binutils-gdb/gdb/procfs.c
Joel Brobecker e2882c8578 Update copyright year range in all GDB files
gdb/ChangeLog:

        Update copyright year range in all GDB files
2018-01-02 07:38:06 +04:00

3966 lines
114 KiB
C

/* Machine independent support for Solaris /proc (process file system) for GDB.
Copyright (C) 1999-2018 Free Software Foundation, Inc.
Written by Michael Snyder at Cygnus Solutions.
Based on work by Fred Fish, Stu Grossman, Geoff Noer, and others.
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 3 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, see <http://www.gnu.org/licenses/>. */
#include "defs.h"
#include "inferior.h"
#include "infrun.h"
#include "target.h"
#include "gdbcore.h"
#include "elf-bfd.h" /* for elfcore_write_* */
#include "gdbcmd.h"
#include "gdbthread.h"
#include "regcache.h"
#include "inf-child.h"
#include "nat/fork-inferior.h"
#include "filestuff.h"
#define _STRUCTURED_PROC 1 /* Should be done by configure script. */
#include <sys/procfs.h>
#include <sys/fault.h>
#include <sys/syscall.h>
#include "gdb_wait.h"
#include <signal.h>
#include <ctype.h>
#include "gdb_bfd.h"
#include "inflow.h"
#include "auxv.h"
#include "procfs.h"
#include "observer.h"
/* This module provides the interface between GDB and the
/proc file system, which is used on many versions of Unix
as a means for debuggers to control other processes.
/proc works by imitating a file system: you open a simulated file
that represents the process you wish to interact with, and perform
operations on that "file" in order to examine or change the state
of the other process.
The most important thing to know about /proc and this module is
that there are two very different interfaces to /proc:
One that uses the ioctl system call, and another that uses read
and write system calls.
This module supports only the Solaris version of the read/write
interface. */
#include <sys/types.h>
#include <dirent.h> /* opendir/readdir, for listing the LWP's */
#include <fcntl.h> /* for O_RDONLY */
#include <unistd.h> /* for "X_OK" */
#include <sys/stat.h> /* for struct stat */
/* Note: procfs-utils.h must be included after the above system header
files, because it redefines various system calls using macros.
This may be incompatible with the prototype declarations. */
#include "proc-utils.h"
/* Prototypes for supply_gregset etc. */
#include "gregset.h"
/* =================== TARGET_OPS "MODULE" =================== */
/* This module defines the GDB target vector and its methods. */
static void procfs_attach (struct target_ops *, const char *, int);
static void procfs_detach (struct target_ops *, const char *, int);
static void procfs_resume (struct target_ops *,
ptid_t, int, enum gdb_signal);
static void procfs_interrupt (struct target_ops *self, ptid_t);
static void procfs_files_info (struct target_ops *);
static void procfs_fetch_registers (struct target_ops *,
struct regcache *, int);
static void procfs_store_registers (struct target_ops *,
struct regcache *, int);
static void procfs_pass_signals (struct target_ops *self,
int, unsigned char *);
static void procfs_kill_inferior (struct target_ops *ops);
static void procfs_mourn_inferior (struct target_ops *ops);
static void procfs_create_inferior (struct target_ops *, const char *,
const std::string &, char **, int);
static ptid_t procfs_wait (struct target_ops *,
ptid_t, struct target_waitstatus *, int);
static enum target_xfer_status procfs_xfer_memory (gdb_byte *,
const gdb_byte *,
ULONGEST, ULONGEST,
ULONGEST *);
static target_xfer_partial_ftype procfs_xfer_partial;
static int procfs_thread_alive (struct target_ops *ops, ptid_t);
static void procfs_update_thread_list (struct target_ops *ops);
static const char *procfs_pid_to_str (struct target_ops *, ptid_t);
static int proc_find_memory_regions (struct target_ops *self,
find_memory_region_ftype, void *);
static char *procfs_make_note_section (struct target_ops *self,
bfd *, int *);
static int procfs_can_use_hw_breakpoint (struct target_ops *self,
enum bptype, int, int);
static void procfs_info_proc (struct target_ops *, const char *,
enum info_proc_what);
#if defined (PR_MODEL_NATIVE) && (PR_MODEL_NATIVE == PR_MODEL_LP64)
/* When GDB is built as 64-bit application on Solaris, the auxv data
is presented in 64-bit format. We need to provide a custom parser
to handle that. */
static int
procfs_auxv_parse (struct target_ops *ops, gdb_byte **readptr,
gdb_byte *endptr, CORE_ADDR *typep, CORE_ADDR *valp)
{
enum bfd_endian byte_order = gdbarch_byte_order (target_gdbarch ());
gdb_byte *ptr = *readptr;
if (endptr == ptr)
return 0;
if (endptr - ptr < 8 * 2)
return -1;
*typep = extract_unsigned_integer (ptr, 4, byte_order);
ptr += 8;
/* The size of data is always 64-bit. If the application is 32-bit,
it will be zero extended, as expected. */
*valp = extract_unsigned_integer (ptr, 8, byte_order);
ptr += 8;
*readptr = ptr;
return 1;
}
#endif
struct target_ops *
procfs_target (void)
{
struct target_ops *t = inf_child_target ();
t->to_create_inferior = procfs_create_inferior;
t->to_kill = procfs_kill_inferior;
t->to_mourn_inferior = procfs_mourn_inferior;
t->to_attach = procfs_attach;
t->to_detach = procfs_detach;
t->to_wait = procfs_wait;
t->to_resume = procfs_resume;
t->to_fetch_registers = procfs_fetch_registers;
t->to_store_registers = procfs_store_registers;
t->to_xfer_partial = procfs_xfer_partial;
t->to_pass_signals = procfs_pass_signals;
t->to_files_info = procfs_files_info;
t->to_interrupt = procfs_interrupt;
t->to_update_thread_list = procfs_update_thread_list;
t->to_thread_alive = procfs_thread_alive;
t->to_pid_to_str = procfs_pid_to_str;
t->to_has_thread_control = tc_schedlock;
t->to_find_memory_regions = proc_find_memory_regions;
t->to_make_corefile_notes = procfs_make_note_section;
t->to_info_proc = procfs_info_proc;
#if defined(PR_MODEL_NATIVE) && (PR_MODEL_NATIVE == PR_MODEL_LP64)
t->to_auxv_parse = procfs_auxv_parse;
#endif
t->to_magic = OPS_MAGIC;
return t;
}
/* =================== END, TARGET_OPS "MODULE" =================== */
/* World Unification:
Put any typedefs, defines etc. here that are required for the
unification of code that handles different versions of /proc. */
enum { READ_WATCHFLAG = WA_READ,
WRITE_WATCHFLAG = WA_WRITE,
EXEC_WATCHFLAG = WA_EXEC,
AFTER_WATCHFLAG = WA_TRAPAFTER
};
/* =================== STRUCT PROCINFO "MODULE" =================== */
/* FIXME: this comment will soon be out of date W.R.T. threads. */
/* The procinfo struct is a wrapper to hold all the state information
concerning a /proc process. There should be exactly one procinfo
for each process, and since GDB currently can debug only one
process at a time, that means there should be only one procinfo.
All of the LWP's of a process can be accessed indirectly thru the
single process procinfo.
However, against the day when GDB may debug more than one process,
this data structure is kept in a list (which for now will hold no
more than one member), and many functions will have a pointer to a
procinfo as an argument.
There will be a separate procinfo structure for use by the (not yet
implemented) "info proc" command, so that we can print useful
information about any random process without interfering with the
inferior's procinfo information. */
/* format strings for /proc paths */
#define MAIN_PROC_NAME_FMT "/proc/%d"
#define CTL_PROC_NAME_FMT "/proc/%d/ctl"
#define AS_PROC_NAME_FMT "/proc/%d/as"
#define MAP_PROC_NAME_FMT "/proc/%d/map"
#define STATUS_PROC_NAME_FMT "/proc/%d/status"
#define MAX_PROC_NAME_SIZE sizeof("/proc/99999/lwp/8096/lstatus")
typedef struct procinfo {
struct procinfo *next;
int pid; /* Process ID */
int tid; /* Thread/LWP id */
/* process state */
int was_stopped;
int ignore_next_sigstop;
int ctl_fd; /* File descriptor for /proc control file */
int status_fd; /* File descriptor for /proc status file */
int as_fd; /* File descriptor for /proc as file */
char pathname[MAX_PROC_NAME_SIZE]; /* Pathname to /proc entry */
fltset_t saved_fltset; /* Saved traced hardware fault set */
sigset_t saved_sigset; /* 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 */
pstatus_t prstatus; /* Current process status info */
struct procinfo *thread_list;
int status_valid : 1;
int gregs_valid : 1;
int fpregs_valid : 1;
int threads_valid: 1;
} procinfo;
static char errmsg[128]; /* shared error msg buffer */
/* Function prototypes for procinfo module: */
static procinfo *find_procinfo_or_die (int pid, int tid);
static procinfo *find_procinfo (int pid, int tid);
static procinfo *create_procinfo (int pid, int tid);
static void destroy_procinfo (procinfo *p);
static void do_destroy_procinfo_cleanup (void *);
static void dead_procinfo (procinfo *p, const char *msg, int killp);
static int open_procinfo_files (procinfo *p, int which);
static void close_procinfo_files (procinfo *p);
static sysset_t *sysset_t_alloc (procinfo *pi);
static int iterate_over_mappings
(procinfo *pi, find_memory_region_ftype child_func, void *data,
int (*func) (struct prmap *map, find_memory_region_ftype child_func,
void *data));
/* The head of the procinfo list: */
static procinfo *procinfo_list;
/* Search the procinfo list. Return a pointer to procinfo, or NULL if
not found. */
static procinfo *
find_procinfo (int pid, int tid)
{
procinfo *pi;
for (pi = procinfo_list; pi; pi = pi->next)
if (pi->pid == pid)
break;
if (pi)
if (tid)
{
/* Don't check threads_valid. If we're updating the
thread_list, we want to find whatever threads are already
here. This means that in general it is the caller's
responsibility to check threads_valid and update before
calling find_procinfo, if the caller wants to find a new
thread. */
for (pi = pi->thread_list; pi; pi = pi->next)
if (pi->tid == tid)
break;
}
return pi;
}
/* Calls find_procinfo, but errors on failure. */
static procinfo *
find_procinfo_or_die (int pid, int tid)
{
procinfo *pi = find_procinfo (pid, tid);
if (pi == NULL)
{
if (tid)
error (_("procfs: couldn't find pid %d "
"(kernel thread %d) in procinfo list."),
pid, tid);
else
error (_("procfs: couldn't find pid %d in procinfo list."), pid);
}
return pi;
}
/* Wrapper for `open'. The appropriate open call is attempted; if
unsuccessful, it will be retried as many times as needed for the
EAGAIN and EINTR conditions.
For other conditions, retry the open a limited number of times. In
addition, a short sleep is imposed prior to retrying the open. The
reason for this sleep is to give the kernel a chance to catch up
and create the file in question in the event that GDB "wins" the
race to open a file before the kernel has created it. */
static int
open_with_retry (const char *pathname, int flags)
{
int retries_remaining, status;
retries_remaining = 2;
while (1)
{
status = open (pathname, flags);
if (status >= 0 || retries_remaining == 0)
break;
else if (errno != EINTR && errno != EAGAIN)
{
retries_remaining--;
sleep (1);
}
}
return status;
}
/* Open the file descriptor for the process or LWP. We only open the
control file descriptor; the others are opened lazily as needed.
Returns the file descriptor, or zero for failure. */
enum { FD_CTL, FD_STATUS, FD_AS };
static int
open_procinfo_files (procinfo *pi, int which)
{
char tmp[MAX_PROC_NAME_SIZE];
int fd;
/* This function is getting ALMOST long enough to break up into
several. Here is some rationale:
There are several file descriptors that may need to be open
for any given process or LWP. The ones we're intereted in are:
- control (ctl) write-only change the state
- status (status) read-only query the state
- address space (as) read/write access memory
- map (map) read-only virtual addr map
Most of these are opened lazily as they are needed.
The pathnames for the 'files' for an LWP look slightly
different from those of a first-class process:
Pathnames for a process (<proc-id>):
/proc/<proc-id>/ctl
/proc/<proc-id>/status
/proc/<proc-id>/as
/proc/<proc-id>/map
Pathnames for an LWP (lwp-id):
/proc/<proc-id>/lwp/<lwp-id>/lwpctl
/proc/<proc-id>/lwp/<lwp-id>/lwpstatus
An LWP has no map or address space file descriptor, since
the memory map and address space are shared by all LWPs. */
/* In this case, there are several different file descriptors that
we might be asked to open. The control file descriptor will be
opened early, but the others will be opened lazily as they are
needed. */
strcpy (tmp, pi->pathname);
switch (which) { /* Which file descriptor to open? */
case FD_CTL:
if (pi->tid)
strcat (tmp, "/lwpctl");
else
strcat (tmp, "/ctl");
fd = open_with_retry (tmp, O_WRONLY);
if (fd < 0)
return 0; /* fail */
pi->ctl_fd = fd;
break;
case FD_AS:
if (pi->tid)
return 0; /* There is no 'as' file descriptor for an lwp. */
strcat (tmp, "/as");
fd = open_with_retry (tmp, O_RDWR);
if (fd < 0)
return 0; /* fail */
pi->as_fd = fd;
break;
case FD_STATUS:
if (pi->tid)
strcat (tmp, "/lwpstatus");
else
strcat (tmp, "/status");
fd = open_with_retry (tmp, O_RDONLY);
if (fd < 0)
return 0; /* fail */
pi->status_fd = fd;
break;
default:
return 0; /* unknown file descriptor */
}
return 1; /* success */
}
/* Allocate a data structure and link it into the procinfo list.
