0)
{
print_sys_errmsg (pi->pathname, errno);
close_proc_file (pi);
error ("PIOCSTOP failed");
}
pi->nopass_next_sigstop = 1;
}
else
{
printf ("Ok, gdb will wait for %s to stop.\n", target_pid_to_str (pid));
}
}
/* Remember some things about the inferior that we will, or might, change
so that we can restore them when we detach. */
ioctl (pi->fd, PIOCGTRACE, &pi->saved_trace);
ioctl (pi->fd, PIOCGHOLD, &pi->saved_sighold);
ioctl (pi->fd, PIOCGFAULT, &pi->saved_fltset);
ioctl (pi->fd, PIOCGENTRY, &pi->saved_entryset);
ioctl (pi->fd, PIOCGEXIT, &pi->saved_exitset);
/* Set up trace and fault sets, as gdb expects them. */
memset (&pi->prrun, 0, sizeof (pi->prrun));
prfillset (&pi->prrun.pr_trace);
procfs_notice_signals (pid);
prfillset (&pi->prrun.pr_fault);
prdelset (&pi->prrun.pr_fault, FLTPAGE);
if (ioctl (pi->fd, PIOCSFAULT, &pi->prrun.pr_fault))
{
print_sys_errmsg ("PIOCSFAULT failed", errno);
}
if (ioctl (pi->fd, PIOCSTRACE, &pi->prrun.pr_trace))
{
print_sys_errmsg ("PIOCSTRACE failed", errno);
}
attach_flag = 1;
return (pid);
}
/*
LOCAL FUNCTION
do_detach -- detach from an attached-to process
SYNOPSIS
void do_detach (int signal)
DESCRIPTION
Detach from the current attachee.
If signal is non-zero, the attachee is started running again and sent
the specified signal.
If signal is zero and the attachee was not already stopped when we
attached to it, then we make it runnable again when we detach.
Otherwise, we query whether or not to make the attachee runnable
again, since we may simply want to leave it in the state it was in
when we attached.
We report any problems, but do not consider them errors, since we
MUST detach even if some things don't seem to go right. This may not
be the ideal situation. (FIXME).
*/
static void
do_detach (signal)
int signal;
{
int result;
struct procinfo *pi;
pi = current_procinfo;
if (signal)
{
set_proc_siginfo (pi, signal);
}
if (ioctl (pi->fd, PIOCSEXIT, &pi->saved_exitset) < 0)
{
print_sys_errmsg (pi->pathname, errno);
printf ("PIOCSEXIT failed.\n");
}
if (ioctl (pi->fd, PIOCSENTRY, &pi->saved_entryset) < 0)
{
print_sys_errmsg (pi->pathname, errno);
printf ("PIOCSENTRY failed.\n");
}
if (ioctl (pi->fd, PIOCSTRACE, &pi->saved_trace) < 0)
{
print_sys_errmsg (pi->pathname, errno);
printf ("PIOCSTRACE failed.\n");
}
if (ioctl (pi->fd, PIOCSHOLD, &pi->saved_sighold) < 0)
{
print_sys_errmsg (pi->pathname, errno);
printf ("PIOSCHOLD failed.\n");
}
if (ioctl (pi->fd, PIOCSFAULT, &pi->saved_fltset) < 0)
{
print_sys_errmsg (pi->pathname, errno);
printf ("PIOCSFAULT failed.\n");
}
if (ioctl (pi->fd, PIOCSTATUS, &pi->prstatus) < 0)
{
print_sys_errmsg (pi->pathname, errno);
printf ("PIOCSTATUS failed.\n");
}
else
{
if (signal || (pi->prstatus.pr_flags & (PR_STOPPED | PR_ISTOP)))
{
if (signal || !pi->was_stopped ||
query ("Was stopped when attached, make it runnable again? "))
{
/* Clear any fault that might have stopped it. */
if (ioctl (pi->fd, PIOCCFAULT, 0))
{
print_sys_errmsg (pi->pathname, errno);
printf ("PIOCCFAULT failed.\n");
}
/* Make it run again when we close it. */
#if defined (PIOCSET) /* New method */
{
long pr_flags;
pr_flags = PR_RLC;
result = ioctl (pi->fd, PIOCSET, &pr_flags);
}
#else
#if defined (PIOCSRLC) /* Original method */
result = ioctl (pi->fd, PIOCSRLC, 0);
#endif
#endif
if (result)
{
print_sys_errmsg (pi->pathname, errno);
printf ("PIOCSRLC or PIOCSET failed.\n");
}
}
}
}
close_proc_file (pi);
attach_flag = 0;
}
/*
LOCAL FUNCTION
procfs_wait -- emulate wait() as much as possible
Wait for child to do something. Return pid of child, or -1 in case
of error; store status through argument pointer STATUS.
SYNOPSIS
int procfs_wait (int pid, int *statloc)
DESCRIPTION
Try to emulate wait() as much as possible. Not sure why we can't
just use wait(), but it seems to have problems when applied to a
process being controlled with the /proc interface.
