mirror of
https://github.com/ruby/ruby.git
synced 2022-11-09 12:17:21 -05:00
708bfd2115
To reduce resource use and reduce CI failure; remove timer-thread. Single-threaded Ruby processes (including forked children) will never see extra thread overhead. This prevents glibc and jemalloc from going into multi-threaded mode and initializing locks or causing fragmentation via arena explosion. The GVL is implements its own wait-queue as a ccan/list to permit controlling wakeup order. Timeslice under contention is handled by a designated timer thread (similar to choosing a "patrol_thread" for current deadlock checking). There is only one self-pipe, now, as wakeups for timeslice are done independently using condition variables. This reduces FD pressure slightly. Signal handling is handled directly by a Ruby Thread (instead of timer-thread) by exposing signal self-pipe to callers of rb_thread_fd_select, native_sleep, rb_wait_for_single_fd, etc... Acquiring, using, and releasing the self-pipe is exposed via 4 new internal functions: 1) rb_sigwait_fd_get - exclusively acquire timer_thread_pipe.normal[0] 2) rb_sigwait_fd_sleep - sleep and wait for signal (and no other FDs) 3) rb_sigwait_fd_put - release acquired result from rb_sigwait_fd_get 4) rb_sigwait_fd_migrate - migrate signal handling to another thread after calling rb_sigwait_fd_put. rb_sigwait_fd_migrate is necessary for waitpid callers because only one thread can wait on self-pipe at a time, otherwise a deadlock will occur if threads fight over the self-pipe. TRAP_INTERRUPT_MASK is now set for the main thread directly in signal handler via rb_thread_wakeup_timer_thread. Originally, I wanted to use POSIX timers (timer_create/timer_settime) for this. Unfortunately, this proved unfeasible as Mutex#sleep resumes on spurious wakeups and test/thread/test_cv.rb::test_condvar_timed_wait failed. Using pthread_sigmask to mask out SIGVTALRM fixed that test, but test/fiddle/test_function.rb::test_nogvl_poll proved there'd be some unavoidable (and frequent) incompatibilities from that approach. Finally, this allows us to drop thread_destruct_lock and interrupt current ec directly. We don't need to rely on vm->thread_destruct_lock or a coherent vm->running_thread on any platform. Separate timer-thread for time slice and signal handling is relegated to thread_win32.c, now. [ruby-core:88088] [Misc #14937] git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@64107 b2dd03c8-39d4-4d8f-98ff-823fe69b080e
1629 lines
36 KiB
C
1629 lines
36 KiB
C
/**********************************************************************
|
|
|
|
signal.c -
|
|
|
|
$Author$
|
|
created at: Tue Dec 20 10:13:44 JST 1994
|
|
|
|
Copyright (C) 1993-2007 Yukihiro Matsumoto
|
|
Copyright (C) 2000 Network Applied Communication Laboratory, Inc.
|
|
Copyright (C) 2000 Information-technology Promotion Agency, Japan
|
|
|
|
**********************************************************************/
|
|
|
|
#include "internal.h"
|
|
#include "vm_core.h"
|
|
#include <signal.h>