First tries to find a pre-existing one (FIXME: why?). Returns the
pointer to new procinfo struct. */
static procinfo *
create_procinfo (int pid, int tid)
{
procinfo *pi, *parent = NULL;
if ((pi = find_procinfo (pid, tid)))
return pi; /* Already exists, nothing to do. */
/* Find parent before doing malloc, to save having to cleanup. */
if (tid != 0)
parent = find_procinfo_or_die (pid, 0); /* FIXME: should I
create it if it
doesn't exist yet? */
pi = XNEW (procinfo);
memset (pi, 0, sizeof (procinfo));
pi->pid = pid;
pi->tid = tid;
pi->saved_entryset = sysset_t_alloc (pi);
pi->saved_exitset = sysset_t_alloc (pi);
/* Chain into list. */
if (tid == 0)
{
sprintf (pi->pathname, MAIN_PROC_NAME_FMT, pid);
pi->next = procinfo_list;
procinfo_list = pi;
}
else
{
sprintf (pi->pathname, "/proc/%05d/lwp/%d", pid, tid);
pi->next = parent->thread_list;
parent->thread_list = pi;
}
return pi;
}
/* Close all file descriptors associated with the procinfo. */
static void
close_procinfo_files (procinfo *pi)
{
if (pi->ctl_fd > 0)
close (pi->ctl_fd);
if (pi->as_fd > 0)
close (pi->as_fd);
if (pi->status_fd > 0)
close (pi->status_fd);
pi->ctl_fd = pi->as_fd = pi->status_fd = 0;
}
/* Destructor function. Close, unlink and deallocate the object. */
static void
destroy_one_procinfo (procinfo **list, procinfo *pi)
{
procinfo *ptr;
/* Step one: unlink the procinfo from its list. */
if (pi == *list)
*list = pi->next;
else
for (ptr = *list; ptr; ptr = ptr->next)
if (ptr->next == pi)
{
ptr->next = pi->next;
break;
}
/* Step two: close any open file descriptors. */
close_procinfo_files (pi);
/* Step three: free the memory. */
xfree (pi->saved_entryset);
xfree (pi->saved_exitset);
xfree (pi);
}
static void
destroy_procinfo (procinfo *pi)
{
procinfo *tmp;
if (pi->tid != 0) /* Destroy a thread procinfo. */
{
tmp = find_procinfo (pi->pid, 0); /* Find the parent process. */
destroy_one_procinfo (&tmp->thread_list, pi);
}
else /* Destroy a process procinfo and all its threads. */
{
/* First destroy the children, if any; */
while (pi->thread_list != NULL)
destroy_one_procinfo (&pi->thread_list, pi->thread_list);
/* Then destroy the parent. Genocide!!! */
destroy_one_procinfo (&procinfo_list, pi);
}
}
static void
do_destroy_procinfo_cleanup (void *pi)
{
destroy_procinfo ((procinfo *) pi);
}
enum { NOKILL, KILL };
/* To be called on a non_recoverable error for a procinfo. Prints
error messages, optionally sends a SIGKILL to the process, then
destroys the data structure. */
static void
dead_procinfo (procinfo *pi, const char *msg, int kill_p)
{
char procfile[80];
if (pi->pathname)
{
print_sys_errmsg (pi->pathname, errno);
}
else
{
sprintf (procfile, "process %d", pi->pid);
print_sys_errmsg (procfile, errno);
}
if (kill_p == KILL)
kill (pi->pid, SIGKILL);
destroy_procinfo (pi);
error ("%s", msg);
}
/* Allocate and (partially) initialize a sysset_t struct. */
static sysset_t *
sysset_t_alloc (procinfo *pi)
{
return (sysset_t *) xmalloc (sizeof (sysset_t));
}
/* =================== END, STRUCT PROCINFO "MODULE" =================== */
/* =================== /proc "MODULE" =================== */
/* This "module" is the interface layer between the /proc system API
and the gdb target vector functions. This layer consists of access
functions that encapsulate each of the basic operations that we
need to use from the /proc API.
The main motivation for this layer is to hide the fact that there
are two very different implementations of the /proc API. Rather
than have a bunch of #ifdefs all thru the gdb target vector
functions, we do our best to hide them all in here. */
static long proc_flags (procinfo *pi);
static int proc_why (procinfo *pi);
static int proc_what (procinfo *pi);
static int proc_set_current_signal (procinfo *pi, int signo);
static int proc_get_current_thread (procinfo *pi);
static int proc_iterate_over_threads
(procinfo *pi,
int (*func) (procinfo *, procinfo *, void *),
void *ptr);
static void
proc_warn (procinfo *pi, const char *func, int line)
{
sprintf (errmsg, "procfs: %s line %d, %s", func, line, pi->pathname);
print_sys_errmsg (errmsg, errno);
}
static void
proc_error (procinfo *pi, const char *func, int line)
{
sprintf (errmsg, "procfs: %s line %d, %s", func, line, pi->pathname);
perror_with_name (errmsg);
}
/* Updates the status struct in the procinfo. There is a 'valid'
flag, to let other functions know when this function needs to be
called (so the status is only read when it is needed). The status
file descriptor is also only opened when it is needed. Returns
non-zero for success, zero for failure. */
static int
proc_get_status (procinfo *pi)
{
/* Status file descriptor is opened "lazily". */
if (pi->status_fd == 0 &&
open_procinfo_files (pi, FD_STATUS) == 0)
{
pi->status_valid = 0;
return 0;
}
if (lseek (pi->status_fd, 0, SEEK_SET) < 0)
pi->status_valid = 0; /* fail */
else
{
/* Sigh... I have to read a different data structure,
depending on whether this is a main process or an LWP. */
if (pi->tid)
pi->status_valid = (read (pi->status_fd,
(char *) &pi->prstatus.pr_lwp,
sizeof (lwpstatus_t))
== sizeof (lwpstatus_t));
else
{
pi->status_valid = (read (pi->status_fd,
(char *) &pi->prstatus,
sizeof (pstatus_t))
== sizeof (pstatus_t));
}
}
if (pi->status_valid)
{
PROC_PRETTYFPRINT_STATUS (proc_flags (pi),
proc_why (pi),
proc_what (pi),
proc_get_current_thread (pi));
}
/* The status struct includes general regs, so mark them valid too. */
pi->gregs_valid = pi->status_valid;
/* In the read/write multiple-fd model, the status struct includes
the fp regs too, so mark them valid too. */
pi->fpregs_valid = pi->status_valid;
return pi->status_valid; /* True if success, false if failure. */
}
/* Returns the process flags (pr_flags field). */
static long
proc_flags (procinfo *pi)
{
if (!pi->status_valid)
if (!proc_get_status (pi))
return 0; /* FIXME: not a good failure value (but what is?) */
return pi->prstatus.pr_lwp.pr_flags;
}
/* Returns the pr_why field (why the process stopped). */
static int
proc_why (procinfo *pi)
{
if (!pi->status_valid)
if (!proc_get_status (pi))
return 0; /* FIXME: not a good failure value (but what is?) */
return pi->prstatus.pr_lwp.pr_why;
}
/* Returns the pr_what field (details of why the process stopped). */
static int
proc_what (procinfo *pi)
{
if (!pi->status_valid)
if (!proc_get_status (pi))
return 0; /* FIXME: not a good failure value (but what is?) */
return pi->prstatus.pr_lwp.pr_what;
}
/* This function is only called when PI is stopped by a watchpoint.
Assuming the OS supports it, write to *ADDR the data address which
triggered it and return 1. Return 0 if it is not possible to know
the address. */
static int
proc_watchpoint_address (procinfo *pi, CORE_ADDR *addr)
{
if (!pi->status_valid)
if (!proc_get_status (pi))
return 0;
*addr = (CORE_ADDR) gdbarch_pointer_to_address (target_gdbarch (),
builtin_type (target_gdbarch ())->builtin_data_ptr,
(gdb_byte *) &pi->prstatus.pr_lwp.pr_info.si_addr);
return 1;
}
/* Returns the pr_nsysarg field (number of args to the current
syscall). */
static int
proc_nsysarg (procinfo *pi)
{
if (!pi->status_valid)
if (!proc_get_status (pi))
return 0;
return pi->prstatus.pr_lwp.pr_nsysarg;
}
/* Returns the pr_sysarg field (pointer to the arguments of current
syscall). */
static long *
proc_sysargs (procinfo *pi)
{
if (!pi->status_valid)
if (!proc_get_status (pi))
return NULL;
return (long *) &pi->prstatus.pr_lwp.pr_sysarg;
}
/* Set or reset any of the following process flags:
PR_FORK -- forked child will inherit trace flags
PR_RLC -- traced process runs when last /proc file closed.
PR_KLC -- traced process is killed when last /proc file closed.
PR_ASYNC -- LWP's get to run/stop independently.
This function is done using read/write [PCSET/PCRESET/PCUNSET].
Arguments:
pi -- the procinfo
flag -- one of PR_FORK, PR_RLC, or PR_ASYNC
mode -- 1 for set, 0 for reset.
Returns non-zero for success, zero for failure. */
enum { FLAG_RESET, FLAG_SET };
static int
proc_modify_flag (procinfo *pi, long flag, long mode)
{
long win = 0; /* default to fail */
/* These operations affect the process as a whole, and applying them
to an individual LWP has the same meaning as applying them to the
main process. Therefore, if we're ever called with a pointer to
an LWP's procinfo, let's substitute the process's procinfo and
avoid opening the LWP's file descriptor unnecessarily. */
if (pi->pid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
procfs_ctl_t arg[2];
if (mode == FLAG_SET) /* Set the flag (RLC, FORK, or ASYNC). */
arg[0] = PCSET;
else /* Reset the flag. */
arg[0] = PCUNSET;
arg[1] = flag;
win = (write (pi->ctl_fd, (void *) &arg, sizeof (arg)) == sizeof (arg));
/* The above operation renders the procinfo's cached pstatus
obsolete. */
pi->status_valid = 0;
if (!win)
warning (_("procfs: modify_flag failed to turn %s %s"),
flag == PR_FORK ? "PR_FORK" :
flag == PR_RLC ? "PR_RLC" :
flag == PR_ASYNC ? "PR_ASYNC" :
flag == PR_KLC ? "PR_KLC" :
"<unknown flag>",
mode == FLAG_RESET ? "off" : "on");
return win;
}
/* Set the run_on_last_close flag. Process with all threads will
become runnable when debugger closes all /proc fds. Returns
non-zero for success, zero for failure. */
static int
proc_set_run_on_last_close (procinfo *pi)
{
return proc_modify_flag (pi, PR_RLC, FLAG_SET);
}
/* Reset the run_on_last_close flag. The process will NOT become
runnable when debugger closes its file handles. Returns non-zero
for success, zero for failure. */
static int
proc_unset_run_on_last_close (procinfo *pi)
{
return proc_modify_flag (pi, PR_RLC, FLAG_RESET);
}
/* Reset inherit_on_fork flag. If the process forks a child while we
are registered for events in the parent, then we will NOT recieve
events from the child. Returns non-zero for success, zero for
failure. */
static int
proc_unset_inherit_on_fork (procinfo *pi)
{
return proc_modify_flag (pi, PR_FORK, FLAG_RESET);
}
/* Set PR_ASYNC flag. If one LWP stops because of a debug event
(signal etc.), the remaining LWPs will continue to run. Returns
non-zero for success, zero for failure. */
static int
proc_set_async (procinfo *pi)
{
return proc_modify_flag (pi, PR_ASYNC, FLAG_SET);
}
/* Reset PR_ASYNC flag. If one LWP stops because of a debug event
(signal etc.), then all other LWPs will stop as well. Returns
non-zero for success, zero for failure. */
static int
proc_unset_async (procinfo *pi)
{
return proc_modify_flag (pi, PR_ASYNC, FLAG_RESET);
}
/* Request the process/LWP to stop. Does not wait. Returns non-zero
for success, zero for failure. */
static int
proc_stop_process (procinfo *pi)
{
int win;
/* We might conceivably apply this operation to an LWP, and the
LWP's ctl file descriptor might not be open. */
if (pi->ctl_fd == 0 &&
open_procinfo_files (pi, FD_CTL) == 0)
return 0;
else
{
procfs_ctl_t cmd = PCSTOP;
win = (write (pi->ctl_fd, (char *) &cmd, sizeof (cmd)) == sizeof (cmd));
}
return win;
}
/* Wait for the process or LWP to stop (block until it does). Returns
non-zero for success, zero for failure. */
static int
proc_wait_for_stop (procinfo *pi)
{
int win;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
procfs_ctl_t cmd = PCWSTOP;
win = (write (pi->ctl_fd, (char *) &cmd, sizeof (cmd)) == sizeof (cmd));
/* We been runnin' and we stopped -- need to update status. */
pi->status_valid = 0;
return win;
}
/* Make the process or LWP runnable.
Options (not all are implemented):
- single-step
- clear current fault
- clear current signal
- abort the current system call
- stop as soon as finished with system call
- (ioctl): set traced signal set
- (ioctl): set held signal set
- (ioctl): set traced fault set
- (ioctl): set start pc (vaddr)
Always clears the current fault. PI is the process or LWP to
operate on. If STEP is true, set the process or LWP to trap after
one instruction. If SIGNO is zero, clear the current signal if
any; if non-zero, set the current signal to this one. Returns
non-zero for success, zero for failure. */
static int
proc_run_process (procinfo *pi, int step, int signo)
{
int win;
int runflags;
/* We will probably have to apply this operation to individual
threads, so make sure the control file descriptor is open. */
if (pi->ctl_fd == 0 &&
open_procinfo_files (pi, FD_CTL) == 0)
{
return 0;
}
runflags = PRCFAULT; /* Always clear current fault. */
if (step)
runflags |= PRSTEP;
if (signo == 0)
runflags |= PRCSIG;
else if (signo != -1) /* -1 means do nothing W.R.T. signals. */
proc_set_current_signal (pi, signo);
procfs_ctl_t cmd[2];
cmd[0] = PCRUN;
cmd[1] = runflags;
win = (write (pi->ctl_fd, (char *) &cmd, sizeof (cmd)) == sizeof (cmd));
return win;
}
/* Register to trace signals in the process or LWP. Returns non-zero
for success, zero for failure. */
static int
proc_set_traced_signals (procinfo *pi, sigset_t *sigset)
{
int win;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
struct {
procfs_ctl_t cmd;
/* Use char array to avoid alignment issues. */
char sigset[sizeof (sigset_t)];
} arg;
arg.cmd = PCSTRACE;
memcpy (&arg.sigset, sigset, sizeof (sigset_t));
win = (write (pi->ctl_fd, (char *) &arg, sizeof (arg)) == sizeof (arg));
/* The above operation renders the procinfo's cached pstatus obsolete. */
pi->status_valid = 0;
if (!win)
warning (_("procfs: set_traced_signals failed"));
return win;
}
/* Register to trace hardware faults in the process or LWP. Returns
non-zero for success, zero for failure. */
static int
proc_set_traced_faults (procinfo *pi, fltset_t *fltset)
{
int win;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
struct {
procfs_ctl_t cmd;
/* Use char array to avoid alignment issues. */
char fltset[sizeof (fltset_t)];
} arg;
arg.cmd = PCSFAULT;
memcpy (&arg.fltset, fltset, sizeof (fltset_t));
win = (write (pi->ctl_fd, (char *) &arg, sizeof (arg)) == sizeof (arg));
/* The above operation renders the procinfo's cached pstatus obsolete. */
pi->status_valid = 0;
return win;
}
/* Register to trace entry to system calls in the process or LWP.