NOTES
We have a race problem here with no obvious solution. We need to let
the inferior run until it stops on an event of interest, which means
that we need to use the PIOCWSTOP ioctl. However, we cannot use this
ioctl if the process is already stopped on something that is not an
event of interest, or the call will hang indefinitely. Thus we first
use PIOCSTATUS to see if the process is not stopped. If not, then we
use PIOCWSTOP. But during the window between the two, if the process
stops for any reason that is not an event of interest (such as a job
control signal) then gdb will hang. One possible workaround is to set
an alarm to wake up every minute of so and check to see if the process
is still running, and if so, then reissue the PIOCWSTOP. But this is
a real kludge, so has not been implemented. FIXME: investigate
alternatives.
FIXME: Investigate why wait() seems to have problems with programs
being control by /proc routines.
*/
static int
procfs_wait (pid, statloc)
int pid;
int *statloc;
{
short what;
short why;
int statval = 0;
int checkerr = 0;
int rtnval = -1;
struct procinfo *pi;
if (pid != -1) /* Non-specific process? */
pi = NULL;
else
for (pi = procinfo_list; pi; pi = pi->next)
if (pi->had_event)
break;
wait_again:
if (!pi)
pi = wait_fd ();
if (pid != -1)
for (pi = procinfo_list; pi; pi = pi->next)
if (pi->pid == pid && pi->had_event)
break;
if (!pi && !checkerr)
goto wait_again;
if (!checkerr && !(pi->prstatus.pr_flags & (PR_STOPPED | PR_ISTOP)))
{
if (ioctl (pi->fd, PIOCWSTOP, &pi->prstatus) < 0)
{
checkerr++;
}
}
if (checkerr)
{
if (errno == ENOENT)
{
rtnval = wait (&statval);
if (rtnval != inferior_pid)
{
print_sys_errmsg (pi->pathname, errno);
error ("PIOCWSTOP, wait failed, returned %d", rtnval);
/* NOTREACHED */
}
}
else
{
print_sys_errmsg (pi->pathname, errno);
error ("PIOCSTATUS or PIOCWSTOP failed.");
/* NOTREACHED */
}
}
else if (pi->prstatus.pr_flags & (PR_STOPPED | PR_ISTOP))
{
rtnval = pi->prstatus.pr_pid;
why = pi->prstatus.pr_why;
what = pi->prstatus.pr_what;
switch (why)
{
case PR_SIGNALLED:
statval = (what << 8) | 0177;
break;
case PR_SYSENTRY:
if (what != SYS_exit)
error ("PR_SYSENTRY, unknown system call %d", what);
pi->prrun.pr_flags = PRCFAULT;
if (ioctl (pi->fd, PIOCRUN, &pi->prrun) != 0)
perror_with_name (pi->pathname);
rtnval = wait (&statval);
break;
case PR_SYSEXIT:
switch (what)
{
#ifdef SYS_exec
case SYS_exec:
#endif
#ifdef SYS_execve
case SYS_execve:
#endif
#ifdef SYS_execv
case SYS_execv:
#endif
statval = (SIGTRAP << 8) | 0177;
break;
#ifdef SYS_sproc
case SYS_sproc:
/* We've just detected the completion of an sproc system call. Now we need to
setup a procinfo struct for this thread, and notify the thread system of the
new arrival. */
/* If sproc failed, then nothing interesting happened. Continue the process and
go back to sleep. */
if (pi->prstatus.pr_errno != 0)
{
pi->prrun.pr_flags &= PRSTEP;
pi->prrun.pr_flags |= PRCFAULT;
if (ioctl (pi->fd, PIOCRUN, &pi->prrun) != 0)
perror_with_name (pi->pathname);
goto wait_again;
}
/* At this point, the new thread is stopped at it's first instruction, and
the parent is stopped at the exit from sproc. */
/* Notify the caller of the arrival of a new thread. */
create_procinfo (pi->prstatus.pr_rval1);
rtnval = pi->prstatus.pr_rval1;
statval = (SIGTRAP << 8) | 0177;
break;
#endif /* SYS_sproc */
default:
error ("PIOCSTATUS (PR_SYSEXIT): Unknown system call %d", what);
}
break;
case PR_REQUESTED:
statval = (SIGSTOP << 8) | 0177;
break;
case PR_JOBCONTROL:
statval = (what << 8) | 0177;
break;
case PR_FAULTED:
switch (what)
{
case FLTPRIV:
case FLTILL:
statval = (SIGILL << 8) | 0177;
break;
case FLTBPT:
case FLTTRACE:
statval = (SIGTRAP << 8) | 0177;
break;
case FLTSTACK:
case FLTACCESS:
case FLTBOUNDS:
statval = (SIGSEGV << 8) | 0177;
break;
case FLTIOVF:
case FLTIZDIV:
case FLTFPE:
statval = (SIGFPE << 8) | 0177;
break;
case FLTPAGE: /* Recoverable page fault */
default:
error ("PIOCWSTOP, unknown why %d, what %d", why, what);
}
break;
default:
error ("PIOCWSTOP, unknown why %d, what %d", why, what);
}
/* Stop all the other threads when any of them stops. */
{
struct procinfo *procinfo;
for (procinfo = procinfo_list; procinfo; procinfo = procinfo->next)
{
if (!procinfo->had_event)
if (ioctl (procinfo->fd, PIOCSTOP, &procinfo->prstatus) < 0)
{
print_sys_errmsg (procinfo->pathname, errno);
error ("PIOCSTOP failed");
}
}
}
}
else
{
error ("PIOCWSTOP, stopped for unknown/unhandled reason, flags %#x",
pi->prstatus.pr_flags);
}
if (statloc)
{
*statloc = statval;
}
if (rtnval == -1) /* No more children to wait for */
{
fprintf (stderr, "Child process unexpectedly missing.\n");
*statloc = 42; /* Claim it exited with signal 42 */
return rtnval;
}
pi->had_event = 0; /* Indicate that we've seen this one */
return (rtnval);
}
/*
LOCAL FUNCTION
set_proc_siginfo - set a process's current signal info
SYNOPSIS
void set_proc_siginfo (struct procinfo *pip, int signo);
DESCRIPTION
Given a pointer to a process info struct in PIP and a signal number
in SIGNO, set the process's current signal and its associated signal
information. The signal will be delivered to the process immediately
after execution is resumed, even if it is being held. In addition,
this particular delivery will not cause another PR_SIGNALLED stop
even if the signal is being traced.