|
|
#include <stdio.h>
|
|
#include <errno.h>
|
|
#include "ruby_atomic.h"
|
|
#include "eval_intern.h"
|
|
#ifdef HAVE_UNISTD_H
|
|
# include <unistd.h>
|
|
#endif
|
|
#ifdef HAVE_SYS_UIO_H
|
|
#include <sys/uio.h>
|
|
#endif
|
|
#ifdef HAVE_UCONTEXT_H
|
|
#include <ucontext.h>
|
|
#endif
|
|
|
|
#ifdef HAVE_VALGRIND_MEMCHECK_H
|
|
# include <valgrind/memcheck.h>
|
|
# ifndef VALGRIND_MAKE_MEM_DEFINED
|
|
# define VALGRIND_MAKE_MEM_DEFINED(p, n) VALGRIND_MAKE_READABLE((p), (n))
|
|
# endif
|
|
# ifndef VALGRIND_MAKE_MEM_UNDEFINED
|
|
# define VALGRIND_MAKE_MEM_UNDEFINED(p, n) VALGRIND_MAKE_WRITABLE((p), (n))
|
|
# endif
|
|
#else
|
|
# define VALGRIND_MAKE_MEM_DEFINED(p, n) 0
|
|
# define VALGRIND_MAKE_MEM_UNDEFINED(p, n) 0
|
|
#endif
|
|
|
|
#ifdef NEED_RUBY_ATOMIC_OPS
|
|
rb_atomic_t
|
|
ruby_atomic_exchange(rb_atomic_t *ptr, rb_atomic_t val)
|
|
{
|
|
rb_atomic_t old = *ptr;
|
|
*ptr = val;
|
|
return old;
|
|
}
|
|
|
|
rb_atomic_t
|
|
ruby_atomic_compare_and_swap(rb_atomic_t *ptr, rb_atomic_t cmp,
|
|
rb_atomic_t newval)
|
|
{
|
|
rb_atomic_t old = *ptr;
|
|
if (old == cmp) {
|
|
*ptr = newval;
|
|
}
|
|
return old;
|
|
}
|
|
#endif
|
|
|
|
#define FOREACH_SIGNAL(sig, offset) \
|
|
for (sig = siglist + (offset); sig < siglist + numberof(siglist); ++sig)
|
|
enum { LONGEST_SIGNAME = 7 }; /* MIGRATE and RETRACT */
|
|
static const struct signals {
|
|
char signm[LONGEST_SIGNAME + 1];
|
|
int signo;
|
|
} siglist [] = {
|
|
{"EXIT", 0},
|
|
#ifdef SIGHUP
|
|
{"HUP", SIGHUP},
|
|
#endif
|
|
{"INT", SIGINT},
|
|
#ifdef SIGQUIT
|
|
{"QUIT", SIGQUIT},
|
|
#endif
|
|
#ifdef SIGILL
|
|
{"ILL", SIGILL},
|
|
#endif
|
|
#ifdef SIGTRAP
|
|
{"TRAP", SIGTRAP},
|
|
#endif
|
|
#ifdef SIGABRT
|
|
{"ABRT", SIGABRT},
|
|
#endif
|
|
#ifdef SIGIOT
|
|
{"IOT", SIGIOT},
|
|
#endif
|
|
#ifdef SIGEMT
|
|
{"EMT", SIGEMT},
|
|
#endif
|
|
#ifdef SIGFPE
|
|
{"FPE", SIGFPE},
|
|
#endif
|
|
#ifdef SIGKILL
|
|
{"KILL", SIGKILL},
|
|
#endif
|
|
#ifdef SIGBUS
|
|
{"BUS", SIGBUS},
|
|
#endif
|
|
#ifdef SIGSEGV
|
|
{"SEGV", SIGSEGV},
|
|
#endif
|
|
#ifdef SIGSYS
|
|
{"SYS", SIGSYS},
|
|
#endif
|
|
#ifdef SIGPIPE
|
|
{"PIPE", SIGPIPE},
|
|
#endif
|
|
#ifdef SIGALRM
|
|
{"ALRM", SIGALRM},
|
|
#endif
|
|
#ifdef SIGTERM
|
|
{"TERM", SIGTERM},
|
|
#endif
|
|
#ifdef SIGURG
|
|
{"URG", SIGURG},
|
|
#endif
|
|
#ifdef SIGSTOP
|
|
{"STOP", SIGSTOP},
|
|
#endif
|
|
#ifdef SIGTSTP
|
|
{"TSTP", SIGTSTP},
|
|
#endif
|
|
#ifdef SIGCONT
|
|
{"CONT", SIGCONT},
|
|
#endif
|
|
#if RUBY_SIGCHLD
|
|
{"CHLD", RUBY_SIGCHLD },
|
|
{"CLD", RUBY_SIGCHLD },
|
|
#endif
|
|
#ifdef SIGTTIN
|
|
{"TTIN", SIGTTIN},
|
|
#endif
|
|
#ifdef SIGTTOU
|
|
{"TTOU", SIGTTOU},
|
|
#endif
|
|
#ifdef SIGIO
|
|
{"IO", SIGIO},
|
|
#endif
|
|
#ifdef SIGXCPU
|
|
{"XCPU", SIGXCPU},
|
|
#endif
|
|
#ifdef SIGXFSZ
|
|
{"XFSZ", SIGXFSZ},
|
|
#endif
|
|
#ifdef SIGVTALRM
|
|
{"VTALRM", SIGVTALRM},
|
|
#endif
|
|
#ifdef SIGPROF
|
|
{"PROF", SIGPROF},
|
|
#endif
|
|
#ifdef SIGWINCH
|
|
{"WINCH", SIGWINCH},
|
|
#endif
|
|
#ifdef SIGUSR1
|
|