Returns non-zero for success, zero for failure. */
static int
proc_set_traced_sysentry (procinfo *pi, sysset_t *sysset)
{
int win;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
struct gdb_proc_ctl_pcsentry {
procfs_ctl_t cmd;
/* Use char array to avoid alignment issues. */
char sysset[sizeof (sysset_t)];
} *argp;
int argp_size = sizeof (struct gdb_proc_ctl_pcsentry);
argp = (struct gdb_proc_ctl_pcsentry *) xmalloc (argp_size);
argp->cmd = PCSENTRY;
memcpy (&argp->sysset, sysset, sizeof (sysset_t));
win = (write (pi->ctl_fd, (char *) argp, argp_size) == argp_size);
xfree (argp);
/* The above operation renders the procinfo's cached pstatus
obsolete. */
pi->status_valid = 0;
return win;
}
/* Register to trace exit from system calls in the process or LWP.
Returns non-zero for success, zero for failure. */
static int
proc_set_traced_sysexit (procinfo *pi, sysset_t *sysset)
{
int win;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
struct gdb_proc_ctl_pcsexit {
procfs_ctl_t cmd;
/* Use char array to avoid alignment issues. */
char sysset[sizeof (sysset_t)];
} *argp;
int argp_size = sizeof (struct gdb_proc_ctl_pcsexit);
argp = (struct gdb_proc_ctl_pcsexit *) xmalloc (argp_size);
argp->cmd = PCSEXIT;
memcpy (&argp->sysset, sysset, sizeof (sysset_t));
win = (write (pi->ctl_fd, (char *) argp, argp_size) == argp_size);
xfree (argp);
/* The above operation renders the procinfo's cached pstatus
obsolete. */
pi->status_valid = 0;
return win;
}
/* Specify the set of blocked / held signals in the process or LWP.
Returns non-zero for success, zero for failure. */
static int
proc_set_held_signals (procinfo *pi, sigset_t *sighold)
{
int win;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
struct {
procfs_ctl_t cmd;
/* Use char array to avoid alignment issues. */
char hold[sizeof (sigset_t)];
} arg;
arg.cmd = PCSHOLD;
memcpy (&arg.hold, sighold, sizeof (sigset_t));
win = (write (pi->ctl_fd, (void *) &arg, sizeof (arg)) == sizeof (arg));
/* The above operation renders the procinfo's cached pstatus
obsolete. */
pi->status_valid = 0;
return win;
}
/* Returns the set of signals that are held / blocked. Will also copy
the sigset if SAVE is non-zero. */
static sigset_t *
proc_get_held_signals (procinfo *pi, sigset_t *save)
{
sigset_t *ret = NULL;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
if (!pi->status_valid)
if (!proc_get_status (pi))
return NULL;
ret = &pi->prstatus.pr_lwp.pr_lwphold;
if (save && ret)
memcpy (save, ret, sizeof (sigset_t));
return ret;
}
/* Returns the set of signals that are traced / debugged. Will also
copy the sigset if SAVE is non-zero. */
static sigset_t *
proc_get_traced_signals (procinfo *pi, sigset_t *save)
{
sigset_t *ret = NULL;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
if (!pi->status_valid)
if (!proc_get_status (pi))
return NULL;
ret = &pi->prstatus.pr_sigtrace;
if (save && ret)
memcpy (save, ret, sizeof (sigset_t));
return ret;
}
/* Returns the set of hardware faults that are traced /debugged. Will
also copy the faultset if SAVE is non-zero. */
static fltset_t *
proc_get_traced_faults (procinfo *pi, fltset_t *save)
{
fltset_t *ret = NULL;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
if (!pi->status_valid)
if (!proc_get_status (pi))
return NULL;
ret = &pi->prstatus.pr_flttrace;
if (save && ret)
memcpy (save, ret, sizeof (fltset_t));
return ret;
}
/* Returns the set of syscalls that are traced /debugged on entry.
Will also copy the syscall set if SAVE is non-zero. */
static sysset_t *
proc_get_traced_sysentry (procinfo *pi, sysset_t *save)
{
sysset_t *ret = NULL;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
if (!pi->status_valid)
if (!proc_get_status (pi))
return NULL;
ret = &pi->prstatus.pr_sysentry;
if (save && ret)
memcpy (save, ret, sizeof (sysset_t));
return ret;
}
/* Returns the set of syscalls that are traced /debugged on exit.
Will also copy the syscall set if SAVE is non-zero. */
static sysset_t *
proc_get_traced_sysexit (procinfo *pi, sysset_t *save)
{
sysset_t *ret = NULL;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
if (!pi->status_valid)
if (!proc_get_status (pi))
return NULL;
ret = &pi->prstatus.pr_sysexit;
if (save && ret)
memcpy (save, ret, sizeof (sysset_t));
return ret;
}
/* The current fault (if any) is cleared; the associated signal will
not be sent to the process or LWP when it resumes. Returns
non-zero for success, zero for failure. */
static int
proc_clear_current_fault (procinfo *pi)
{
int win;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
procfs_ctl_t cmd = PCCFAULT;
win = (write (pi->ctl_fd, (void *) &cmd, sizeof (cmd)) == sizeof (cmd));
return win;
}
/* Set the "current signal" that will be delivered next to the
process. NOTE: semantics are different from those of KILL. This
signal will be delivered to the process or LWP immediately when it
is resumed (even if the signal is held/blocked); it will NOT
immediately cause another event of interest, and will NOT first
trap back to the debugger. Returns non-zero for success, zero for
failure. */
static int
proc_set_current_signal (procinfo *pi, int signo)
{
int win;
struct {
procfs_ctl_t cmd;
/* Use char array to avoid alignment issues. */
char sinfo[sizeof (siginfo_t)];
} arg;
siginfo_t mysinfo;
ptid_t wait_ptid;
struct target_waitstatus wait_status;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
/* The pointer is just a type alias. */
get_last_target_status (&wait_ptid, &wait_status);
if (ptid_equal (wait_ptid, inferior_ptid)
&& wait_status.kind == TARGET_WAITKIND_STOPPED
&& wait_status.value.sig == gdb_signal_from_host (signo)
&& proc_get_status (pi)
&& pi->prstatus.pr_lwp.pr_info.si_signo == signo
)
/* Use the siginfo associated with the signal being
redelivered. */
memcpy (arg.sinfo, &pi->prstatus.pr_lwp.pr_info, sizeof (siginfo_t));
else
{
mysinfo.si_signo = signo;
mysinfo.si_code = 0;
mysinfo.si_pid = getpid (); /* ?why? */
mysinfo.si_uid = getuid (); /* ?why? */
memcpy (arg.sinfo, &mysinfo, sizeof (siginfo_t));
}
arg.cmd = PCSSIG;
win = (write (pi->ctl_fd, (void *) &arg, sizeof (arg)) == sizeof (arg));
return win;
}
/* The current signal (if any) is cleared, and is not sent to the
process or LWP when it resumes. Returns non-zero for success, zero
for failure. */
static int
proc_clear_current_signal (procinfo *pi)
{
int win;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
struct {
procfs_ctl_t cmd;
/* Use char array to avoid alignment issues. */
char sinfo[sizeof (siginfo_t)];
} arg;
siginfo_t mysinfo;
arg.cmd = PCSSIG;
/* The pointer is just a type alias. */
mysinfo.si_signo = 0;
mysinfo.si_code = 0;
mysinfo.si_errno = 0;
mysinfo.si_pid = getpid (); /* ?why? */
mysinfo.si_uid = getuid (); /* ?why? */
memcpy (arg.sinfo, &mysinfo, sizeof (siginfo_t));
win = (write (pi->ctl_fd, (void *) &arg, sizeof (arg)) == sizeof (arg));
return win;
}
/* Return the general-purpose registers for the process or LWP
corresponding to PI. Upon failure, return NULL. */
static gdb_gregset_t *
proc_get_gregs (procinfo *pi)
{
if (!pi->status_valid || !pi->gregs_valid)
if (!proc_get_status (pi))
return NULL;
return &pi->prstatus.pr_lwp.pr_reg;
}
/* Return the general-purpose registers for the process or LWP
corresponding to PI. Upon failure, return NULL. */
static gdb_fpregset_t *
proc_get_fpregs (procinfo *pi)
{
if (!pi->status_valid || !pi->fpregs_valid)
if (!proc_get_status (pi))
return NULL;
return &pi->prstatus.pr_lwp.pr_fpreg;
}
/* Write the general-purpose registers back to the process or LWP
corresponding to PI. Return non-zero for success, zero for
failure. */
static int
proc_set_gregs (procinfo *pi)
{
gdb_gregset_t *gregs;
int win;
gregs = proc_get_gregs (pi);
if (gregs == NULL)
return 0; /* proc_get_regs has already warned. */
if (pi->ctl_fd == 0 && open_procinfo_files (pi, FD_CTL) == 0)
{
return 0;
}
else
{
struct {
procfs_ctl_t cmd;
/* Use char array to avoid alignment issues. */
char gregs[sizeof (gdb_gregset_t)];
} arg;
arg.cmd = PCSREG;
memcpy (&arg.gregs, gregs, sizeof (arg.gregs));
win = (write (pi->ctl_fd, (void *) &arg, sizeof (arg)) == sizeof (arg));
}
/* Policy: writing the registers invalidates our cache. */
pi->gregs_valid = 0;
return win;
}
/* Write the floating-pointer registers back to the process or LWP
corresponding to PI. Return non-zero for success, zero for
failure. */
static int
proc_set_fpregs (procinfo *pi)
{
gdb_fpregset_t *fpregs;
int win;
fpregs = proc_get_fpregs (pi);
if (fpregs == NULL)
return 0; /* proc_get_fpregs has already warned. */
if (pi->ctl_fd == 0 && open_procinfo_files (pi, FD_CTL) == 0)
{
return 0;
}
else
{
struct {
procfs_ctl_t cmd;
/* Use char array to avoid alignment issues. */
char fpregs[sizeof (gdb_fpregset_t)];
} arg;
arg.cmd = PCSFPREG;
memcpy (&arg.fpregs, fpregs, sizeof (arg.fpregs));
win = (write (pi->ctl_fd, (void *) &arg, sizeof (arg)) == sizeof (arg));
}
/* Policy: writing the registers invalidates our cache. */
pi->fpregs_valid = 0;
return win;
}
/* Send a signal to the proc or lwp with the semantics of "kill()".
Returns non-zero for success, zero for failure. */
static int
proc_kill (procinfo *pi, int signo)
{
int win;
/* We might conceivably apply this operation to an LWP, and the
LWP's ctl file descriptor might not be open. */
if (pi->ctl_fd == 0 &&
open_procinfo_files (pi, FD_CTL) == 0)
{
return 0;
}
else
{
procfs_ctl_t cmd[2];
cmd[0] = PCKILL;
cmd[1] = signo;
win = (write (pi->ctl_fd, (char *) &cmd, sizeof (cmd)) == sizeof (cmd));
}
return win;
}
/* Find the pid of the process that started this one. Returns the
parent process pid, or zero. */
static int
proc_parent_pid (procinfo *pi)
{
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
if (!pi->status_valid)
if (!proc_get_status (pi))
return 0;
return pi->prstatus.pr_ppid;
}
/* Convert a target address (a.k.a. CORE_ADDR) into a host address
(a.k.a void pointer)! */
static void *
procfs_address_to_host_pointer (CORE_ADDR addr)
{
struct type *ptr_type = builtin_type (target_gdbarch ())->builtin_data_ptr;
void *ptr;
gdb_assert (sizeof (ptr) == TYPE_LENGTH (ptr_type));
gdbarch_address_to_pointer (target_gdbarch (), ptr_type,
(gdb_byte *) &ptr, addr);
return ptr;
}
static int
proc_set_watchpoint (procinfo *pi, CORE_ADDR addr, int len, int wflags)
{
struct {
procfs_ctl_t cmd;
char watch[sizeof (prwatch_t)];
} arg;
prwatch_t pwatch;
/* NOTE: cagney/2003-02-01: Even more horrible hack. Need to
convert a target address into something that can be stored in a
native data structure. */
pwatch.pr_vaddr = (uintptr_t) procfs_address_to_host_pointer (addr);
pwatch.pr_size = len;
pwatch.pr_wflags = wflags;
arg.cmd = PCWATCH;
memcpy (arg.watch, &pwatch, sizeof (prwatch_t));
return (write (pi->ctl_fd, &arg, sizeof (arg)) == sizeof (arg));
}
#if (defined(__i386__) || defined(__x86_64__)) && defined (sun)
#include <sys/sysi86.h>
/* The KEY is actually the value of the lower 16 bits of the GS
register for the LWP that we're interested in. Returns the
matching ssh struct (LDT entry). */
static struct ssd *
proc_get_LDT_entry (procinfo *pi, int key)
{
static struct ssd *ldt_entry = NULL;
char pathname[MAX_PROC_NAME_SIZE];
struct cleanup *old_chain = NULL;
int fd;
/* Allocate space for one LDT entry.
