/** Determine if we're running under a debugger by attempting to attach using pattach
*
* @return 0 if we're not, 1 if we are, -1 if we can't tell because of an error,
* -2 if we can't tell because we don't have the CAP_SYS_PTRACE capability.
*/
static int fr_get_debug_state(void)
{
int pid;
int from_child[2] = {-1, -1};
#ifdef HAVE_CAPABILITY_H
cap_flag_value_t value;
cap_t current;
/*
* If we're running under linux, we first need to check if we have
* permission to to ptrace. We do that using the capabilities
* functions.
*/
current = cap_get_proc();
if (!current) {
fr_strerror_printf("Failed getting process capabilities: %s", fr_syserror(errno));
return DEBUG_STATE_UNKNOWN;
}
if (cap_get_flag(current, CAP_SYS_PTRACE, CAP_PERMITTED, &value) < 0) {
fr_strerror_printf("Failed getting permitted ptrace capability state: %s",
fr_syserror(errno));
cap_free(current);
return DEBUG_STATE_UNKNOWN;
}
if ((value == CAP_SET) && (cap_get_flag(current, CAP_SYS_PTRACE, CAP_EFFECTIVE, &value) < 0)) {
fr_strerror_printf("Failed getting effective ptrace capability state: %s",
fr_syserror(errno));
cap_free(current);
return DEBUG_STATE_UNKNOWN;
}
/*
* We don't have permission to ptrace, so this test will always fail.
*/
if (value == CAP_CLEAR) {
fr_strerror_printf("ptrace capability not set. If debugger detection is required run as root or: "
"setcap cap_sys_ptrace+ep <path_to_radiusd>");
cap_free(current);
return DEBUG_STATE_UNKNOWN_NO_PTRACE_CAP;
}
cap_free(current);
#endif
if (pipe(from_child) < 0) {
fr_strerror_printf("Error opening internal pipe: %s", fr_syserror(errno));
return DEBUG_STATE_UNKNOWN;
}
pid = fork();
if (pid == -1) {
fr_strerror_printf("Error forking: %s", fr_syserror(errno));
return DEBUG_STATE_UNKNOWN;
}
/* Child */
if (pid == 0) {
int8_t ret = DEBUG_STATE_NOT_ATTACHED;
int ppid = getppid();
/* Close parent's side */
close(from_child[0]);
/*
* FreeBSD is extremely picky about the order of operations here
* we need to attach, wait *then* write whilst the parent is still
* suspended, then detach, continuing the process.
*
* If we don't do it in that order the read in the parent triggers
* a SIGKILL.
*/
if (_PTRACE(PT_ATTACH, ppid) == 0) {
/* Wait for the parent to stop */
waitpid(ppid, NULL, 0);
/* Tell the parent what happened */
if (write(from_child[1], &ret, sizeof(ret)) < 0) {
fprintf(stderr, "Writing ptrace status to parent failed: %s", fr_syserror(errno));
}
/* Detach */
_PTRACE_DETACH(ppid);
exit(0);
}
ret = DEBUG_STATE_ATTACHED;
/* Tell the parent what happened */
if (write(from_child[1], &ret, sizeof(ret)) < 0) {
fprintf(stderr, "Writing ptrace status to parent failed: %s", fr_syserror(errno));
}
exit(0);
/* Parent */
} else {
int8_t ret = DEBUG_STATE_UNKNOWN;
/*
* The child writes errno (reason) if pattach failed else 0.
*
* This read may be interrupted by pattach,
* which is why we need the loop.
*/
while ((read(from_child[0], &ret, sizeof(ret)) < 0) && (errno == EINTR));
/* Close the pipes here (if we did it above, it might race with pattach) */
close(from_child[1]);
close(from_child[0]);
/* Collect the status of the child */
waitpid(pid, NULL, 0);
return ret;
}
}
/*
* Prepare task for parallel debugger attach
* Returns SLURM_SUCCESS or SLURM_ERROR.
*/
int
pdebug_trace_process(stepd_step_rec_t *job, pid_t pid)
{
/* If task to be debugged, wait for it to stop via
* child's ptrace(PTRACE_TRACEME), then SIGSTOP, and
* ptrace(PTRACE_DETACH).