{"USR1", SIGUSR1},
|
|
#endif
|
|
#ifdef SIGUSR2
|
|
{"USR2", SIGUSR2},
|
|
#endif
|
|
#ifdef SIGLOST
|
|
{"LOST", SIGLOST},
|
|
#endif
|
|
#ifdef SIGMSG
|
|
{"MSG", SIGMSG},
|
|
#endif
|
|
#ifdef SIGPWR
|
|
{"PWR", SIGPWR},
|
|
#endif
|
|
#ifdef SIGPOLL
|
|
{"POLL", SIGPOLL},
|
|
#endif
|
|
#ifdef SIGDANGER
|
|
{"DANGER", SIGDANGER},
|
|
#endif
|
|
#ifdef SIGMIGRATE
|
|
{"MIGRATE", SIGMIGRATE},
|
|
#endif
|
|
#ifdef SIGPRE
|
|
{"PRE", SIGPRE},
|
|
#endif
|
|
#ifdef SIGGRANT
|
|
{"GRANT", SIGGRANT},
|
|
#endif
|
|
#ifdef SIGRETRACT
|
|
{"RETRACT", SIGRETRACT},
|
|
#endif
|
|
#ifdef SIGSOUND
|
|
{"SOUND", SIGSOUND},
|
|
#endif
|
|
#ifdef SIGINFO
|
|
{"INFO", SIGINFO},
|
|
#endif
|
|
};
|
|
|
|
static const char signame_prefix[3] = "SIG";
|
|
static const int signame_prefix_len = (int)sizeof(signame_prefix);
|
|
|
|
static int
|
|
signm2signo(VALUE *sig_ptr, int negative, int exit, int *prefix_ptr)
|
|
{
|
|
const struct signals *sigs;
|
|
VALUE vsig = *sig_ptr;
|
|
const char *nm;
|
|
long len, nmlen;
|
|
int prefix = 0;
|
|
|
|
if (RB_SYMBOL_P(vsig)) {
|
|
*sig_ptr = vsig = rb_sym2str(vsig);
|
|
}
|
|
else if (!RB_TYPE_P(vsig, T_STRING)) {
|
|
VALUE str = rb_check_string_type(vsig);
|
|
if (NIL_P(str)) {
|
|
rb_raise(rb_eArgError, "bad signal type %s",
|
|
rb_obj_classname(vsig));
|
|
}
|
|
*sig_ptr = vsig = str;
|
|
}
|
|
|
|
rb_must_asciicompat(vsig);
|
|
RSTRING_GETMEM(vsig, nm, len);
|
|
if (memchr(nm, '\0', len)) {
|
|
rb_raise(rb_eArgError, "signal name with null byte");
|
|
}
|
|
|
|
if (len > 0 && nm[0] == '-') {
|
|
if (!negative)
|
|
rb_raise(rb_eArgError, "negative signal name: % "PRIsVALUE, vsig);
|
|
prefix = 1;
|
|
}
|
|
else {
|
|
negative = 0;
|
|
}
|
|
if (len >= prefix + signame_prefix_len) {
|
|
if (memcmp(nm + prefix, signame_prefix, sizeof(signame_prefix)) == 0)
|
|
prefix += signame_prefix_len;
|
|
}
|
|
if (len <= (long)prefix) {
|
|
unsupported:
|
|
if (prefix == signame_prefix_len) {
|
|
prefix = 0;
|
|
}
|
|
else if (prefix > signame_prefix_len) {
|
|
prefix -= signame_prefix_len;
|
|
len -= prefix;
|
|
vsig = rb_str_subseq(vsig, prefix, len);
|
|
prefix = 0;
|
|
}
|
|
else {
|
|
len -= prefix;
|
|
vsig = rb_str_subseq(vsig, prefix, len);
|
|
prefix = signame_prefix_len;
|
|
}
|
|
rb_raise(rb_eArgError, "unsupported signal `%.*s%"PRIsVALUE"'",
|
|
prefix, signame_prefix, vsig);
|
|
}
|
|
|
|
if (prefix_ptr) *prefix_ptr = prefix;
|
|
nmlen = len - prefix;
|
|
nm += prefix;
|
|
if (nmlen > LONGEST_SIGNAME) goto unsupported;
|
|
FOREACH_SIGNAL(sigs, !exit) {
|
|
if (memcmp(sigs->signm, nm, nmlen) == 0 &&
|
|
sigs->signm[nmlen] == '\0') {
|
|
return negative ? -sigs->signo : sigs->signo;
|
|
}
|
|
}
|
|
goto unsupported;
|
|
}
|
|
|
|
static const char*
|
|
signo2signm(int no)
|
|
{
|
|
const struct signals *sigs;
|
|
|
|
FOREACH_SIGNAL(sigs, 0) {
|
|
if (sigs->signo == no)
|
|
return sigs->signm;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* Signal.signame(signo) -> string or nil
|
|
*
|
|
* Convert signal number to signal name.