This alloc must persist, because we return a pointer to it. */
if (ldt_entry == NULL)
ldt_entry = XNEW (struct ssd);
/* Open the file descriptor for the LDT table. */
sprintf (pathname, "/proc/%d/ldt", pi->pid);
if ((fd = open_with_retry (pathname, O_RDONLY)) < 0)
{
proc_warn (pi, "proc_get_LDT_entry (open)", __LINE__);
return NULL;
}
/* Make sure it gets closed again! */
old_chain = make_cleanup_close (fd);
/* Now 'read' thru the table, find a match and return it. */
while (read (fd, ldt_entry, sizeof (struct ssd)) == sizeof (struct ssd))
{
if (ldt_entry->sel == 0 &&
ldt_entry->bo == 0 &&
ldt_entry->acc1 == 0 &&
ldt_entry->acc2 == 0)
break; /* end of table */
/* If key matches, return this entry. */
if (ldt_entry->sel == key)
{
do_cleanups (old_chain);
return ldt_entry;
}
}
/* Loop ended, match not found. */
do_cleanups (old_chain);
return NULL;
}
/* Returns the pointer to the LDT entry of PTID. */
struct ssd *
procfs_find_LDT_entry (ptid_t ptid)
{
gdb_gregset_t *gregs;
int key;
procinfo *pi;
/* Find procinfo for the lwp. */
if ((pi = find_procinfo (ptid_get_pid (ptid), ptid_get_lwp (ptid))) == NULL)
{
warning (_("procfs_find_LDT_entry: could not find procinfo for %d:%ld."),
ptid_get_pid (ptid), ptid_get_lwp (ptid));
return NULL;
}
/* get its general registers. */
if ((gregs = proc_get_gregs (pi)) == NULL)
{
warning (_("procfs_find_LDT_entry: could not read gregs for %d:%ld."),
ptid_get_pid (ptid), ptid_get_lwp (ptid));
return NULL;
}
/* Now extract the GS register's lower 16 bits. */
key = (*gregs)[GS] & 0xffff;
/* Find the matching entry and return it. */
return proc_get_LDT_entry (pi, key);
}
#endif
/* =============== END, non-thread part of /proc "MODULE" =============== */
/* =================== Thread "MODULE" =================== */
/* NOTE: you'll see more ifdefs and duplication of functions here,
since there is a different way to do threads on every OS. */
/* Returns the number of threads for the process. */
static int
proc_get_nthreads (procinfo *pi)
{
if (!pi->status_valid)
if (!proc_get_status (pi))
return 0;
/* Only works for the process procinfo, because the LWP procinfos do not
get prstatus filled in. */
if (pi->tid != 0) /* Find the parent process procinfo. */
pi = find_procinfo_or_die (pi->pid, 0);
return pi->prstatus.pr_nlwp;
}
/* LWP version.
Return the ID of the thread that had an event of interest.
(ie. the one that hit a breakpoint or other traced event). All
other things being equal, this should be the ID of a thread that is
currently executing. */
static int
proc_get_current_thread (procinfo *pi)
{
/* Note: this should be applied to the root procinfo for the
process, not to the procinfo for an LWP. If applied to the
procinfo for an LWP, it will simply return that LWP's ID. In
that case, find the parent process procinfo. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
if (!pi->status_valid)
if (!proc_get_status (pi))
return 0;
return pi->prstatus.pr_lwp.pr_lwpid;
}
/* Discover the IDs of all the threads within the process, and create
a procinfo for each of them (chained to the parent). This
unfortunately requires a different method on every OS. Returns
non-zero for success, zero for failure. */
static int
proc_delete_dead_threads (procinfo *parent, procinfo *thread, void *ignore)
{
if (thread && parent) /* sanity */
{
thread->status_valid = 0;
if (!proc_get_status (thread))
destroy_one_procinfo (&parent->thread_list, thread);
}
return 0; /* keep iterating */
}
static void
do_closedir_cleanup (void *dir)
{
closedir ((DIR *) dir);
}
static int
proc_update_threads (procinfo *pi)
{
char pathname[MAX_PROC_NAME_SIZE + 16];
struct dirent *direntry;
struct cleanup *old_chain = NULL;
procinfo *thread;
DIR *dirp;
int lwpid;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
proc_iterate_over_threads (pi, proc_delete_dead_threads, NULL);
/* Note: this brute-force method was originally devised for Unixware
(support removed since), and will also work on Solaris 2.6 and
2.7. The original comment mentioned the existence of a much
simpler and more elegant way to do this on Solaris, but didn't
point out what that was. */
strcpy (pathname, pi->pathname);
strcat (pathname, "/lwp");
if ((dirp = opendir (pathname)) == NULL)
proc_error (pi, "update_threads, opendir", __LINE__);
old_chain = make_cleanup (do_closedir_cleanup, dirp);
while ((direntry = readdir (dirp)) != NULL)
if (direntry->d_name[0] != '.') /* skip '.' and '..' */
{
lwpid = atoi (&direntry->d_name[0]);
if ((thread = create_procinfo (pi->pid, lwpid)) == NULL)
proc_error (pi, "update_threads, create_procinfo", __LINE__);
}
pi->threads_valid = 1;
do_cleanups (old_chain);
return 1;
}
/* Given a pointer to a function, call that function once for each lwp
in the procinfo list, until the function returns non-zero, in which
event return the value returned by the function.
Note: this function does NOT call update_threads. If you want to
discover new threads first, you must call that function explicitly.
This function just makes a quick pass over the currently-known
procinfos.
PI is the parent process procinfo. FUNC is the per-thread
function. PTR is an opaque parameter for function. Returns the
first non-zero return value from the callee, or zero. */
static int
proc_iterate_over_threads (procinfo *pi,
int (*func) (procinfo *, procinfo *, void *),
void *ptr)
{
procinfo *thread, *next;
int retval = 0;
/* We should never have to apply this operation to any procinfo
except the one for the main process. If that ever changes for
any reason, then take out the following clause and replace it
with one that makes sure the ctl_fd is open. */
if (pi->tid != 0)
pi = find_procinfo_or_die (pi->pid, 0);
for (thread = pi->thread_list; thread != NULL; thread = next)
{
next = thread->next; /* In case thread is destroyed. */
if ((retval = (*func) (pi, thread, ptr)) != 0)
break;
}
return retval;
}
/* =================== END, Thread "MODULE" =================== */
/* =================== END, /proc "MODULE" =================== */
/* =================== GDB "MODULE" =================== */
/* Here are all of the gdb target vector functions and their
friends. */
static ptid_t do_attach (ptid_t ptid);
static void do_detach (int signo);
static void proc_trace_syscalls_1 (procinfo *pi, int syscallnum,
int entry_or_exit, int mode, int from_tty);
/* Sets up the inferior to be debugged. Registers to trace signals,
hardware faults, and syscalls. Note: does not set RLC flag: caller
may want to customize that. Returns zero for success (note!
unlike most functions in this module); on failure, returns the LINE
NUMBER where it failed! */
static int
procfs_debug_inferior (procinfo *pi)
{
fltset_t traced_faults;
sigset_t traced_signals;
sysset_t *traced_syscall_entries;
sysset_t *traced_syscall_exits;
int status;
/* Register to trace hardware faults in the child. */
prfillset (&traced_faults); /* trace all faults... */
prdelset (&traced_faults, FLTPAGE); /* except page fault. */
if (!proc_set_traced_faults (pi, &traced_faults))
return __LINE__;
/* Initially, register to trace all signals in the child. */
prfillset (&traced_signals);
if (!proc_set_traced_signals (pi, &traced_signals))
return __LINE__;
/* Register to trace the 'exit' system call (on entry). */
traced_syscall_entries = sysset_t_alloc (pi);
premptyset (traced_syscall_entries);
praddset (traced_syscall_entries, SYS_exit);
praddset (traced_syscall_entries, SYS_lwp_exit);
status = proc_set_traced_sysentry (pi, traced_syscall_entries);
xfree (traced_syscall_entries);
if (!status)
return __LINE__;
/* Method for tracing exec syscalls. */
/* 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. */
traced_syscall_exits = sysset_t_alloc (pi);
premptyset (traced_syscall_exits);
#ifdef SYS_exec
praddset (traced_syscall_exits, SYS_exec);
#endif
praddset (traced_syscall_exits, SYS_execve);
praddset (traced_syscall_exits, SYS_lwp_create);
praddset (traced_syscall_exits, SYS_lwp_exit);
status = proc_set_traced_sysexit (pi, traced_syscall_exits);
xfree (traced_syscall_exits);
if (!status)
return __LINE__;
return 0;
}
static void
procfs_attach (struct target_ops *ops, const char *args, int from_tty)
{
char *exec_file;
int pid;
pid = parse_pid_to_attach (args);
if (pid == getpid ())
error (_("Attaching GDB to itself is not a good idea..."));
if (from_tty)
{
exec_file = get_exec_file (0);
if (exec_file)
printf_filtered (_("Attaching to program `%s', %s\n"),
exec_file, target_pid_to_str (pid_to_ptid (pid)));
else
printf_filtered (_("Attaching to %s\n"),
target_pid_to_str (pid_to_ptid (pid)));
fflush (stdout);
}
inferior_ptid = do_attach (pid_to_ptid (pid));
if (!target_is_pushed (ops))
push_target (ops);
}
static void
procfs_detach (struct target_ops *ops, const char *args, int from_tty)
{
int sig = 0;
int pid = ptid_get_pid (inferior_ptid);
if (args)
sig = atoi (args);
if (from_tty)
{
const char *exec_file;
exec_file = get_exec_file (0);
if (exec_file == NULL)
exec_file = "";
printf_filtered (_("Detaching from program: %s, %s\n"), exec_file,
target_pid_to_str (pid_to_ptid (pid)));
gdb_flush (gdb_stdout);
}
do_detach (sig);
inferior_ptid = null_ptid;
detach_inferior (pid);
inf_child_maybe_unpush_target (ops);
}
static ptid_t
do_attach (ptid_t ptid)
{
procinfo *pi;
struct inferior *inf;
int fail;
int lwpid;
if ((pi = create_procinfo (ptid_get_pid (ptid), 0)) == NULL)
perror (_("procfs: out of memory in 'attach'"));
if (!open_procinfo_files (pi, FD_CTL))
{
fprintf_filtered (gdb_stderr, "procfs:%d -- ", __LINE__);
sprintf (errmsg, "do_attach: couldn't open /proc file for process %d",
ptid_get_pid (ptid));
dead_procinfo (pi, errmsg, NOKILL);
}
/* Stop the process (if it isn't already stopped). */
if (proc_flags (pi) & (PR_STOPPED | PR_ISTOP))
{
pi->was_stopped = 1;
proc_prettyprint_why (proc_why (pi), proc_what (pi), 1);
}
else
{
pi->was_stopped = 0;
/* Set the process to run again when we close it. */
if (!proc_set_run_on_last_close (pi))
dead_procinfo (pi, "do_attach: couldn't set RLC.", NOKILL);
/* Now stop the process. */
if (!proc_stop_process (pi))
dead_procinfo (pi, "do_attach: couldn't stop the process.", NOKILL);
pi->ignore_next_sigstop = 1;
}
/* Save some of the /proc state to be restored if we detach. */
if (!proc_get_traced_faults (pi, &pi->saved_fltset))
dead_procinfo (pi, "do_attach: couldn't save traced faults.", NOKILL);
if (!proc_get_traced_signals (pi, &pi->saved_sigset))
dead_procinfo (pi, "do_attach: couldn't save traced signals.", NOKILL);
if (!proc_get_traced_sysentry (pi, pi->saved_entryset))
dead_procinfo (pi, "do_attach: couldn't save traced syscall entries.",
NOKILL);
if (!proc_get_traced_sysexit (pi, pi->saved_exitset))
dead_procinfo (pi, "do_attach: couldn't save traced syscall exits.",
NOKILL);
if (!proc_get_held_signals (pi, &pi->saved_sighold))
dead_procinfo (pi, "do_attach: couldn't save held signals.", NOKILL);
if ((fail = procfs_debug_inferior (pi)) != 0)
dead_procinfo (pi, "do_attach: failed in procfs_debug_inferior", NOKILL);
inf = current_inferior ();
inferior_appeared (inf, pi->pid);
/* Let GDB know that the inferior was attached. */
inf->attach_flag = 1;
/* Create a procinfo for the current lwp. */
lwpid = proc_get_current_thread (pi);
create_procinfo (pi->pid, lwpid);
/* Add it to gdb's thread list. */
ptid = ptid_build (pi->pid, lwpid, 0);
add_thread (ptid);
return ptid;
}
static void
do_detach (int signo)
{
procinfo *pi;
/* Find procinfo for the main process. */
pi = find_procinfo_or_die (ptid_get_pid (inferior_ptid),
0); /* FIXME: threads */
if (signo)
if (!proc_set_current_signal (pi, signo))
proc_warn (pi, "do_detach, set_current_signal", __LINE__);
if (!proc_set_traced_signals (pi, &pi->saved_sigset))
proc_warn (pi, "do_detach, set_traced_signal", __LINE__);
if (!proc_set_traced_faults (pi, &pi->saved_fltset))
proc_warn (pi, "do_detach, set_traced_faults", __LINE__);
if (!proc_set_traced_sysentry (pi, pi->saved_entryset))
proc_warn (pi, "do_detach, set_traced_sysentry", __LINE__);
if (!proc_set_traced_sysexit (pi, pi->saved_exitset))
proc_warn (pi, "do_detach, set_traced_sysexit", __LINE__);
if (!proc_set_held_signals (pi, &pi->saved_sighold))
proc_warn (pi, "do_detach, set_held_signals", __LINE__);
if (signo || (proc_flags (pi) & (PR_STOPPED | PR_ISTOP)))
if (signo || !(pi->was_stopped) ||
query (_("Was stopped when attached, make it runnable again? ")))
{
/* Clear any pending signal. */
if (!proc_clear_current_fault (pi))
proc_warn (pi, "do_detach, clear_current_fault", __LINE__);
if (signo == 0 && !proc_clear_current_signal (pi))
proc_warn (pi, "do_detach, clear_current_signal", __LINE__);
if (!proc_set_run_on_last_close (pi))
proc_warn (pi, "do_detach, set_rlc", __LINE__);
}
destroy_procinfo (pi);
}
/* Fetch register REGNUM from the inferior. If REGNUM is -1, do this
for all registers.