If we are not delivering the same signal that the prstatus siginfo
struct contains information about, then synthesize a siginfo struct
to match the signal we are doing to deliver, make it of the type
"generated by a user process", and send this synthesized copy. When
used to set the inferior's signal state, this will be required if we
are not currently stopped because of a traced signal, or if we decide
to continue with a different signal.
Note that when continuing the inferior from a stop due to receipt
of a traced signal, we either have set PRCSIG to clear the existing
signal, or we have to call this function to do a PIOCSSIG with either
the existing siginfo struct from pr_info, or one we have synthesized
appropriately for the signal we want to deliver. Otherwise if the
signal is still being traced, the inferior will immediately stop
again.
See siginfo(5) for more details.
*/
static void
set_proc_siginfo (pip, signo)
struct procinfo *pip;
int signo;
{
struct siginfo newsiginfo;
struct siginfo *sip;
if (signo == pip -> prstatus.pr_info.si_signo)
{
sip = &pip -> prstatus.pr_info;
}
else
{
memset ((char *) &newsiginfo, 0, sizeof (newsiginfo));
sip = &newsiginfo;
sip -> si_signo = signo;
sip -> si_code = 0;
sip -> si_errno = 0;
sip -> si_pid = getpid ();
sip -> si_uid = getuid ();
}
if (ioctl (pip -> fd, PIOCSSIG, sip) < 0)
{
print_sys_errmsg (pip -> pathname, errno);
warning ("PIOCSSIG failed");
}
}
/* Resume execution of process PID. If STEP is nozero, then
just single step it. If SIGNAL is nonzero, restart it with that
signal activated. */
static void
procfs_resume (pid, step, signo)
int pid;
int step;
int signo;
{
int signal_to_pass;
struct procinfo *pi, *procinfo;
pi = find_procinfo (pid == -1 ? inferior_pid : pid, 0);
errno = 0;
pi->prrun.pr_flags = PRSTRACE | PRSFAULT | PRCFAULT;
#if 0
/* It should not be necessary. If the user explicitly changes the value,
value_assign calls write_register_bytes, which writes it. */
/* It may not be absolutely necessary to specify the PC value for
restarting, but to be safe we use the value that gdb considers
to be current. One case where this might be necessary is if the
user explicitly changes the PC value that gdb considers to be
current. FIXME: Investigate if this is necessary or not. */
#ifdef PRSVADDR_BROKEN
/* Can't do this under Solaris running on a Sparc, as there seems to be no
place to put nPC. In fact, if you use this, nPC seems to be set to some
random garbage. We have to rely on the fact that PC and nPC have been
written previously via PIOCSREG during a register flush. */
pi->prrun.pr_vaddr = (caddr_t) *(int *) &registers[REGISTER_BYTE (PC_REGNUM)];
pi->prrun.pr_flags != PRSVADDR;
#endif
#endif
if (signo == SIGSTOP && pi->nopass_next_sigstop)
/* When attaching to a child process, if we forced it to stop with
a PIOCSTOP, then we will have set the nopass_next_sigstop flag.
Upon resuming the first time after such a stop, we explicitly
inhibit sending it another SIGSTOP, which would be the normal
result of default signal handling. One potential drawback to
this is that we will also ignore any attempt to by the user
to explicitly continue after the attach with a SIGSTOP. Ultimately
this problem should be dealt with by making the routines that
deal with the inferior a little smarter, and possibly even allow
an inferior to continue running at the same time as gdb. (FIXME?) */
signal_to_pass = 0;
else if (signo == SIGTSTP
&& pi->prstatus.pr_cursig == SIGTSTP
&& pi->prstatus.pr_action.sa_handler == SIG_DFL)
/* We are about to pass the inferior a SIGTSTP whose action is
SIG_DFL. The SIG_DFL action for a SIGTSTP is to stop
(notifying the parent via wait()), and then keep going from the
same place when the parent is ready for you to keep going. So
under the debugger, it should do nothing (as if the program had
been stopped and then later resumed. Under ptrace, this
happens for us, but under /proc, the system obligingly stops
the process, and wait_for_inferior would have no way of
distinguishing that type of stop (which indicates that we
should just start it again), with a stop due to the pr_trace
field of the prrun_t struct.