*/
if (job->task_flags & TASK_PARALLEL_DEBUG) {
int status;
waitpid(pid, &status, WUNTRACED);
if (!WIFSTOPPED(status)) {
int i;
error("pdebug_trace_process WIFSTOPPED false"
" for pid %d", pid);
if (WIFEXITED(status)) {
error("Process %d exited \"normally\""
" with return code %d",
pid,
WEXITSTATUS(status));
} else if (WIFSIGNALED(status)) {
error("Process %d killed by signal %d",
pid, WTERMSIG(status));
}
/*
* Mark this process as complete since it died
* prematurely.
*/
for (i = 0; i < job->node_tasks; i++) {
if (job->task[i]->pid == pid) {
job->task[i]->state =
STEPD_STEP_TASK_COMPLETE;
}
}
return SLURM_ERROR;
}
if ((pid > (pid_t) 0) && (kill(pid, SIGSTOP) < 0)) {
error("kill(%lu): %m", (unsigned long) pid);
return SLURM_ERROR;
}
#ifdef BSD
if (_PTRACE(PT_DETACH, pid, (caddr_t)1, 0)) {
#elif defined(PT_DETACH)
if (_PTRACE(PT_DETACH, pid, NULL, 0)) {
#elif defined(__sun)
if (_PTRACE(7, pid, NULL, 0)) {
#elif defined(__CYGWIN__)
if (1) {
debug3("No ptrace for cygwin");
} else {
#else
if (_PTRACE(PTRACE_DETACH, pid, NULL, 0)) {
#endif
error("ptrace(%lu): %m", (unsigned long) pid);
return SLURM_ERROR;
}
}
return SLURM_SUCCESS;
}
/*
* Stop current task on exec() for connection from a parallel debugger
*/
void
pdebug_stop_current(stepd_step_rec_t *job)
{
/*
* Stop the task on exec for TotalView to connect
*/
if ( (job->task_flags & TASK_PARALLEL_DEBUG)
#ifdef BSD
&& (_PTRACE(PT_TRACE_ME, 0, (caddr_t)0, 0) < 0) )
#elif defined(PT_TRACE_ME)
&& (_PTRACE(PT_TRACE_ME, 0, NULL, 0) < 0) )
#elif defined(__sun)
&& (_PTRACE(0, 0, NULL, 0) < 0))
#elif defined(__CYGWIN__)
&& 0)
#else
&& (_PTRACE(PTRACE_TRACEME, 0, NULL, 0) < 0) )
#endif
error("ptrace: %m");
}
/* Check if this PID should be woken for TotalView partitial attach */
static bool _pid_to_wake(pid_t pid)
{
#ifdef CLONE_PTRACE
char proc_stat[1024], proc_name[22], state[1], *str_ptr;
int len, proc_fd, ppid, pgrp, session, tty, tpgid;
long unsigned flags;
sprintf (proc_name, "/proc/%d/stat", (int) pid);
if ((proc_fd = open(proc_name, O_RDONLY, 0)) == -1)
return false; /* process is now gone */
len = read(proc_fd, proc_stat, sizeof(proc_stat));
close(proc_fd);
if (len < 14)
return false;
/* skip over "PID (CMD) " */
if ((str_ptr = (char *)strrchr(proc_stat, ')')) == NULL)
return false;
if (sscanf(str_ptr + 2,
"%c %d %d %d %d %d %lu ",
state, &ppid, &pgrp, &session, &tty, &tpgid, &flags) != 7)
return false;
if ((flags & CLONE_PTRACE) == 0)
return true;
return false;
#else
int status;
waitpid(pid, &status, (WUNTRACED | WNOHANG));
if (WIFSTOPPED(status))
return true;
return false;
#endif
}
/*
* Wake tasks currently stopped for parallel debugger attach
*/
void pdebug_wake_process(stepd_step_rec_t *job, pid_t pid)
{
if ((job->task_flags & TASK_PARALLEL_DEBUG) && (pid > (pid_t) 0)) {
if (_pid_to_wake(pid)) {
if (kill(pid, SIGCONT) < 0)
error("kill(%lu): %m", (unsigned long) pid);
else
debug("woke pid %lu", (unsigned long) pid);
} else {
debug("pid %lu not stopped", (unsigned long) pid);
}
}
}