??? Is the following note still relevant? We can't get individual
registers with the PT_GETREGS ptrace(2) request either, yet we
don't bother with caching at all in that case.
NOTE: Since the /proc interface cannot give us individual
registers, we pay no attention to REGNUM, and just fetch them all.
This results in the possibility that we will do unnecessarily many
fetches, since we may be called repeatedly for individual
registers. So we cache the results, and mark the cache invalid
when the process is resumed. */
static void
procfs_fetch_registers (struct target_ops *ops,
struct regcache *regcache, int regnum)
{
gdb_gregset_t *gregs;
procinfo *pi;
ptid_t ptid = regcache_get_ptid (regcache);
int pid = ptid_get_pid (ptid);
int tid = ptid_get_lwp (ptid);
struct gdbarch *gdbarch = regcache->arch ();
pi = find_procinfo_or_die (pid, tid);
if (pi == NULL)
error (_("procfs: fetch_registers failed to find procinfo for %s"),
target_pid_to_str (ptid));
gregs = proc_get_gregs (pi);
if (gregs == NULL)
proc_error (pi, "fetch_registers, get_gregs", __LINE__);
supply_gregset (regcache, (const gdb_gregset_t *) gregs);
if (gdbarch_fp0_regnum (gdbarch) >= 0) /* Do we have an FPU? */
{
gdb_fpregset_t *fpregs;
if ((regnum >= 0 && regnum < gdbarch_fp0_regnum (gdbarch))
|| regnum == gdbarch_pc_regnum (gdbarch)
|| regnum == gdbarch_sp_regnum (gdbarch))
return; /* Not a floating point register. */
fpregs = proc_get_fpregs (pi);
if (fpregs == NULL)
proc_error (pi, "fetch_registers, get_fpregs", __LINE__);
supply_fpregset (regcache, (const gdb_fpregset_t *) fpregs);
}
}
/* Store register REGNUM back into the inferior. If REGNUM is -1, do
this for all registers.
NOTE: Since the /proc interface will not read individual registers,
we will cache these requests until the process is resumed, and only
then write them back to the inferior process.
FIXME: is that a really bad idea? Have to think about cases where
writing one register might affect the value of others, etc. */
static void
procfs_store_registers (struct target_ops *ops,
struct regcache *regcache, int regnum)
{
gdb_gregset_t *gregs;
procinfo *pi;
ptid_t ptid = regcache_get_ptid (regcache);
int pid = ptid_get_pid (ptid);
int tid = ptid_get_lwp (ptid);
struct gdbarch *gdbarch = regcache->arch ();
pi = find_procinfo_or_die (pid, tid);
if (pi == NULL)
error (_("procfs: store_registers: failed to find procinfo for %s"),
target_pid_to_str (ptid));
gregs = proc_get_gregs (pi);
if (gregs == NULL)
proc_error (pi, "store_registers, get_gregs", __LINE__);
fill_gregset (regcache, gregs, regnum);
if (!proc_set_gregs (pi))
proc_error (pi, "store_registers, set_gregs", __LINE__);
if (gdbarch_fp0_regnum (gdbarch) >= 0) /* Do we have an FPU? */
{
gdb_fpregset_t *fpregs;
if ((regnum >= 0 && regnum < gdbarch_fp0_regnum (gdbarch))
|| regnum == gdbarch_pc_regnum (gdbarch)
|| regnum == gdbarch_sp_regnum (gdbarch))
return; /* Not a floating point register. */
fpregs = proc_get_fpregs (pi);
if (fpregs == NULL)
proc_error (pi, "store_registers, get_fpregs", __LINE__);
fill_fpregset (regcache, fpregs, regnum);
if (!proc_set_fpregs (pi))
proc_error (pi, "store_registers, set_fpregs", __LINE__);
}
}
static int
syscall_is_lwp_exit (procinfo *pi, int scall)
{
if (scall == SYS_lwp_exit)
return 1;
return 0;
}
static int
syscall_is_exit (procinfo *pi, int scall)
{
if (scall == SYS_exit)
return 1;
return 0;
}
static int
syscall_is_exec (procinfo *pi, int scall)
{
#ifdef SYS_exec
if (scall == SYS_exec)
return 1;
#endif
if (scall == SYS_execve)
return 1;
return 0;
}
static int
syscall_is_lwp_create (procinfo *pi, int scall)
{
if (scall == SYS_lwp_create)
return 1;
return 0;
}
/* Retrieve the next stop event from the child process. If child has
not stopped yet, wait for it to stop. Translate /proc eventcodes
(or possibly wait eventcodes) into gdb internal event codes.
Returns the id of process (and possibly thread) that incurred the
event. Event codes are returned through a pointer parameter. */
static ptid_t
procfs_wait (struct target_ops *ops,
ptid_t ptid, struct target_waitstatus *status, int options)
{
/* First cut: loosely based on original version 2.1. */
procinfo *pi;
int wstat;
int temp_tid;
ptid_t retval, temp_ptid;
int why, what, flags;
int retry = 0;
wait_again:
retry++;
wstat = 0;
retval = pid_to_ptid (-1);
/* Find procinfo for main process. */
pi = find_procinfo_or_die (ptid_get_pid (inferior_ptid), 0);
if (pi)
{
/* We must assume that the status is stale now... */
pi->status_valid = 0;
pi->gregs_valid = 0;
pi->fpregs_valid = 0;
#if 0 /* just try this out... */
flags = proc_flags (pi);
why = proc_why (pi);
if ((flags & PR_STOPPED) && (why == PR_REQUESTED))
pi->status_valid = 0; /* re-read again, IMMEDIATELY... */
#endif
/* If child is not stopped, wait for it to stop. */
if (!(proc_flags (pi) & (PR_STOPPED | PR_ISTOP)) &&
!proc_wait_for_stop (pi))
{
/* wait_for_stop failed: has the child terminated? */
if (errno == ENOENT)
{
int wait_retval;
/* /proc file not found; presumably child has terminated. */
wait_retval = wait (&wstat); /* "wait" for the child's exit. */
/* Wrong child? */
if (wait_retval != ptid_get_pid (inferior_ptid))
error (_("procfs: couldn't stop "
"process %d: wait returned %d."),
ptid_get_pid (inferior_ptid), wait_retval);
/* FIXME: might I not just use waitpid?
Or try find_procinfo to see if I know about this child? */
retval = pid_to_ptid (wait_retval);
}
else if (errno == EINTR)
goto wait_again;
else
{
/* Unknown error from wait_for_stop. */
proc_error (pi, "target_wait (wait_for_stop)", __LINE__);
}
}
else
{
/* This long block is reached if either:
a) the child was already stopped, or
b) we successfully waited for the child with wait_for_stop.
This block will analyze the /proc status, and translate it
into a waitstatus for GDB.
If we actually had to call wait because the /proc file
is gone (child terminated), then we skip this block,
because we already have a waitstatus. */
flags = proc_flags (pi);
why = proc_why (pi);
what = proc_what (pi);
if (flags & (PR_STOPPED | PR_ISTOP))
{
/* If it's running async (for single_thread control),
set it back to normal again. */
if (flags & PR_ASYNC)
if (!proc_unset_async (pi))
proc_error (pi, "target_wait, unset_async", __LINE__);
if (info_verbose)
proc_prettyprint_why (why, what, 1);
/* The 'pid' we will return to GDB is composed of
the process ID plus the lwp ID. */
retval = ptid_build (pi->pid, proc_get_current_thread (pi), 0);
switch (why) {
case PR_SIGNALLED:
wstat = (what << 8) | 0177;
break;
case PR_SYSENTRY:
if (syscall_is_lwp_exit (pi, what))
{
if (print_thread_events)
printf_unfiltered (_("[%s exited]\n"),
target_pid_to_str (retval));
delete_thread (retval);
status->kind = TARGET_WAITKIND_SPURIOUS;
return retval;
}
else if (syscall_is_exit (pi, what))
{
struct inferior *inf;
/* Handle SYS_exit call only. */
/* Stopped at entry to SYS_exit.
Make it runnable, resume it, then use
the wait system call to get its exit code.
Proc_run_process always clears the current
fault and signal.
Then return its exit status. */
pi->status_valid = 0;
wstat = 0;
/* FIXME: what we should do is return
TARGET_WAITKIND_SPURIOUS. */
if (!proc_run_process (pi, 0, 0))
proc_error (pi, "target_wait, run_process", __LINE__);
inf = find_inferior_pid (pi->pid);
if (inf->attach_flag)
{
/* Don't call wait: simulate waiting for exit,
return a "success" exit code. Bogus: what if
it returns something else? */
wstat = 0;
retval = inferior_ptid; /* ? ? ? */
}
else
{
int temp = wait (&wstat);
/* FIXME: shouldn't I make sure I get the right
event from the right process? If (for
instance) I have killed an earlier inferior
process but failed to clean up after it
somehow, I could get its termination event
here. */
/* If wait returns -1, that's what we return
to GDB. */
if (temp < 0)
retval = pid_to_ptid (temp);
}
}
else
{
printf_filtered (_("procfs: trapped on entry to "));
proc_prettyprint_syscall (proc_what (pi), 0);
printf_filtered ("\n");
long i, nsysargs, *sysargs;
if ((nsysargs = proc_nsysarg (pi)) > 0 &&
(sysargs = proc_sysargs (pi)) != NULL)
{
printf_filtered (_("%ld syscall arguments:\n"),
nsysargs);
for (i = 0; i < nsysargs; i++)
printf_filtered ("#%ld: 0x%08lx\n",
i, sysargs[i]);
}
if (status)
{
/* How to exit gracefully, returning "unknown
event". */
status->kind = TARGET_WAITKIND_SPURIOUS;
return inferior_ptid;
}
else
{
/* How to keep going without returning to wfi: */
target_continue_no_signal (ptid);
goto wait_again;
}
}
break;
case PR_SYSEXIT:
if (syscall_is_exec (pi, what))
{
/* Hopefully this is our own "fork-child" execing
the real child. Hoax this event into a trap, and
GDB will see the child about to execute its start
address. */
wstat = (SIGTRAP << 8) | 0177;
}
else if (syscall_is_lwp_create (pi, what))
{
/* This syscall is somewhat like fork/exec. We
will get the event twice: once for the parent
LWP, and once for the child. We should already
know about the parent LWP, but the child will
be new to us. So, whenever we get this event,
if it represents a new thread, simply add the
thread to the list. */
/* If not in procinfo list, add it. */
temp_tid = proc_get_current_thread (pi);
if (!find_procinfo (pi->pid, temp_tid))
create_procinfo (pi->pid, temp_tid);
temp_ptid = ptid_build (pi->pid, temp_tid, 0);
/* If not in GDB's thread list, add it. */
if (!in_thread_list (temp_ptid))
add_thread (temp_ptid);
/* Return to WFI, but tell it to immediately resume. */
status->kind = TARGET_WAITKIND_SPURIOUS;
return inferior_ptid;
}
else if (syscall_is_lwp_exit (pi, what))
{
if (print_thread_events)
printf_unfiltered (_("[%s exited]\n"),
target_pid_to_str (retval));
delete_thread (retval);
status->kind = TARGET_WAITKIND_SPURIOUS;
return retval;
}
else if (0)
{
/* FIXME: Do we need to handle SYS_sproc,
SYS_fork, or SYS_vfork here? The old procfs
seemed to use this event to handle threads on
older (non-LWP) systems, where I'm assuming
that threads were actually separate processes.
Irix, maybe? Anyway, low priority for now. */
}
else
{
printf_filtered (_("procfs: trapped on exit from "));
proc_prettyprint_syscall (proc_what (pi), 0);
printf_filtered ("\n");
long i, nsysargs, *sysargs;
if ((nsysargs = proc_nsysarg (pi)) > 0 &&
(sysargs = proc_sysargs (pi)) != NULL)
{
printf_filtered (_("%ld syscall arguments:\n"),
nsysargs);
for (i = 0; i < nsysargs; i++)
printf_filtered ("#%ld: 0x%08lx\n",
i, sysargs[i]);
}
status->kind = TARGET_WAITKIND_SPURIOUS;
return inferior_ptid;
}
break;
case PR_REQUESTED:
#if 0 /* FIXME */
wstat = (SIGSTOP << 8) | 0177;
break;
#else
if (retry < 5)
{
printf_filtered (_("Retry #%d:\n"), retry);
pi->status_valid = 0;
goto wait_again;
}
else
{
/* If not in procinfo list, add it. */
temp_tid = proc_get_current_thread (pi);
if (!find_procinfo (pi->pid, temp_tid))
create_procinfo (pi->pid, temp_tid);
/* If not in GDB's thread list, add it. */
temp_ptid = ptid_build (pi->pid, temp_tid, 0);
if (!in_thread_list (temp_ptid))
add_thread (temp_ptid);
status->kind = TARGET_WAITKIND_STOPPED;
status->value.sig = GDB_SIGNAL_0;
return retval;
}
#endif
case PR_JOBCONTROL:
wstat = (what << 8) | 0177;
break;
case PR_FAULTED:
switch (what) {
case FLTWATCH:
wstat = (SIGTRAP << 8) | 0177;
break;
/* FIXME: use si_signo where possible. */
case FLTPRIV:
case FLTILL:
wstat = (SIGILL << 8) | 0177;
break;
case FLTBPT:
case FLTTRACE:
wstat = (SIGTRAP << 8) | 0177;
break;
case FLTSTACK:
case FLTACCESS:
case FLTBOUNDS:
wstat = (SIGSEGV << 8) | 0177;
break;
case FLTIOVF:
case FLTIZDIV:
case FLTFPE:
wstat = (SIGFPE << 8) | 0177;
break;
case FLTPAGE: /* Recoverable page fault */
default: /* FIXME: use si_signo if possible for
fault. */
retval = pid_to_ptid (-1);
printf_filtered ("procfs:%d -- ", __LINE__);
printf_filtered (_("child stopped for unknown reason:\n"));
proc_prettyprint_why (why, what, 1);
error (_("... giving up..."));
break;
}
break; /* case PR_FAULTED: */
default: /* switch (why) unmatched */
printf_filtered ("procfs:%d -- ", __LINE__);
printf_filtered (_("child stopped for unknown reason:\n"));
proc_prettyprint_why (why, what, 1);
error (_("... giving up..."));
break;
}
/* Got this far without error: If retval isn't in the
threads database, add it. */
if (ptid_get_pid (retval) > 0 &&
!ptid_equal (retval, inferior_ptid) &&
!in_thread_list (retval))
{
/* We have a new thread. We need to add it both to
GDB's list and to our own. If we don't create a
procinfo, resume may be unhappy later. */
add_thread (retval);
if (find_procinfo (ptid_get_pid (retval),
ptid_get_lwp (retval)) == NULL)
create_procinfo (ptid_get_pid (retval),
ptid_get_lwp (retval));
}
}
else /* Flags do not indicate STOPPED. */
{
/* surely this can't happen... */
printf_filtered ("procfs:%d -- process not stopped.\n",
__LINE__);
proc_prettyprint_flags (flags, 1);
error (_("procfs: ...giving up..."));
}
}
if (status)
store_waitstatus (status, wstat);
}
return retval;
}
/* Perform a partial transfer to/from the specified object. For
memory transfers, fall back to the old memory xfer functions. */
static enum target_xfer_status
procfs_xfer_partial (struct target_ops *ops, enum target_object object,
const char *annex, gdb_byte *readbuf,
const gdb_byte *writebuf, ULONGEST offset, ULONGEST len,
ULONGEST *xfered_len)
{
switch (object)
{
case TARGET_OBJECT_MEMORY:
return procfs_xfer_memory (readbuf, writebuf, offset, len, xfered_len);
case TARGET_OBJECT_AUXV:
return memory_xfer_auxv (ops, object, annex, readbuf, writebuf,
offset, len, xfered_len);
default:
return ops->beneath->to_xfer_partial (ops->beneath, object, annex,
readbuf, writebuf, offset, len,
xfered_len);
}
}
/* Helper for procfs_xfer_partial that handles memory transfers.