Note that if the SIGTSTP is being caught, we *do* need to pass it,
because the handler needs to get executed. */
signal_to_pass = 0;
else
signal_to_pass = signo;
if (signal_to_pass)
{
set_proc_siginfo (pi, signal_to_pass);
}
else
{
pi->prrun.pr_flags |= PRCSIG;
}
pi->nopass_next_sigstop = 0;
if (step)
{
pi->prrun.pr_flags |= PRSTEP;
}
if (ioctl (pi->fd, PIOCRUN, &pi->prrun) != 0)
{
perror_with_name (pi->pathname);
/* NOTREACHED */
}
pi->had_event = 0;
/* Continue all the other threads that haven't had an event of
interest. */
if (pid == -1)
for (procinfo = procinfo_list; procinfo; procinfo = procinfo->next)
{
if (pi != procinfo && !procinfo->had_event)
{
procinfo->prrun.pr_flags &= PRSTEP;
procinfo->prrun.pr_flags |= PRCFAULT | PRCSIG;
ioctl (procinfo->fd, PIOCSTATUS, &procinfo->prstatus);
if (ioctl (procinfo->fd, PIOCRUN, &procinfo->prrun) < 0)
{
if (ioctl (procinfo->fd, PIOCSTATUS, &procinfo->prstatus) < 0)
{
fprintf(stderr, "PIOCSTATUS failed, errno=%d\n", errno);
}
print_sys_errmsg (procinfo->pathname, errno);
error ("PIOCRUN failed");
}
ioctl (procinfo->fd, PIOCSTATUS, &procinfo->prstatus);
}
}
}
/*
LOCAL FUNCTION
procfs_fetch_registers -- fetch current registers from inferior
SYNOPSIS
void procfs_fetch_registers (int regno)
DESCRIPTION
Read the current values of the inferior's registers, both the
general register set and floating point registers (if supported)
and update gdb's idea of their current values.
*/
static void
procfs_fetch_registers (regno)
int regno;
{
struct procinfo *pi;
pi = current_procinfo;
if (ioctl (pi->fd, PIOCGREG, &pi->gregset) != -1)
{
supply_gregset (&pi->gregset);
}
#if defined (FP0_REGNUM)
if (ioctl (pi->fd, PIOCGFPREG, &pi->fpregset) != -1)
{
supply_fpregset (&pi->fpregset);
}
#endif
}
/*
LOCAL FUNCTION
proc_init_failed - called whenever /proc access initialization
fails
SYNOPSIS
static void proc_init_failed (struct procinfo *pi, char *why)
DESCRIPTION
This function is called whenever initialization of access to a /proc
entry fails. It prints a suitable error message, does some cleanup,
and then invokes the standard error processing routine which dumps
us back into the command loop.
*/
static void
proc_init_failed (pi, why)
struct procinfo *pi;
char *why;
{
print_sys_errmsg (pi->pathname, errno);
kill (pi->pid, SIGKILL);
close_proc_file (pi);
error (why);
/* NOTREACHED */
}
/*
LOCAL FUNCTION
close_proc_file - close any currently open /proc entry
SYNOPSIS
static void close_proc_file (struct procinfo *pip)
DESCRIPTION
Close any currently open /proc entry and mark the process information
entry as invalid. In order to ensure that we don't try to reuse any
stale information, the pid, fd, and pathnames are explicitly
invalidated, which may be overkill.
*/
static void
close_proc_file (pip)
struct procinfo *pip;
{
struct procinfo *procinfo;
remove_fd (pip); /* Remove fd from poll/select list */
close (pip -> fd);
free (pip -> pathname);
/* Unlink pip from the procinfo chain. Note pip might not be on the list. */
if (procinfo_list == pip)
procinfo_list = pip->next;
else
for (procinfo = procinfo_list; procinfo; procinfo = procinfo->next)
if (procinfo->next == pip)
procinfo->next = pip->next;
free (pip);
}
/*
LOCAL FUNCTION
open_proc_file - open a /proc entry for a given process id
SYNOPSIS
static int open_proc_file (int pid, struct procinfo *pip, int mode)
DESCRIPTION
Given a process id and a mode, close the existing open /proc
entry (if any) and open one for the new process id, in the
specified mode. Once it is open, then mark the local process
information structure as valid, which guarantees that the pid,
fd, and pathname fields match an open /proc entry. Returns
zero if the open fails, nonzero otherwise.
Note that the pathname is left intact, even when the open fails,
so that callers can use it to construct meaningful error messages
rather than just "file open failed".