|
|
* Returns +nil+ if the signo is an invalid signal number.
|
|
*
|
|
* Signal.trap("INT") { |signo| puts Signal.signame(signo) }
|
|
* Process.kill("INT", 0)
|
|
*
|
|
* <em>produces:</em>
|
|
*
|
|
* INT
|
|
*/
|
|
static VALUE
|
|
sig_signame(VALUE recv, VALUE signo)
|
|
{
|
|
const char *signame = signo2signm(NUM2INT(signo));
|
|
if (!signame) return Qnil;
|
|
return rb_str_new_cstr(signame);
|
|
}
|
|
|
|
const char *
|
|
ruby_signal_name(int no)
|
|
{
|
|
return signo2signm(no);
|
|
}
|
|
|
|
static VALUE
|
|
rb_signo2signm(int signo)
|
|
{
|
|
const char *const signm = signo2signm(signo);
|
|
if (signm) {
|
|
return rb_sprintf("SIG%s", signm);
|
|
}
|
|
else {
|
|
return rb_sprintf("SIG%u", signo);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* SignalException.new(sig_name) -> signal_exception
|
|
* SignalException.new(sig_number [, name]) -> signal_exception
|
|
*
|
|
* Construct a new SignalException object. +sig_name+ should be a known
|
|
* signal name.
|
|
*/
|
|
|
|
static VALUE
|
|
esignal_init(int argc, VALUE *argv, VALUE self)
|
|
{
|
|
int argnum = 1;
|
|
VALUE sig = Qnil;
|
|
int signo;
|
|
|
|
if (argc > 0) {
|
|
sig = rb_check_to_integer(argv[0], "to_int");
|
|
if (!NIL_P(sig)) argnum = 2;
|
|
else sig = argv[0];
|
|
}
|
|
rb_check_arity(argc, 1, argnum);
|
|
if (argnum == 2) {
|
|
signo = NUM2INT(sig);
|
|
if (signo < 0 || signo > NSIG) {
|
|
rb_raise(rb_eArgError, "invalid signal number (%d)", signo);
|
|
}
|
|
if (argc > 1) {
|
|
sig = argv[1];
|
|
}
|
|
else {
|
|
sig = rb_signo2signm(signo);
|
|
}
|
|
}
|
|
else {
|
|
int prefix;
|
|
signo = signm2signo(&sig, FALSE, FALSE, &prefix);
|
|
if (prefix != signame_prefix_len) {
|
|
sig = rb_str_append(rb_str_new_cstr("SIG"), sig);
|
|
}
|
|
}
|
|
rb_call_super(1, &sig);
|
|
rb_ivar_set(self, id_signo, INT2NUM(signo));
|
|
|
|
return self;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* signal_exception.signo -> num
|
|
*
|
|
* Returns a signal number.
|
|
*/
|
|
|
|
static VALUE
|
|
esignal_signo(VALUE self)
|
|
{
|
|
return rb_ivar_get(self, id_signo);
|
|
}
|
|
|
|
/* :nodoc: */
|
|
static VALUE
|
|
interrupt_init(int argc, VALUE *argv, VALUE self)
|
|
{
|
|
VALUE args[2];
|
|
|
|
args[0] = INT2FIX(SIGINT);
|
|
rb_scan_args(argc, argv, "01", &args[1]);
|
|
return rb_call_super(2, args);
|
|
}
|
|
|
|
void
|
|
ruby_default_signal(int sig)
|
|
{
|
|
signal(sig, SIG_DFL);
|
|
raise(sig);
|
|
}
|
|
|
|
static RETSIGTYPE sighandler(int sig);
|
|
static int signal_ignored(int sig);
|
|
static void signal_enque(int sig);
|
|
|
|
/*
|
|
* call-seq:
|
|
* Process.kill(signal, pid, ...) -> integer
|
|
*
|
|
* Sends the given signal to the specified process id(s) if _pid_ is positive.