Arguments are like target_xfer_partial. */
static enum target_xfer_status
procfs_xfer_memory (gdb_byte *readbuf, const gdb_byte *writebuf,
ULONGEST memaddr, ULONGEST len, ULONGEST *xfered_len)
{
procinfo *pi;
int nbytes;
/* Find procinfo for main process. */
pi = find_procinfo_or_die (ptid_get_pid (inferior_ptid), 0);
if (pi->as_fd == 0 &&
open_procinfo_files (pi, FD_AS) == 0)
{
proc_warn (pi, "xfer_memory, open_proc_files", __LINE__);
return TARGET_XFER_E_IO;
}
if (lseek (pi->as_fd, (off_t) memaddr, SEEK_SET) != (off_t) memaddr)
return TARGET_XFER_E_IO;
if (writebuf != NULL)
{
PROCFS_NOTE ("write memory:\n");
nbytes = write (pi->as_fd, writebuf, len);
}
else
{
PROCFS_NOTE ("read memory:\n");
nbytes = read (pi->as_fd, readbuf, len);
}
if (nbytes <= 0)
return TARGET_XFER_E_IO;
*xfered_len = nbytes;
return TARGET_XFER_OK;
}
/* Called by target_resume before making child runnable. Mark cached
registers and status's invalid. If there are "dirty" caches that
need to be written back to the child process, do that.
File descriptors are also cached. As they are a limited resource,
we cannot hold onto them indefinitely. However, as they are
expensive to open, we don't want to throw them away
indescriminately either. As a compromise, we will keep the file
descriptors for the parent process, but discard any file
descriptors we may have accumulated for the threads.
As this function is called by iterate_over_threads, it always
returns zero (so that iterate_over_threads will keep
iterating). */
static int
invalidate_cache (procinfo *parent, procinfo *pi, void *ptr)
{
/* About to run the child; invalidate caches and do any other
cleanup. */
#if 0
if (pi->gregs_dirty)
if (parent == NULL ||
proc_get_current_thread (parent) != pi->tid)
if (!proc_set_gregs (pi)) /* flush gregs cache */
proc_warn (pi, "target_resume, set_gregs",
__LINE__);
if (gdbarch_fp0_regnum (target_gdbarch ()) >= 0)
if (pi->fpregs_dirty)
if (parent == NULL ||
proc_get_current_thread (parent) != pi->tid)
if (!proc_set_fpregs (pi)) /* flush fpregs cache */
proc_warn (pi, "target_resume, set_fpregs",
__LINE__);
#endif
if (parent != NULL)
{
/* The presence of a parent indicates that this is an LWP.
Close any file descriptors that it might have open.
We don't do this to the master (parent) procinfo. */
close_procinfo_files (pi);
}
pi->gregs_valid = 0;
pi->fpregs_valid = 0;
#if 0
pi->gregs_dirty = 0;
pi->fpregs_dirty = 0;
#endif
pi->status_valid = 0;
pi->threads_valid = 0;
return 0;
}
#if 0
/* A callback function for iterate_over_threads. Find the
asynchronous signal thread, and make it runnable. See if that
helps matters any. */
static int
make_signal_thread_runnable (procinfo *process, procinfo *pi, void *ptr)
{
#ifdef PR_ASLWP
if (proc_flags (pi) & PR_ASLWP)
{
if (!proc_run_process (pi, 0, -1))
proc_error (pi, "make_signal_thread_runnable", __LINE__);
return 1;
}
#endif
return 0;
}
#endif
/* Make the child process runnable. Normally we will then call
procfs_wait and wait for it to stop again (unless gdb is async).
If STEP is true, then arrange for the child to stop again after
executing a single instruction. If SIGNO is zero, then cancel any
pending signal; if non-zero, then arrange for the indicated signal
to be delivered to the child when it runs. If PID is -1, then
allow any child thread to run; if non-zero, then allow only the
indicated thread to run. (not implemented yet). */
static void
procfs_resume (struct target_ops *ops,
ptid_t ptid, int step, enum gdb_signal signo)
{
procinfo *pi, *thread;
int native_signo;
/* 2.1:
prrun.prflags |= PRSVADDR;
prrun.pr_vaddr = $PC; set resume address
prrun.prflags |= PRSTRACE; trace signals in pr_trace (all)
prrun.prflags |= PRSFAULT; trace faults in pr_fault (all but PAGE)
prrun.prflags |= PRCFAULT; clear current fault.
PRSTRACE and PRSFAULT can be done by other means
(proc_trace_signals, proc_trace_faults)
PRSVADDR is unnecessary.
PRCFAULT may be replaced by a PIOCCFAULT call (proc_clear_current_fault)
This basically leaves PRSTEP and PRCSIG.
PRCSIG is like PIOCSSIG (proc_clear_current_signal).
So basically PR_STEP is the sole argument that must be passed
to proc_run_process (for use in the prrun struct by ioctl). */
/* Find procinfo for main process. */
pi = find_procinfo_or_die (ptid_get_pid (inferior_ptid), 0);
/* First cut: ignore pid argument. */
errno = 0;
/* Convert signal to host numbering. */
if (signo == 0 ||
(signo == GDB_SIGNAL_STOP && pi->ignore_next_sigstop))
native_signo = 0;
else
native_signo = gdb_signal_to_host (signo);
pi->ignore_next_sigstop = 0;
/* Running the process voids all cached registers and status. */
/* Void the threads' caches first. */
proc_iterate_over_threads (pi, invalidate_cache, NULL);
/* Void the process procinfo's caches. */
invalidate_cache (NULL, pi, NULL);
if (ptid_get_pid (ptid) != -1)
{
/* Resume a specific thread, presumably suppressing the
others. */
thread = find_procinfo (ptid_get_pid (ptid), ptid_get_lwp (ptid));
if (thread != NULL)
{
if (thread->tid != 0)
{
/* We're to resume a specific thread, and not the
others. Set the child process's PR_ASYNC flag. */
if (!proc_set_async (pi))
proc_error (pi, "target_resume, set_async", __LINE__);
#if 0
proc_iterate_over_threads (pi,
make_signal_thread_runnable,
NULL);
#endif
pi = thread; /* Substitute the thread's procinfo
for run. */
}
}
}
if (!proc_run_process (pi, step, native_signo))
{
if (errno == EBUSY)
warning (_("resume: target already running. "
"Pretend to resume, and hope for the best!"));
else
proc_error (pi, "target_resume", __LINE__);
}
}
/* Set up to trace signals in the child process. */
static void
procfs_pass_signals (struct target_ops *self,
int numsigs, unsigned char *pass_signals)
{
sigset_t signals;
procinfo *pi = find_procinfo_or_die (ptid_get_pid (inferior_ptid), 0);
int signo;
prfillset (&signals);
for (signo = 0; signo < NSIG; signo++)
{
int target_signo = gdb_signal_from_host (signo);
if (target_signo < numsigs && pass_signals[target_signo])
prdelset (&signals, signo);
}
if (!proc_set_traced_signals (pi, &signals))
proc_error (pi, "pass_signals", __LINE__);
}
/* Print status information about the child process. */
static void
procfs_files_info (struct target_ops *ignore)
{
struct inferior *inf = current_inferior ();
printf_filtered (_("\tUsing the running image of %s %s via /proc.\n"),
inf->attach_flag? "attached": "child",
target_pid_to_str (inferior_ptid));
}
/* Stop the child process asynchronously, as when the gdb user types
control-c or presses a "stop" button. Works by sending
kill(SIGINT) to the child's process group. */
static void
procfs_interrupt (struct target_ops *self, ptid_t ptid)
{
kill (-inferior_process_group (), SIGINT);
}
/* Make it die. Wait for it to die. Clean up after it. Note: this
should only be applied to the real process, not to an LWP, because
of the check for parent-process. If we need this to work for an
LWP, it needs some more logic. */
static void
unconditionally_kill_inferior (procinfo *pi)
{
int parent_pid;
parent_pid = proc_parent_pid (pi);
if (!proc_kill (pi, SIGKILL))
proc_error (pi, "unconditionally_kill, proc_kill", __LINE__);
destroy_procinfo (pi);
/* If pi is GDB's child, wait for it to die. */
if (parent_pid == getpid ())
/* FIXME: should we use waitpid to make sure we get the right event?
Should we check the returned event? */
{
#if 0
int status, ret;
ret = waitpid (pi->pid, &status, 0);
#else
wait (NULL);
#endif
}
}
/* We're done debugging it, and we want it to go away. Then we want
GDB to forget all about it. */
static void
procfs_kill_inferior (struct target_ops *ops)
{
if (!ptid_equal (inferior_ptid, null_ptid)) /* ? */
{
/* Find procinfo for main process. */
procinfo *pi = find_procinfo (ptid_get_pid (inferior_ptid), 0);
if (pi)
unconditionally_kill_inferior (pi);
target_mourn_inferior (inferior_ptid);
}
}
/* Forget we ever debugged this thing! */
static void
procfs_mourn_inferior (struct target_ops *ops)
{
procinfo *pi;
if (!ptid_equal (inferior_ptid, null_ptid))
{
/* Find procinfo for main process. */
pi = find_procinfo (ptid_get_pid (inferior_ptid), 0);
if (pi)
destroy_procinfo (pi);
}
generic_mourn_inferior ();
inf_child_maybe_unpush_target (ops);
}
/* When GDB forks to create a runnable inferior process, this function
is called on the parent side of the fork. It's job is to do
whatever is necessary to make the child ready to be debugged, and
then wait for the child to synchronize. */
static void
procfs_init_inferior (struct target_ops *ops, int pid)
{
procinfo *pi;
sigset_t signals;
int fail;
int lwpid;
/* This routine called on the parent side (GDB side)
after GDB forks the inferior. */
if (!target_is_pushed (ops))
push_target (ops);
if ((pi = create_procinfo (pid, 0)) == NULL)
perror (_("procfs: out of memory in 'init_inferior'"));
if (!open_procinfo_files (pi, FD_CTL))
proc_error (pi, "init_inferior, open_proc_files", __LINE__);
/*
xmalloc // done
open_procinfo_files // done
link list // done
prfillset (trace)
procfs_notice_signals
prfillset (fault)
prdelset (FLTPAGE)
PIOCWSTOP
PIOCSFAULT
*/
/* If not stopped yet, wait for it to stop. */
if (!(proc_flags (pi) & PR_STOPPED) &&
!(proc_wait_for_stop (pi)))
dead_procinfo (pi, "init_inferior: wait_for_stop failed", KILL);
/* Save some of the /proc state to be restored if we detach. */
/* FIXME: Why? In case another debugger was debugging it?
We're it's parent, for Ghu's sake! */
if (!proc_get_traced_signals (pi, &pi->saved_sigset))
proc_error (pi, "init_inferior, get_traced_signals", __LINE__);
if (!proc_get_held_signals (pi, &pi->saved_sighold))
proc_error (pi, "init_inferior, get_held_signals", __LINE__);
if (!proc_get_traced_faults (pi, &pi->saved_fltset))
proc_error (pi, "init_inferior, get_traced_faults", __LINE__);
if (!proc_get_traced_sysentry (pi, pi->saved_entryset))
proc_error (pi, "init_inferior, get_traced_sysentry", __LINE__);
if (!proc_get_traced_sysexit (pi, pi->saved_exitset))
proc_error (pi, "init_inferior, get_traced_sysexit", __LINE__);
if ((fail = procfs_debug_inferior (pi)) != 0)
proc_error (pi, "init_inferior (procfs_debug_inferior)", fail);
/* FIXME: logically, we should really be turning OFF run-on-last-close,
and possibly even turning ON kill-on-last-close at this point. But
I can't make that change without careful testing which I don't have
time to do right now... */
/* Turn on run-on-last-close flag so that the child
will die if GDB goes away for some reason. */
if (!proc_set_run_on_last_close (pi))
proc_error (pi, "init_inferior, set_RLC", __LINE__);
/* We now have have access to the lwpid of the main thread/lwp. */
lwpid = proc_get_current_thread (pi);
/* Create a procinfo for the main lwp. */
create_procinfo (pid, lwpid);
/* We already have a main thread registered in the thread table at
this point, but it didn't have any lwp info yet. Notify the core
about it. This changes inferior_ptid as well. */
thread_change_ptid (pid_to_ptid (pid),
ptid_build (pid, lwpid, 0));
gdb_startup_inferior (pid, START_INFERIOR_TRAPS_EXPECTED);
}
/* When GDB forks to create a new process, this function is called on
the child side of the fork before GDB exec's the user program. Its
job is to make the child minimally debuggable, so that the parent
GDB process can connect to the child and take over. This function
should do only the minimum to make that possible, and to
synchronize with the parent process. The parent process should
take care of the details. */
static void
procfs_set_exec_trap (void)
{
/* This routine called on the child side (inferior side)
after GDB forks the inferior. It must use only local variables,
because it may be sharing data space with its parent. */
procinfo *pi;
sysset_t *exitset;
if ((pi = create_procinfo (getpid (), 0)) == NULL)
perror_with_name (_("procfs: create_procinfo failed in child."));
if (open_procinfo_files (pi, FD_CTL) == 0)
{
proc_warn (pi, "set_exec_trap, open_proc_files", __LINE__);
gdb_flush (gdb_stderr);
/* No need to call "dead_procinfo", because we're going to
exit. */
_exit (127);
}
/* Method for tracing exec syscalls. */
/* 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. */
exitset = sysset_t_alloc (pi);
premptyset (exitset);
#ifdef SYS_exec
praddset (exitset, SYS_exec);
#endif
praddset (exitset, SYS_execve);
if (!proc_set_traced_sysexit (pi, exitset))
{
proc_warn (pi, "set_exec_trap, set_traced_sysexit", __LINE__);
gdb_flush (gdb_stderr);
_exit (127);
}
/* FIXME: should this be done in the parent instead? */
/* Turn off inherit on fork flag so that all grand-children
of gdb start with tracing flags cleared. */
if (!proc_unset_inherit_on_fork (pi))
proc_warn (pi, "set_exec_trap, unset_inherit", __LINE__);
/* Turn off run on last close flag, so that the child process
cannot run away just because we close our handle on it.