*/
static int
open_proc_file (pid, pip, mode)
int pid;
struct procinfo *pip;
int mode;
{
pip -> next = NULL;
pip -> had_event = 0;
pip -> pathname = xmalloc (32);
pip -> pid = pid;
sprintf (pip -> pathname, PROC_NAME_FMT, pid);
if ((pip -> fd = open (pip -> pathname, mode)) < 0)
return 0;
return 1;
}
static char *
mappingflags (flags)
long flags;
{
static char asciiflags[8];
strcpy (asciiflags, "-------");
#if defined (MA_PHYS)
if (flags & MA_PHYS) asciiflags[0] = 'd';
#endif
if (flags & MA_STACK) asciiflags[1] = 's';
if (flags & MA_BREAK) asciiflags[2] = 'b';
if (flags & MA_SHARED) asciiflags[3] = 's';
if (flags & MA_READ) asciiflags[4] = 'r';
if (flags & MA_WRITE) asciiflags[5] = 'w';
if (flags & MA_EXEC) asciiflags[6] = 'x';
return (asciiflags);
}
static void
info_proc_flags (pip, summary)
struct procinfo *pip;
int summary;
{
struct trans *transp;
printf_filtered ("%-32s", "Process status flags:");
if (!summary)
{
printf_filtered ("\n\n");
}
for (transp = pr_flag_table; transp -> name != NULL; transp++)
{
if (pip -> prstatus.pr_flags & transp -> value)
{
if (summary)
{
printf_filtered ("%s ", transp -> name);
}
else
{
printf_filtered ("\t%-16s %s.\n", transp -> name, transp -> desc);
}
}
}
printf_filtered ("\n");
}
static void
info_proc_stop (pip, summary)
struct procinfo *pip;
int summary;
{
struct trans *transp;
int why;
int what;
why = pip -> prstatus.pr_why;
what = pip -> prstatus.pr_what;
if (pip -> prstatus.pr_flags & PR_STOPPED)
{
printf_filtered ("%-32s", "Reason for stopping:");
if (!summary)
{
printf_filtered ("\n\n");
}
for (transp = pr_why_table; transp -> name != NULL; transp++)
{
if (why == transp -> value)
{
if (summary)
{
printf_filtered ("%s ", transp -> name);
}
else
{
printf_filtered ("\t%-16s %s.\n",
transp -> name, transp -> desc);
}
break;
}
}
/* Use the pr_why field to determine what the pr_what field means, and
print more information. */
switch (why)
{
case PR_REQUESTED:
/* pr_what is unused for this case */
break;
case PR_JOBCONTROL:
case PR_SIGNALLED:
if (summary)
{
printf_filtered ("%s ", signalname (what));
}
else
{
printf_filtered ("\t%-16s %s.\n", signalname (what),
safe_strsignal (what));
}
break;
case PR_SYSENTRY:
if (summary)
{
printf_filtered ("%s ", syscallname (what));
}
else
{
printf_filtered ("\t%-16s %s.\n", syscallname (what),
"Entered this system call");
}
break;
case PR_SYSEXIT:
if (summary)
{
printf_filtered ("%s ", syscallname (what));
}
else
{
printf_filtered ("\t%-16s %s.\n", syscallname (what),
"Returned from this system call");
}
break;
case PR_FAULTED:
if (summary)
{
printf_filtered ("%s ",
lookupname (faults_table, what, "fault"));
}
else
{
printf_filtered ("\t%-16s %s.\n",
lookupname (faults_table, what, "fault"),
lookupdesc (faults_table, what));
}
break;
}
printf_filtered ("\n");
}
}
static void
info_proc_siginfo (pip, summary)
struct procinfo *pip;
int summary;
{
struct siginfo *sip;
if ((pip -> prstatus.pr_flags & PR_STOPPED) &&
(pip -> prstatus.pr_why == PR_SIGNALLED ||
pip -> prstatus.pr_why == PR_FAULTED))
{
printf_filtered ("%-32s", "Additional signal/fault info:");
sip = &pip -> prstatus.pr_info;
if (summary)
{
printf_filtered ("%s ", signalname (sip -> si_signo));
if (sip -> si_errno > 0)
{
printf_filtered ("%s ", errnoname (sip -> si_errno));
}
if (sip -> si_code <= 0)
{
printf_filtered ("sent by %s, uid %d ",
target_pid_to_str (sip -> si_pid),
sip -> si_uid);
}
else
{
printf_filtered ("%s ", sigcodename (sip));
if ((sip -> si_signo == SIGILL) ||
(sip -> si_signo == SIGFPE) ||
(sip -> si_signo == SIGSEGV) ||
(sip -> si_signo == SIGBUS))
{
printf_filtered ("addr=%#x ", sip -> si_addr);
}
else if ((sip -> si_signo == SIGCHLD))
{
printf_filtered ("child %s, status %u ",
target_pid_to_str (sip -> si_pid),
sip -> si_status);
}
else if ((sip -> si_signo == SIGPOLL))
{
printf_filtered ("band %u ", sip -> si_band);
}
}
}
else
{
printf_filtered ("\n\n");
printf_filtered ("\t%-16s %s.\n", signalname (sip -> si_signo),
safe_strsignal (sip -> si_signo));
if (sip -> si_errno > 0)
{
printf_filtered ("\t%-16s %s.\n",
errnoname (sip -> si_errno),
safe_strerror (sip -> si_errno));
}
if (sip -> si_code <= 0)
{
printf_filtered ("\t%-16u %s\n", sip -> si_pid, /* XXX need target_pid_to_str() */
"PID of process sending signal");
printf_filtered ("\t%-16u %s\n", sip -> si_uid,
"UID of process sending signal");
}
else
{
printf_filtered ("\t%-16s %s.\n", sigcodename (sip),
sigcodedesc (sip));
if ((sip -> si_signo == SIGILL) ||
(sip -> si_signo == SIGFPE))
{
printf_filtered ("\t%-16#x %s.\n", sip -> si_addr,
"Address of faulting instruction");
}
else if ((sip -> si_signo == SIGSEGV) ||
(sip -> si_signo == SIGBUS))
{
printf_filtered ("\t%-16#x %s.\n", sip -> si_addr,