|
|
* If _pid_ is zero _signal_ is sent to all processes whose group ID is equal
|
|
* to the group ID of the process. _signal_ may be an integer signal number or
|
|
* a POSIX signal name (either with or without a +SIG+ prefix). If _signal_ is
|
|
* negative (or starts with a minus sign), kills process groups instead of
|
|
* processes. Not all signals are available on all platforms.
|
|
* The keys and values of +Signal.list+ are known signal names and numbers,
|
|
* respectively.
|
|
*
|
|
* pid = fork do
|
|
* Signal.trap("HUP") { puts "Ouch!"; exit }
|
|
* # ... do some work ...
|
|
* end
|
|
* # ...
|
|
* Process.kill("HUP", pid)
|
|
* Process.wait
|
|
*
|
|
* <em>produces:</em>
|
|
*
|
|
* Ouch!
|
|
*
|
|
* If _signal_ is an integer but wrong for signal,
|
|
* <code>Errno::EINVAL</code> or +RangeError+ will be raised.
|
|
* Otherwise unless _signal_ is a +String+ or a +Symbol+, and a known
|
|
* signal name, +ArgumentError+ will be raised.
|
|
*
|
|
* Also, <code>Errno::ESRCH</code> or +RangeError+ for invalid _pid_,
|
|
* <code>Errno::EPERM</code> when failed because of no privilege,
|
|
* will be raised. In these cases, signals may have been sent to
|
|
* preceding processes.
|
|
*/
|
|
|
|
VALUE
|
|
rb_f_kill(int argc, const VALUE *argv)
|
|
{
|
|
#ifndef HAVE_KILLPG
|
|
#define killpg(pg, sig) kill(-(pg), (sig))
|
|
#endif
|
|
int sig;
|
|
int i;
|
|
VALUE str;
|
|
|
|
rb_check_arity(argc, 2, UNLIMITED_ARGUMENTS);
|
|
|
|
if (FIXNUM_P(argv[0])) {
|
|
sig = FIX2INT(argv[0]);
|
|
}
|
|
else {
|
|
str = argv[0];
|
|
sig = signm2signo(&str, TRUE, FALSE, NULL);
|
|
}
|
|
|
|
if (argc <= 1) return INT2FIX(0);
|
|
|
|
if (sig < 0) {
|
|
sig = -sig;
|
|
for (i=1; i<argc; i++) {
|
|
if (killpg(NUM2PIDT(argv[i]), sig) < 0)
|
|
rb_sys_fail(0);
|
|
}
|
|
}
|
|
else {
|
|
const rb_pid_t self = (GET_THREAD() == GET_VM()->main_thread) ? getpid() : -1;
|
|
int wakeup = 0;
|
|
|
|
for (i=1; i<argc; i++) {
|
|
rb_pid_t pid = NUM2PIDT(argv[i]);
|
|
|
|
if ((sig != 0) && (self != -1) && (pid == self)) {
|
|
int t;
|
|
/*
|
|
* When target pid is self, many caller assume signal will be
|
|
* delivered immediately and synchronously.