We want it to wait for the parent to attach. */
if (!proc_unset_run_on_last_close (pi))
proc_warn (pi, "set_exec_trap, unset_RLC", __LINE__);
/* FIXME: No need to destroy the procinfo --
we have our own address space, and we're about to do an exec! */
/*destroy_procinfo (pi);*/
}
/* This function is called BEFORE gdb forks the inferior process. Its
only real responsibility is to set things up for the fork, and tell
GDB which two functions to call after the fork (one for the parent,
and one for the child).
This function does a complicated search for a unix shell program,
which it then uses to parse arguments and environment variables to
be sent to the child. I wonder whether this code could not be
abstracted out and shared with other unix targets such as
inf-ptrace? */
static void
procfs_create_inferior (struct target_ops *ops, const char *exec_file,
const std::string &allargs, char **env, int from_tty)
{
char *shell_file = getenv ("SHELL");
char *tryname;
int pid;
if (shell_file != NULL && strchr (shell_file, '/') == NULL)
{
/* We will be looking down the PATH to find shell_file. If we
just do this the normal way (via execlp, which operates by
attempting an exec for each element of the PATH until it
finds one which succeeds), then there will be an exec for
each failed attempt, each of which will cause a PR_SYSEXIT
stop, and we won't know how to distinguish the PR_SYSEXIT's
for these failed execs with the ones for successful execs
(whether the exec has succeeded is stored at that time in the
carry bit or some such architecture-specific and
non-ABI-specified place).
So I can't think of anything better than to search the PATH
now. This has several disadvantages: (1) There is a race
condition; if we find a file now and it is deleted before we
exec it, we lose, even if the deletion leaves a valid file
further down in the PATH, (2) there is no way to know exactly
what an executable (in the sense of "capable of being
exec'd") file is. Using access() loses because it may lose
if the caller is the superuser; failing to use it loses if
there are ACLs or some such. */
const char *p;
const char *p1;
/* FIXME-maybe: might want "set path" command so user can change what
path is used from within GDB. */
const char *path = getenv ("PATH");
int len;
struct stat statbuf;
if (path == NULL)
path = "/bin:/usr/bin";
tryname = (char *) alloca (strlen (path) + strlen (shell_file) + 2);
for (p = path; p != NULL; p = p1 ? p1 + 1: NULL)
{
p1 = strchr (p, ':');
if (p1 != NULL)
len = p1 - p;
else
len = strlen (p);
strncpy (tryname, p, len);
tryname[len] = '\0';
strcat (tryname, "/");
strcat (tryname, shell_file);
if (access (tryname, X_OK) < 0)
continue;
if (stat (tryname, &statbuf) < 0)
continue;
if (!S_ISREG (statbuf.st_mode))
/* We certainly need to reject directories. I'm not quite
as sure about FIFOs, sockets, etc., but I kind of doubt
that people want to exec() these things. */
continue;
break;
}
if (p == NULL)
/* Not found. This must be an error rather than merely passing
the file to execlp(), because execlp() would try all the
exec()s, causing GDB to get confused. */
error (_("procfs:%d -- Can't find shell %s in PATH"),
__LINE__, shell_file);
shell_file = tryname;
}
pid = fork_inferior (exec_file, allargs, env, procfs_set_exec_trap,
NULL, NULL, shell_file, NULL);
/* We have something that executes now. We'll be running through
the shell at this point (if startup-with-shell is true), but the
pid shouldn't change. */
add_thread_silent (pid_to_ptid (pid));
procfs_init_inferior (ops, pid);
}
/* An observer for the "inferior_created" event. */
static void
procfs_inferior_created (struct target_ops *ops, int from_tty)
{
}
/* Callback for update_thread_list. Calls "add_thread". */
static int
procfs_notice_thread (procinfo *pi, procinfo *thread, void *ptr)
{
ptid_t gdb_threadid = ptid_build (pi->pid, thread->tid, 0);
if (!in_thread_list (gdb_threadid) || is_exited (gdb_threadid))
add_thread (gdb_threadid);
return 0;
}
/* Query all the threads that the target knows about, and give them
back to GDB to add to its list. */
static void
procfs_update_thread_list (struct target_ops *ops)
{
procinfo *pi;
prune_threads ();
/* Find procinfo for main process. */
pi = find_procinfo_or_die (ptid_get_pid (inferior_ptid), 0);
proc_update_threads (pi);
proc_iterate_over_threads (pi, procfs_notice_thread, NULL);
}
/* Return true if the thread is still 'alive'. This guy doesn't
really seem to be doing his job. Got to investigate how to tell
when a thread is really gone. */
static int
procfs_thread_alive (struct target_ops *ops, ptid_t ptid)
{
int proc, thread;
procinfo *pi;
proc = ptid_get_pid (ptid);
thread = ptid_get_lwp (ptid);
/* If I don't know it, it ain't alive! */
if ((pi = find_procinfo (proc, thread)) == NULL)
return 0;
/* If I can't get its status, it ain't alive!
What's more, I need to forget about it! */
if (!proc_get_status (pi))
{
destroy_procinfo (pi);
return 0;
}
/* I couldn't have got its status if it weren't alive, so it's
alive. */
return 1;
}
/* Convert PTID to a string. Returns the string in a static
buffer. */
static const char *
procfs_pid_to_str (struct target_ops *ops, ptid_t ptid)
{
static char buf[80];
if (ptid_get_lwp (ptid) == 0)
sprintf (buf, "process %d", ptid_get_pid (ptid));
else
sprintf (buf, "LWP %ld", ptid_get_lwp (ptid));
return buf;
}
/* Insert a watchpoint. */
static int
procfs_set_watchpoint (ptid_t ptid, CORE_ADDR addr, int len, int rwflag,
int after)
{
int pflags = 0;
procinfo *pi;
pi = find_procinfo_or_die (ptid_get_pid (ptid) == -1 ?
ptid_get_pid (inferior_ptid) : ptid_get_pid (ptid),
0);
/* Translate from GDB's flags to /proc's. */
if (len > 0) /* len == 0 means delete watchpoint. */
{
switch (rwflag) { /* FIXME: need an enum! */
case hw_write: /* default watchpoint (write) */
pflags = WRITE_WATCHFLAG;
break;
case hw_read: /* read watchpoint */
pflags = READ_WATCHFLAG;
break;
case hw_access: /* access watchpoint */
pflags = READ_WATCHFLAG | WRITE_WATCHFLAG;
break;
case hw_execute: /* execution HW breakpoint */
pflags = EXEC_WATCHFLAG;
break;
default: /* Something weird. Return error. */
return -1;
}
if (after) /* Stop after r/w access is completed. */
pflags |= AFTER_WATCHFLAG;
}
if (!proc_set_watchpoint (pi, addr, len, pflags))
{
if (errno == E2BIG) /* Typical error for no resources. */
return -1; /* fail */
/* GDB may try to remove the same watchpoint twice.
If a remove request returns no match, don't error. */
if (errno == ESRCH && len == 0)
return 0; /* ignore */
proc_error (pi, "set_watchpoint", __LINE__);
}
return 0;
}
/* Return non-zero if we can set a hardware watchpoint of type TYPE. TYPE
is one of bp_hardware_watchpoint, bp_read_watchpoint, bp_write_watchpoint,
or bp_hardware_watchpoint. CNT is the number of watchpoints used so
far.
Note: procfs_can_use_hw_breakpoint() is not yet used by all
procfs.c targets due to the fact that some of them still define
target_can_use_hardware_watchpoint. */
static int
procfs_can_use_hw_breakpoint (struct target_ops *self,
enum bptype type,
int cnt, int othertype)
{
/* Due to the way that proc_set_watchpoint() is implemented, host
and target pointers must be of the same size. If they are not,
we can't use hardware watchpoints. This limitation is due to the
fact that proc_set_watchpoint() calls
procfs_address_to_host_pointer(); a close inspection of
procfs_address_to_host_pointer will reveal that an internal error
will be generated when the host and target pointer sizes are
different. */
struct type *ptr_type = builtin_type (target_gdbarch ())->builtin_data_ptr;
if (sizeof (void *) != TYPE_LENGTH (ptr_type))
return 0;
/* Other tests here??? */
return 1;
}
/* Returns non-zero if process is stopped on a hardware watchpoint
fault, else returns zero. */
static int
procfs_stopped_by_watchpoint (struct target_ops *ops)
{
procinfo *pi;
pi = find_procinfo_or_die (ptid_get_pid (inferior_ptid), 0);
if (proc_flags (pi) & (PR_STOPPED | PR_ISTOP))
{
if (proc_why (pi) == PR_FAULTED)
{
if (proc_what (pi) == FLTWATCH)
return 1;
}
}
return 0;
}
/* Returns 1 if the OS knows the position of the triggered watchpoint,
and sets *ADDR to that address. Returns 0 if OS cannot report that
address. This function is only called if
procfs_stopped_by_watchpoint returned 1, thus no further checks are
done. The function also assumes that ADDR is not NULL. */
static int
procfs_stopped_data_address (struct target_ops *targ, CORE_ADDR *addr)
{
procinfo *pi;
pi = find_procinfo_or_die (ptid_get_pid (inferior_ptid), 0);
return proc_watchpoint_address (pi, addr);
}
static int
procfs_insert_watchpoint (struct target_ops *self,
CORE_ADDR addr, int len,
enum target_hw_bp_type type,
struct expression *cond)
{
if (!target_have_steppable_watchpoint
&& !gdbarch_have_nonsteppable_watchpoint (target_gdbarch ()))
{
/* When a hardware watchpoint fires off the PC will be left at
the instruction following the one which caused the
watchpoint. It will *NOT* be necessary for GDB to step over
the watchpoint. */
return procfs_set_watchpoint (inferior_ptid, addr, len, type, 1);
}
else
{
/* When a hardware watchpoint fires off the PC will be left at
the instruction which caused the watchpoint. It will be
necessary for GDB to step over the watchpoint. */
return procfs_set_watchpoint (inferior_ptid, addr, len, type, 0);
}
}
static int
procfs_remove_watchpoint (struct target_ops *self,
CORE_ADDR addr, int len,
enum target_hw_bp_type type,
struct expression *cond)
{
return procfs_set_watchpoint (inferior_ptid, addr, 0, 0, 0);
}
static int
procfs_region_ok_for_hw_watchpoint (struct target_ops *self,
CORE_ADDR addr, int len)
{
/* The man page for proc(4) on Solaris 2.6 and up says that the
system can support "thousands" of hardware watchpoints, but gives
no method for finding out how many; It doesn't say anything about
the allowed size for the watched area either. So we just tell
GDB 'yes'. */
return 1;
}
void
procfs_use_watchpoints (struct target_ops *t)
{
t->to_stopped_by_watchpoint = procfs_stopped_by_watchpoint;
t->to_insert_watchpoint = procfs_insert_watchpoint;
t->to_remove_watchpoint = procfs_remove_watchpoint;
t->to_region_ok_for_hw_watchpoint = procfs_region_ok_for_hw_watchpoint;
t->to_can_use_hw_breakpoint = procfs_can_use_hw_breakpoint;
t->to_stopped_data_address = procfs_stopped_data_address;
}
/* Memory Mappings Functions: */
/* Call a callback function once for each mapping, passing it the
mapping, an optional secondary callback function, and some optional
opaque data. Quit and return the first non-zero value returned
from the callback.
PI is the procinfo struct for the process to be mapped. FUNC is
the callback function to be called by this iterator. DATA is the
optional opaque data to be passed to the callback function.
CHILD_FUNC is the optional secondary function pointer to be passed
to the child function. Returns the first non-zero return value
from the callback function, or zero. */
static int
iterate_over_mappings (procinfo *pi, find_memory_region_ftype child_func,
void *data,
int (*func) (struct prmap *map,
find_memory_region_ftype child_func,
void *data))
{
char pathname[MAX_PROC_NAME_SIZE];
struct prmap *prmaps;
struct prmap *prmap;
int funcstat;
int map_fd;
int nmap;
struct cleanup *cleanups = make_cleanup (null_cleanup, NULL);
struct stat sbuf;
/* Get the number of mappings, allocate space,
and read the mappings into prmaps. */
/* Open map fd. */
sprintf (pathname, "/proc/%d/map", pi->pid);
if ((map_fd = open (pathname, O_RDONLY)) < 0)
proc_error (pi, "iterate_over_mappings (open)", __LINE__);
/* Make sure it gets closed again. */
make_cleanup_close (map_fd);
/* Use stat to determine the file size, and compute
the number of prmap_t objects it contains. */
if (fstat (map_fd, &sbuf) != 0)
proc_error (pi, "iterate_over_mappings (fstat)", __LINE__);
nmap = sbuf.st_size / sizeof (prmap_t);
prmaps = (struct prmap *) alloca ((nmap + 1) * sizeof (*prmaps));
if (read (map_fd, (char *) prmaps, nmap * sizeof (*prmaps))
!= (nmap * sizeof (*prmaps)))
proc_error (pi, "iterate_over_mappings (read)", __LINE__);
for (prmap = prmaps; nmap > 0; prmap++, nmap--)
if ((funcstat = (*func) (prmap, child_func, data)) != 0)
{
do_cleanups (cleanups);
return funcstat;
}
do_cleanups (cleanups);
return 0;
}
/* Implements the to_find_memory_regions method. Calls an external
function for each memory region.