"Address of faulting memory reference");
}
else if ((sip -> si_signo == SIGCHLD))
{
printf_filtered ("\t%-16u %s.\n", sip -> si_pid, /* XXX need target_pid_to_str() */
"Child process ID");
printf_filtered ("\t%-16u %s.\n", sip -> si_status,
"Child process exit value or signal");
}
else if ((sip -> si_signo == SIGPOLL))
{
printf_filtered ("\t%-16u %s.\n", sip -> si_band,
"Band event for POLL_{IN,OUT,MSG}");
}
}
}
printf_filtered ("\n");
}
}
static void
info_proc_syscalls (pip, summary)
struct procinfo *pip;
int summary;
{
int syscallnum;
if (!summary)
{
#if 0 /* FIXME: Needs to use gdb-wide configured info about system calls. */
if (pip -> prstatus.pr_flags & PR_ASLEEP)
{
int syscallnum = pip -> prstatus.pr_reg[R_D0];
if (summary)
{
printf_filtered ("%-32s", "Sleeping in system call:");
printf_filtered ("%s", syscallname (syscallnum));
}
else
{
printf_filtered ("Sleeping in system call '%s'.\n",
syscallname (syscallnum));
}
}
#endif
if (ioctl (pip -> fd, PIOCGENTRY, &pip -> entryset) < 0)
{
print_sys_errmsg (pip -> pathname, errno);
error ("PIOCGENTRY failed");
}
if (ioctl (pip -> fd, PIOCGEXIT, &pip -> exitset) < 0)
{
print_sys_errmsg (pip -> pathname, errno);
error ("PIOCGEXIT failed");
}
printf_filtered ("System call tracing information:\n\n");
printf_filtered ("\t%-12s %-8s %-8s\n",
"System call",
"Entry",
"Exit");
for (syscallnum = 0; syscallnum < MAX_SYSCALLS; syscallnum++)
{
QUIT;
if (syscall_table[syscallnum] != NULL)
{
printf_filtered ("\t%-12s ", syscall_table[syscallnum]);
printf_filtered ("%-8s ",
prismember (&pip -> entryset, syscallnum)
? "on" : "off");
printf_filtered ("%-8s ",
prismember (&pip -> exitset, syscallnum)
? "on" : "off");
printf_filtered ("\n");
}
}
printf_filtered ("\n");
}
}
static char *
signalname (signo)
int signo;
{
char *name;
static char locbuf[32];
name = strsigno (signo);
if (name == NULL)
{
sprintf (locbuf, "Signal %d", signo);
}
else
{
sprintf (locbuf, "%s (%d)", name, signo);
}
return (locbuf);
}
static char *
errnoname (errnum)
int errnum;
{
char *name;
static char locbuf[32];
name = strerrno (errnum);
if (name == NULL)
{
sprintf (locbuf, "Errno %d", errnum);
}
else
{
sprintf (locbuf, "%s (%d)", name, errnum);
}
return (locbuf);
}
static void
info_proc_signals (pip, summary)
struct procinfo *pip;
int summary;
{
int signo;
if (!summary)
{
if (ioctl (pip -> fd, PIOCGTRACE, &pip -> trace) < 0)
{
print_sys_errmsg (pip -> pathname, errno);
error ("PIOCGTRACE failed");
}
printf_filtered ("Disposition of signals:\n\n");
printf_filtered ("\t%-15s %-8s %-8s %-8s %s\n\n",
"Signal", "Trace", "Hold", "Pending", "Description");
for (signo = 0; signo < NSIG; signo++)
{
QUIT;
printf_filtered ("\t%-15s ", signalname (signo));
printf_filtered ("%-8s ",
prismember (&pip -> trace, signo)
? "on" : "off");
printf_filtered ("%-8s ",
prismember (&pip -> prstatus.pr_sighold, signo)
? "on" : "off");
printf_filtered ("%-8s ",
prismember (&pip -> prstatus.pr_sigpend, signo)
? "yes" : "no");
printf_filtered (" %s\n", safe_strsignal (signo));
}
printf_filtered ("\n");
}
}
static void
info_proc_faults (pip, summary)
struct procinfo *pip;
int summary;
{
struct trans *transp;
if (!summary)
{
if (ioctl (pip -> fd, PIOCGFAULT, &pip -> fltset) < 0)
{
print_sys_errmsg (pip -> pathname, errno);
error ("PIOCGFAULT failed");
}
printf_filtered ("Current traced hardware fault set:\n\n");
printf_filtered ("\t%-12s %-8s\n", "Fault", "Trace");
for (transp = faults_table; transp -> name != NULL; transp++)
{
QUIT;
printf_filtered ("\t%-12s ", transp -> name);
printf_filtered ("%-8s", prismember (&pip -> fltset, transp -> value)
? "on" : "off");
printf_filtered ("\n");
}
printf_filtered ("\n");
}
}
static void
info_proc_mappings (pip, summary)
struct procinfo *pip;
int summary;
{
int nmap;
struct prmap *prmaps;
struct prmap *prmap;
if (!summary)
{
printf_filtered ("Mapped address spaces:\n\n");
printf_filtered ("\t%10s %10s %10s %10s %7s\n",
"Start Addr",
" End Addr",
" Size",
" Offset",
"Flags");
if (ioctl (pip -> fd, PIOCNMAP, &nmap) == 0)
{
prmaps = (struct prmap *) alloca ((nmap + 1) * sizeof (*prmaps));
if (ioctl (pip -> fd, PIOCMAP, prmaps) == 0)
{
for (prmap = prmaps; prmap -> pr_size; ++prmap)
{
printf_filtered ("\t%#10x %#10x %#10x %#10x %7s\n",
prmap -> pr_vaddr,
prmap -> pr_vaddr + prmap -> pr_size - 1,
prmap -> pr_size,
prmap -> pr_off,
mappingflags (prmap -> pr_mflags));
}
}
}
printf_filtered ("\n");
}
}
/*
LOCAL FUNCTION
info_proc -- implement the "info proc" command
SYNOPSIS
void info_proc (char *args, int from_tty)
DESCRIPTION
Implement gdb's "info proc" command by using the /proc interface
to print status information about any currently running process.