|
|
*/
|
|
switch (sig) {
|
|
case SIGSEGV:
|
|
#ifdef SIGBUS
|
|
case SIGBUS:
|
|
#endif
|
|
#ifdef SIGKILL
|
|
case SIGKILL:
|
|
#endif
|
|
#ifdef SIGILL
|
|
case SIGILL:
|
|
#endif
|
|
#ifdef SIGFPE
|
|
case SIGFPE:
|
|
#endif
|
|
#ifdef SIGSTOP
|
|
case SIGSTOP:
|
|
#endif
|
|
kill(pid, sig);
|
|
break;
|
|
default:
|
|
t = signal_ignored(sig);
|
|
if (t) {
|
|
if (t < 0 && kill(pid, sig))
|
|
rb_sys_fail(0);
|
|
break;
|
|
}
|
|
signal_enque(sig);
|
|
wakeup = 1;
|
|
}
|
|
}
|
|
else if (kill(pid, sig) < 0) {
|
|
rb_sys_fail(0);
|
|
}
|
|
}
|
|
if (wakeup) {
|
|
rb_threadptr_check_signal(GET_VM()->main_thread);
|
|
}
|
|
}
|
|
rb_thread_execute_interrupts(rb_thread_current());
|
|
|
|
return INT2FIX(i-1);
|
|
}
|
|
|
|
static struct {
|
|
rb_atomic_t cnt[RUBY_NSIG];
|
|
rb_atomic_t size;
|
|
} signal_buff;
|
|
volatile unsigned int ruby_nocldwait;
|
|
|
|
#ifdef __dietlibc__
|
|
#define sighandler_t sh_t
|
|
#else
|
|
#define sighandler_t ruby_sighandler_t
|
|
#endif
|
|
|
|
typedef RETSIGTYPE (*sighandler_t)(int);
|
|
#ifdef USE_SIGALTSTACK
|
|
typedef void ruby_sigaction_t(int, siginfo_t*, void*);
|
|
#define SIGINFO_ARG , siginfo_t *info, void *ctx
|
|
#define SIGINFO_CTX ctx
|
|
#else
|
|
typedef RETSIGTYPE ruby_sigaction_t(int);
|
|
#define SIGINFO_ARG
|
|
#define SIGINFO_CTX 0
|
|
#endif
|
|
|
|
#ifdef USE_SIGALTSTACK
|
|
static int
|
|
rb_sigaltstack_size(void)
|
|
{
|
|
/* XXX: BSD_vfprintf() uses >1500KiB stack and x86-64 need >5KiB stack. */
|
|
int size = 16*1024;
|
|
|
|
#ifdef MINSIGSTKSZ
|
|
if (size < MINSIGSTKSZ)
|
|
size = MINSIGSTKSZ;
|
|
#endif
|
|
#if defined(HAVE_SYSCONF) && defined(_SC_PAGE_SIZE)
|
|
{
|
|
int pagesize;
|
|
pagesize = (int)sysconf(_SC_PAGE_SIZE);
|
|
if (size < pagesize)
|
|
size = pagesize;
|
|
}
|
|
#endif
|
|
|
|
return size;
|
|
}
|
|
|
|
/* alternate stack for SIGSEGV */
|
|
void *
|
|
rb_register_sigaltstack(void)
|
|
{
|
|
stack_t newSS, oldSS;
|
|
|
|
newSS.ss_size = rb_sigaltstack_size();
|
|
newSS.ss_sp = xmalloc(newSS.ss_size);
|
|
newSS.ss_flags = 0;
|
|
|
|
sigaltstack(&newSS, &oldSS); /* ignore error. */
|
|
|
|
return newSS.ss_sp;
|
|
}
|
|
#endif /* USE_SIGALTSTACK */
|
|
|
|
#ifdef POSIX_SIGNAL
|
|
static sighandler_t
|
|
ruby_signal(int signum, sighandler_t handler)
|
|
{
|
|
struct sigaction sigact, old;
|
|
|
|
#if 0
|
|
rb_trap_accept_nativethreads[signum] = 0;
|
|
#endif
|
|
|
|
sigemptyset(&sigact.sa_mask);
|
|
#ifdef USE_SIGALTSTACK
|
|
if (handler == SIG_IGN || handler == SIG_DFL) {
|
|
sigact.sa_handler = handler;
|
|
sigact.sa_flags = 0;
|
|
}
|
|
else {
|
|
sigact.sa_sigaction = (ruby_sigaction_t*)handler;
|
|
sigact.sa_flags = SA_SIGINFO;
|
|
}
|
|
#else
|
|
sigact.sa_handler = handler;
|
|
sigact.sa_flags = 0;
|
|
#endif
|
|
|
|
switch (signum) {
|
|
case SIGCHLD:
|
|
if (handler == SIG_IGN) {
|
|
ruby_nocldwait = 1;
|
|
if (sigact.sa_flags & SA_SIGINFO) {
|
|
sigact.sa_sigaction = (ruby_sigaction_t*)sighandler;
|
|
}
|
|
else {
|
|
sigact.sa_handler = sighandler;