Returns the integer value returned by the callback. */
static int
find_memory_regions_callback (struct prmap *map,
find_memory_region_ftype func, void *data)
{
return (*func) ((CORE_ADDR) map->pr_vaddr,
map->pr_size,
(map->pr_mflags & MA_READ) != 0,
(map->pr_mflags & MA_WRITE) != 0,
(map->pr_mflags & MA_EXEC) != 0,
1, /* MODIFIED is unknown, pass it as true. */
data);
}
/* External interface. Calls a callback function once for each
mapped memory region in the child process, passing as arguments:
CORE_ADDR virtual_address,
unsigned long size,
int read, TRUE if region is readable by the child
int write, TRUE if region is writable by the child
int execute TRUE if region is executable by the child.
Stops iterating and returns the first non-zero value returned by
the callback. */
static int
proc_find_memory_regions (struct target_ops *self,
find_memory_region_ftype func, void *data)
{
procinfo *pi = find_procinfo_or_die (ptid_get_pid (inferior_ptid), 0);
return iterate_over_mappings (pi, func, data,
find_memory_regions_callback);
}
/* Returns an ascii representation of a memory mapping's flags. */
static char *
mappingflags (long flags)
{
static char asciiflags[8];
strcpy (asciiflags, "-------");
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);
}
/* Callback function, does the actual work for 'info proc
mappings'. */
static int
info_mappings_callback (struct prmap *map, find_memory_region_ftype ignore,
void *unused)
{
unsigned int pr_off;
pr_off = (unsigned int) map->pr_offset;
if (gdbarch_addr_bit (target_gdbarch ()) == 32)
printf_filtered ("\t%#10lx %#10lx %#10lx %#10x %7s\n",
(unsigned long) map->pr_vaddr,
(unsigned long) map->pr_vaddr + map->pr_size - 1,
(unsigned long) map->pr_size,
pr_off,
mappingflags (map->pr_mflags));
else
printf_filtered (" %#18lx %#18lx %#10lx %#10x %7s\n",
(unsigned long) map->pr_vaddr,
(unsigned long) map->pr_vaddr + map->pr_size - 1,
(unsigned long) map->pr_size,
pr_off,
mappingflags (map->pr_mflags));
return 0;
}
/* Implement the "info proc mappings" subcommand. */
static void
info_proc_mappings (procinfo *pi, int summary)
{
if (summary)
return; /* No output for summary mode. */
printf_filtered (_("Mapped address spaces:\n\n"));
if (gdbarch_ptr_bit (target_gdbarch ()) == 32)
printf_filtered ("\t%10s %10s %10s %10s %7s\n",
"Start Addr",
" End Addr",
" Size",
" Offset",
"Flags");
else
printf_filtered (" %18s %18s %10s %10s %7s\n",
"Start Addr",
" End Addr",
" Size",
" Offset",
"Flags");
iterate_over_mappings (pi, NULL, NULL, info_mappings_callback);
printf_filtered ("\n");
}
/* Implement the "info proc" command. */
static void
procfs_info_proc (struct target_ops *ops, const char *args,
enum info_proc_what what)
{
struct cleanup *old_chain;
procinfo *process = NULL;
procinfo *thread = NULL;
char *tmp = NULL;
int pid = 0;
int tid = 0;
int mappings = 0;
switch (what)
{
case IP_MINIMAL:
break;
case IP_MAPPINGS:
case IP_ALL:
mappings = 1;
break;
default:
error (_("Not supported on this target."));
}
old_chain = make_cleanup (null_cleanup, 0);
gdb_argv built_argv (args);
for (char *arg : built_argv)
{
if (isdigit (arg[0]))
{
pid = strtoul (arg, &tmp, 10);
if (*tmp == '/')
tid = strtoul (++tmp, NULL, 10);
}
else if (arg[0] == '/')
{
tid = strtoul (arg + 1, NULL, 10);
}
}
if (pid == 0)
pid = ptid_get_pid (inferior_ptid);
if (pid == 0)
error (_("No current process: you must name one."));
else
{
/* Have pid, will travel.
First see if it's a process we're already debugging. */
process = find_procinfo (pid, 0);
if (process == NULL)
{
/* No. So open a procinfo for it, but
remember to close it again when finished. */
process = create_procinfo (pid, 0);
make_cleanup (do_destroy_procinfo_cleanup, process);
if (!open_procinfo_files (process, FD_CTL))
proc_error (process, "info proc, open_procinfo_files", __LINE__);
}
}
if (tid != 0)
thread = create_procinfo (pid, tid);
if (process)
{
printf_filtered (_("process %d flags:\n"), process->pid);
proc_prettyprint_flags (proc_flags (process), 1);
if (proc_flags (process) & (PR_STOPPED | PR_ISTOP))
proc_prettyprint_why (proc_why (process), proc_what (process), 1);
if (proc_get_nthreads (process) > 1)
printf_filtered ("Process has %d threads.\n",
proc_get_nthreads (process));
}
if (thread)
{
printf_filtered (_("thread %d flags:\n"), thread->tid);
proc_prettyprint_flags (proc_flags (thread), 1);
if (proc_flags (thread) & (PR_STOPPED | PR_ISTOP))
proc_prettyprint_why (proc_why (thread), proc_what (thread), 1);
}
if (mappings)
{
info_proc_mappings (process, 0);
}
do_cleanups (old_chain);
}
/* Modify the status of the system call identified by SYSCALLNUM in
the set of syscalls that are currently traced/debugged.
If ENTRY_OR_EXIT is set to PR_SYSENTRY, then the entry syscalls set
will be updated. Otherwise, the exit syscalls set will be updated.
If MODE is FLAG_SET, then traces will be enabled. Otherwise, they
will be disabled. */
static void
proc_trace_syscalls_1 (procinfo *pi, int syscallnum, int entry_or_exit,
int mode, int from_tty)
{
sysset_t *sysset;
if (entry_or_exit == PR_SYSENTRY)
sysset = proc_get_traced_sysentry (pi, NULL);
else
sysset = proc_get_traced_sysexit (pi, NULL);
if (sysset == NULL)
proc_error (pi, "proc-trace, get_traced_sysset", __LINE__);
if (mode == FLAG_SET)
praddset (sysset, syscallnum);
else
prdelset (sysset, syscallnum);
if (entry_or_exit == PR_SYSENTRY)
{
if (!proc_set_traced_sysentry (pi, sysset))
proc_error (pi, "proc-trace, set_traced_sysentry", __LINE__);
}
else
{
if (!proc_set_traced_sysexit (pi, sysset))
proc_error (pi, "proc-trace, set_traced_sysexit", __LINE__);
}
}
static void
proc_trace_syscalls (const char *args, int from_tty, int entry_or_exit, int mode)
{
procinfo *pi;
if (ptid_get_pid (inferior_ptid) <= 0)
error (_("you must be debugging a process to use this command."));
if (args == NULL || args[0] == 0)
error_no_arg (_("system call to trace"));
pi = find_procinfo_or_die (ptid_get_pid (inferior_ptid), 0);
if (isdigit (args[0]))
{
const int syscallnum = atoi (args);
proc_trace_syscalls_1 (pi, syscallnum, entry_or_exit, mode, from_tty);
}
}
static void
proc_trace_sysentry_cmd (const char *args, int from_tty)
{
proc_trace_syscalls (args, from_tty, PR_SYSENTRY, FLAG_SET);
}
static void
proc_trace_sysexit_cmd (const char *args, int from_tty)
{
proc_trace_syscalls (args, from_tty, PR_SYSEXIT, FLAG_SET);
}
static void
proc_untrace_sysentry_cmd (const char *args, int from_tty)
{
proc_trace_syscalls (args, from_tty, PR_SYSENTRY, FLAG_RESET);
}
static void
proc_untrace_sysexit_cmd (const char *args, int from_tty)
{
proc_trace_syscalls (args, from_tty, PR_SYSEXIT, FLAG_RESET);
}
void
_initialize_procfs (void)
{
observer_attach_inferior_created (procfs_inferior_created);
add_com ("proc-trace-entry", no_class, proc_trace_sysentry_cmd,
_("Give a trace of entries into the syscall."));
add_com ("proc-trace-exit", no_class, proc_trace_sysexit_cmd,
_("Give a trace of exits from the syscall."));
add_com ("proc-untrace-entry", no_class, proc_untrace_sysentry_cmd,
_("Cancel a trace of entries into the syscall."));
add_com ("proc-untrace-exit", no_class, proc_untrace_sysexit_cmd,
_("Cancel a trace of exits from the syscall."));
}
/* =================== END, GDB "MODULE" =================== */
/* miscellaneous stubs: */
/* The following satisfy a few random symbols mostly created by the
solaris threads implementation, which I will chase down later. */
/* Return a pid for which we guarantee we will be able to find a
'live' procinfo. */
ptid_t
procfs_first_available (void)
{
return pid_to_ptid (procinfo_list ? procinfo_list->pid : -1);
}
/* =================== GCORE .NOTE "MODULE" =================== */
static char *
procfs_do_thread_registers (bfd *obfd, ptid_t ptid,
char *note_data, int *note_size,
enum gdb_signal stop_signal)
{
struct regcache *regcache = get_thread_regcache (ptid);
gdb_gregset_t gregs;
gdb_fpregset_t fpregs;
unsigned long merged_pid;
merged_pid = ptid_get_lwp (ptid) << 16 | ptid_get_pid (ptid);
/* This part is the old method for fetching registers.
It should be replaced by the newer one using regsets
once it is implemented in this platform:
gdbarch_iterate_over_regset_sections(). */
scoped_restore save_inferior_ptid = make_scoped_restore (&inferior_ptid);
inferior_ptid = ptid;
target_fetch_registers (regcache, -1);
fill_gregset (regcache, &gregs, -1);
note_data = (char *) elfcore_write_lwpstatus (obfd,
note_data,
note_size,
merged_pid,
stop_signal,
&gregs);
fill_fpregset (regcache, &fpregs, -1);
note_data = (char *) elfcore_write_prfpreg (obfd,
note_data,
note_size,
&fpregs,
sizeof (fpregs));
return note_data;
}
struct procfs_corefile_thread_data {
bfd *obfd;
char *note_data;
int *note_size;
enum gdb_signal stop_signal;
};
static int
procfs_corefile_thread_callback (procinfo *pi, procinfo *thread, void *data)
{
struct procfs_corefile_thread_data *args
= (struct procfs_corefile_thread_data *) data;
if (pi != NULL)
{
ptid_t ptid = ptid_build (pi->pid, thread->tid, 0);
args->note_data = procfs_do_thread_registers (args->obfd, ptid,
args->note_data,
args->note_size,
args->stop_signal);
}
return 0;
}
static int
find_signalled_thread (struct thread_info *info, void *data)
{
if (info->suspend.stop_signal != GDB_SIGNAL_0
&& ptid_get_pid (info->ptid) == ptid_get_pid (inferior_ptid))
return 1;
return 0;
}
static enum gdb_signal
find_stop_signal (void)
{
struct thread_info *info =
iterate_over_threads (find_signalled_thread, NULL);
if (info)
return info->suspend.stop_signal;
else
return GDB_SIGNAL_0;
}
static char *
procfs_make_note_section (struct target_ops *self, bfd *obfd, int *note_size)
{
struct cleanup *old_chain;
gdb_gregset_t gregs;
gdb_fpregset_t fpregs;
char fname[16] = {'\0'};
char psargs[80] = {'\0'};
procinfo *pi = find_procinfo_or_die (ptid_get_pid (inferior_ptid), 0);
char *note_data = NULL;
char *inf_args;
struct procfs_corefile_thread_data thread_args;
gdb_byte *auxv;
int auxv_len;
enum gdb_signal stop_signal;
if (get_exec_file (0))
{
strncpy (fname, lbasename (get_exec_file (0)), sizeof (fname));
fname[sizeof (fname) - 1] = 0;
strncpy (psargs, get_exec_file (0), sizeof (psargs));
psargs[sizeof (psargs) - 1] = 0;
inf_args = get_inferior_args ();
if (inf_args && *inf_args &&
strlen (inf_args) < ((int) sizeof (psargs) - (int) strlen (psargs)))
{
strncat (psargs, " ",
sizeof (psargs) - strlen (psargs));
strncat (psargs, inf_args,
sizeof (psargs) - strlen (psargs));
}
}
note_data = (char *) elfcore_write_prpsinfo (obfd,
note_data,
note_size,
fname,
psargs);
stop_signal = find_stop_signal ();
fill_gregset (get_current_regcache (), &gregs, -1);
note_data = elfcore_write_pstatus (obfd, note_data, note_size,
ptid_get_pid (inferior_ptid),
stop_signal, &gregs);
thread_args.obfd = obfd;
thread_args.note_data = note_data;
thread_args.note_size = note_size;
thread_args.stop_signal = stop_signal;
proc_iterate_over_threads (pi, procfs_corefile_thread_callback,
&thread_args);
note_data = thread_args.note_data;
auxv_len = target_read_alloc (&current_target, TARGET_OBJECT_AUXV,
NULL, &auxv);
if (auxv_len > 0)
{
note_data = elfcore_write_note (obfd, note_data, note_size,
"CORE", NT_AUXV, auxv, auxv_len);
xfree (auxv);
}
return note_data;
}
/* =================== END GCORE .NOTE "MODULE" =================== */