Examples of the use of "info proc" are:
info proc (prints summary info for current inferior)
info proc 123 (prints summary info for process with pid 123)
info proc mappings (prints address mappings)
info proc times (prints process/children times)
info proc id (prints pid, ppid, gid, sid, etc)
FIXME: i proc id not implemented.
info proc status (prints general process state info)
FIXME: i proc status not implemented.
info proc signals (prints info about signal handling)
info proc all (prints all info)
*/
static void
info_proc (args, from_tty)
char *args;
int from_tty;
{
int pid;
struct procinfo *pip;
struct cleanup *old_chain;
char **argv;
int argsize;
int summary = 1;
int flags = 0;
int syscalls = 0;
int signals = 0;
int faults = 0;
int mappings = 0;
int times = 0;
int id = 0;
int status = 0;
int all = 0;
old_chain = make_cleanup (null_cleanup, 0);
/* Default to using the current inferior if no pid specified. Note
that inferior_pid may be 0, hence we set okerr. */
pip = find_procinfo (inferior_pid, 1);
if (args != NULL)
{
if ((argv = buildargv (args)) == NULL)
{
nomem (0);
}
make_cleanup (freeargv, (char *) argv);
while (*argv != NULL)
{
argsize = strlen (*argv);
if (argsize >= 1 && strncmp (*argv, "all", argsize) == 0)
{
summary = 0;
all = 1;
}
else if (argsize >= 2 && strncmp (*argv, "faults", argsize) == 0)
{
summary = 0;
faults = 1;
}
else if (argsize >= 2 && strncmp (*argv, "flags", argsize) == 0)
{
summary = 0;
flags = 1;
}
else if (argsize >= 1 && strncmp (*argv, "id", argsize) == 0)
{
summary = 0;
id = 1;
}
else if (argsize >= 1 && strncmp (*argv, "mappings", argsize) == 0)
{
summary = 0;
mappings = 1;
}
else if (argsize >= 2 && strncmp (*argv, "signals", argsize) == 0)
{
summary = 0;
signals = 1;
}
else if (argsize >= 2 && strncmp (*argv, "status", argsize) == 0)
{
summary = 0;
status = 1;
}
else if (argsize >= 2 && strncmp (*argv, "syscalls", argsize) == 0)
{
summary = 0;
syscalls = 1;
}
else if (argsize >= 1 && strncmp (*argv, "times", argsize) == 0)
{
summary = 0;
times = 1;
}
else if ((pid = atoi (*argv)) > 0)
{
pip = (struct procinfo *) xmalloc (sizeof (struct procinfo));
memset (pip, 0, sizeof (*pip));
pip->pid = pid;
if (!open_proc_file (pid, pip, O_RDONLY))
{
perror_with_name (pip -> pathname);
/* NOTREACHED */
}
make_cleanup (close_proc_file, pip);
}
else if (**argv != '\000')
{
error ("Unrecognized or ambiguous keyword `%s'.", *argv);
}
argv++;
}
}
/* If we don't have a valid open process at this point, then we have no
inferior or didn't specify a specific pid. */
if (!pip)
{
error ("\
No process. Start debugging a program or specify an explicit process ID.");
}
if (ioctl (pip -> fd, PIOCSTATUS, &(pip -> prstatus)) < 0)
{
print_sys_errmsg (pip -> pathname, errno);
error ("PIOCSTATUS failed");
}
/* Print verbose information of the requested type(s), or just a summary
of the information for all types. */
printf_filtered ("\nInformation for %s:\n\n", pip -> pathname);
if (summary || all || flags)
{
info_proc_flags (pip, summary);
}
if (summary || all)
{
info_proc_stop (pip, summary);
}
if (summary || all || signals || faults)
{
info_proc_siginfo (pip, summary);
}
if (summary || all || syscalls)
{
info_proc_syscalls (pip, summary);
}
if (summary || all || mappings)
{
info_proc_mappings (pip, summary);
}
if (summary || all || signals)
{
info_proc_signals (pip, summary);
}
if (summary || all || faults)
{
info_proc_faults (pip, summary);
}
printf_filtered ("\n");
/* All done, deal with closing any temporary process info structure,
freeing temporary memory , etc. */
do_cleanups (old_chain);
}
/*
LOCAL FUNCTION
procfs_set_sproc_trap -- arrange for exec'd child stop on sproc
SYNOPSIS
void procfs_set_sproc_trap (void)
DESCRIPTION
This function sets up a trap on sproc system call exits so that we can
detect the arrival of a new thread. We are called with the child
stopped prior to it's first instruction.