|
|
}
|
|
}
|
|
else {
|
|
ruby_nocldwait = 0;
|
|
}
|
|
break;
|
|
#if defined(SA_ONSTACK) && defined(USE_SIGALTSTACK)
|
|
case SIGSEGV:
|
|
#ifdef SIGBUS
|
|
case SIGBUS:
|
|
#endif
|
|
sigact.sa_flags |= SA_ONSTACK;
|
|
break;
|
|
#endif
|
|
}
|
|
(void)VALGRIND_MAKE_MEM_DEFINED(&old, sizeof(old));
|
|
if (sigaction(signum, &sigact, &old) < 0) {
|
|
return SIG_ERR;
|
|
}
|
|
if (old.sa_flags & SA_SIGINFO)
|
|
handler = (sighandler_t)old.sa_sigaction;
|
|
else
|
|
handler = old.sa_handler;
|
|
ASSUME(handler != SIG_ERR);
|
|
return handler;
|
|
}
|
|
|
|
sighandler_t
|
|
posix_signal(int signum, sighandler_t handler)
|
|
{
|
|
return ruby_signal(signum, handler);
|
|
}
|
|
|
|
#elif defined _WIN32
|
|
static inline sighandler_t
|
|
ruby_signal(int signum, sighandler_t handler)
|
|
{
|
|
if (signum == SIGKILL) {
|
|
errno = EINVAL;
|
|
return SIG_ERR;
|
|
}
|
|
return signal(signum, handler);
|
|
}
|
|
|
|
#else /* !POSIX_SIGNAL */
|
|
#define ruby_signal(sig,handler) (/* rb_trap_accept_nativethreads[(sig)] = 0,*/ signal((sig),(handler)))
|
|
#if 0 /* def HAVE_NATIVETHREAD */
|
|
static sighandler_t
|
|
ruby_nativethread_signal(int signum, sighandler_t handler)
|
|
{
|
|
sighandler_t old;
|
|
|
|
old = signal(signum, handler);
|
|
rb_trap_accept_nativethreads[signum] = 1;
|
|
return old;
|
|
}
|
|
#endif
|
|
#endif
|
|
|
|
static int
|
|
signal_ignored(int sig)
|
|
{
|
|
sighandler_t func;
|
|
#ifdef POSIX_SIGNAL
|
|
struct sigaction old;
|
|
(void)VALGRIND_MAKE_MEM_DEFINED(&old, sizeof(old));
|
|
if (sigaction(sig, NULL, &old) < 0) return FALSE;
|
|
func = old.sa_handler;
|
|
#else
|
|
sighandler_t old = signal(sig, SIG_DFL);
|
|
signal(sig, old);
|
|
func = old;
|
|
#endif
|
|
if (func == SIG_IGN) return 1;
|
|
return func == sighandler ? 0 : -1;
|
|
}
|
|
|
|
static void
|
|
signal_enque(int sig)
|
|
{
|
|
ATOMIC_INC(signal_buff.cnt[sig]);
|
|
ATOMIC_INC(signal_buff.size);
|
|
}
|
|
|
|
static rb_atomic_t sigchld_hit;
|
|
|
|
static RETSIGTYPE
|
|
sighandler(int sig)
|
|
{
|
|
int old_errnum = errno;
|
|
|
|
/* the VM always needs to handle SIGCHLD for rb_waitpid */
|
|
if (sig == RUBY_SIGCHLD) {
|
|
rb_vm_t *vm = GET_VM();
|
|
ATOMIC_EXCHANGE(sigchld_hit, 1);
|
|
|
|
/* avoid spurious wakeup in main thread iff nobody uses trap(:CHLD) */
|
|
if (vm && ACCESS_ONCE(VALUE, vm->trap_list.cmd[sig])) {
|
|
signal_enque(sig);
|
|
}
|
|
}
|
|
else {
|
|
signal_enque(sig);
|
|
}
|
|
rb_thread_wakeup_timer_thread(sig);
|
|
#if !defined(BSD_SIGNAL) && !defined(POSIX_SIGNAL)
|
|
ruby_signal(sig, sighandler);
|
|
#endif
|
|
|
|
errno = old_errnum;
|
|
}
|
|
|
|
int
|
|
rb_signal_buff_size(void)
|
|
{
|
|
return signal_buff.size;
|
|
}
|
|
|
|
#if HAVE_PTHREAD_H
|
|
#include <pthread.h>
|
|
#endif
|
|
|
|
static void
|
|
rb_disable_interrupt(void)
|
|
{
|
|
#ifdef HAVE_PTHREAD_SIGMASK
|
|
sigset_t mask;
|
|
sigfillset(&mask);
|
|
pthread_sigmask(SIG_SETMASK, &mask, NULL);
|
|
#endif
|
|
}
|
|
|
|
static void
|
|
rb_enable_interrupt(void)