Also note that we turn on the inherit-on-fork flag in the child process
so that any grand-children start with all tracing flags set.
*/
#ifdef SYS_sproc
static void
procfs_set_sproc_trap (pi)
struct procinfo *pi;
{
sysset_t exitset;
if (ioctl (pi->fd, PIOCGEXIT, &exitset) < 0)
{
print_sys_errmsg (pi->pathname, errno);
error ("PIOCGEXIT failed");
}
praddset (&exitset, SYS_sproc);
if (ioctl (pi->fd, PIOCSEXIT, &exitset) < 0)
{
print_sys_errmsg (pi->pathname, errno);
error ("PIOCSEXIT failed");
}
/* Turn on inherit-on-fork flag so that all grand-children of gdb start with
tracing flags set. */
#ifdef PIOCSET /* New method */
{
long pr_flags;
pr_flags = PR_FORK;
ioctl (pi->fd, PIOCSET, &pr_flags);
}
#else
#ifdef PIOCSFORK /* Original method */
ioctl (pi->fd, PIOCSFORK, NULL);
#endif
#endif
}
#endif /* SYS_sproc */
/* Fork an inferior process, and start debugging it with /proc. */
static void
procfs_create_inferior (exec_file, allargs, env)
char *exec_file;
char *allargs;
char **env;
{
fork_inferior (exec_file, allargs, env,
proc_set_exec_trap, procfs_init_inferior);
/* We are at the first instruction we care about. */
/* Pedal to the metal... */
/* Setup traps on exit from sproc() */
#ifdef SYS_sproc
procfs_set_sproc_trap (current_procinfo);
#endif
proceed ((CORE_ADDR) -1, 0, 0);
}
/* Clean up after the inferior dies. */
static void
procfs_mourn_inferior ()
{
struct procinfo *pi;
for (pi = procinfo_list; pi; pi = pi->next)
unconditionally_kill_inferior (pi);
unpush_target (&procfs_ops);
generic_mourn_inferior ();
}
/* Mark our target-struct as eligible for stray "run" and "attach" commands. */
static int
procfs_can_run ()
{
return(1);
}
struct target_ops procfs_ops = {
"procfs", /* to_shortname */
"Unix /proc child process", /* to_longname */
"Unix /proc child process (started by the \"run\" command).", /* to_doc */
procfs_open, /* to_open */
0, /* to_close */
procfs_attach, /* to_attach */
procfs_detach, /* to_detach */
procfs_resume, /* to_resume */
procfs_wait, /* to_wait */
procfs_fetch_registers, /* to_fetch_registers */
procfs_store_registers, /* to_store_registers */
procfs_prepare_to_store, /* to_prepare_to_store */
procfs_xfer_memory, /* to_xfer_memory */
procfs_files_info, /* to_files_info */
memory_insert_breakpoint, /* to_insert_breakpoint */
memory_remove_breakpoint, /* to_remove_breakpoint */
terminal_init_inferior, /* to_terminal_init */
terminal_inferior, /* to_terminal_inferior */
terminal_ours_for_output, /* to_terminal_ours_for_output */
terminal_ours, /* to_terminal_ours */
child_terminal_info, /* to_terminal_info */
procfs_kill_inferior, /* to_kill */
0, /* to_load */
0, /* to_lookup_symbol */
procfs_create_inferior, /* to_create_inferior */
procfs_mourn_inferior, /* to_mourn_inferior */
procfs_can_run, /* to_can_run */
procfs_notice_signals, /* to_notice_signals */
process_stratum, /* to_stratum */
0, /* to_next */
1, /* to_has_all_memory */
1, /* to_has_memory */
1, /* to_has_stack */
1, /* to_has_registers */
1, /* to_has_execution */
0, /* sections */
0, /* sections_end */
OPS_MAGIC /* to_magic */
};
/*
GLOBAL FUNCTION
_initialize_procfs -- initialize the process file system stuff
SYNOPSIS
void _initialize_procfs (void)
DESCRIPTION
Do required initializations during gdb startup for using the
/proc file system interface.
*/
void
_initialize_procfs ()
{
add_target (&procfs_ops);
add_info ("proc", info_proc,
"Show process status information using /proc entry.\n\
Specify process id or use current inferior by default.\n\
Specify keywords for detailed information; default is summary.\n\
Keywords are: `all', `faults', `flags', `id', `mappings', `signals',\n\
`status', `syscalls', and `times'.\n\
Unambiguous abbreviations may be used.");
init_syscall_table ();
}