|
|
{
|
|
#ifdef HAVE_PTHREAD_SIGMASK
|
|
sigset_t mask;
|
|
sigemptyset(&mask);
|
|
pthread_sigmask(SIG_SETMASK, &mask, NULL);
|
|
#endif
|
|
}
|
|
|
|
int
|
|
rb_get_next_signal(void)
|
|
{
|
|
int i, sig = 0;
|
|
|
|
if (signal_buff.size != 0) {
|
|
for (i=1; i<RUBY_NSIG; i++) {
|
|
if (signal_buff.cnt[i] > 0) {
|
|
ATOMIC_DEC(signal_buff.cnt[i]);
|
|
ATOMIC_DEC(signal_buff.size);
|
|
sig = i;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
return sig;
|
|
}
|
|
|
|
#if defined SIGSEGV || defined SIGBUS || defined SIGILL || defined SIGFPE
|
|
static const char *received_signal;
|
|
# define clear_received_signal() (void)(ruby_disable_gc = 0, received_signal = 0)
|
|
#else
|
|
# define clear_received_signal() ((void)0)
|
|
#endif
|
|
|
|
#if defined(USE_SIGALTSTACK) || defined(_WIN32)
|
|
NORETURN(void rb_ec_stack_overflow(rb_execution_context_t *ec, int crit));
|
|
# if defined __HAIKU__
|
|
# define USE_UCONTEXT_REG 1
|
|
# elif !(defined(HAVE_UCONTEXT_H) && (defined __i386__ || defined __x86_64__ || defined __amd64__))
|
|
# elif defined __linux__
|
|
# define USE_UCONTEXT_REG 1
|
|
# elif defined __APPLE__
|
|
# define USE_UCONTEXT_REG 1
|
|
# elif defined __FreeBSD__
|
|
# define USE_UCONTEXT_REG 1
|
|
# endif
|
|
#if defined(HAVE_PTHREAD_SIGMASK)
|
|
# define ruby_sigunmask pthread_sigmask
|
|
#elif defined(HAVE_SIGPROCMASK)
|
|
# define ruby_sigunmask sigprocmask
|
|
#endif
|
|
static void
|
|
reset_sigmask(int sig)
|
|
{
|
|
#if defined(ruby_sigunmask)
|
|
sigset_t mask;
|
|
#endif
|
|
clear_received_signal();
|
|
#if defined(ruby_sigunmask)
|
|
sigemptyset(&mask);
|
|
sigaddset(&mask, sig);
|
|
if (ruby_sigunmask(SIG_UNBLOCK, &mask, NULL)) {
|
|
rb_bug_errno(STRINGIZE(ruby_sigunmask)":unblock", errno);
|
|
}
|
|
#endif
|
|
}
|
|
|
|
# ifdef USE_UCONTEXT_REG
|
|
static void
|
|
check_stack_overflow(int sig, const uintptr_t addr, const ucontext_t *ctx)
|
|
{
|
|
const DEFINE_MCONTEXT_PTR(mctx, ctx);
|
|
# if defined __linux__
|
|
# if defined REG_RSP
|
|
const greg_t sp = mctx->gregs[REG_RSP];
|
|
const greg_t bp = mctx->gregs[REG_RBP];
|
|
# else
|
|
const greg_t sp = mctx->gregs[REG_ESP];
|
|
const greg_t bp = mctx->gregs[REG_EBP];
|
|
# endif
|
|
# elif defined __APPLE__
|
|
# if defined(__LP64__)
|
|
const uintptr_t sp = mctx->__ss.__rsp;
|
|
const uintptr_t bp = mctx->__ss.__rbp;
|
|
# else
|
|
const uintptr_t sp = mctx->__ss.__esp;
|
|
const uintptr_t bp = mctx->__ss.__ebp;
|
|
# endif
|
|
# elif defined __FreeBSD__
|
|
# if defined(__amd64__)
|
|
const __register_t sp = mctx->mc_rsp;
|
|
const __register_t bp = mctx->mc_rbp;
|
|
# else
|
|
const __register_t sp = mctx->mc_esp;
|
|
const __register_t bp = mctx->mc_ebp;
|
|
# endif
|
|
# elif defined __HAIKU__
|
|
# if defined(__amd64__)
|
|
const unsigned long sp = mctx->rsp;
|
|
const unsigned long bp = mctx->rbp;
|
|
# else
|
|
const unsigned long sp = mctx->esp;
|
|
const unsigned long bp = mctx->ebp;
|
|
# endif
|
|
# endif
|
|
enum {pagesize = 4096};
|
|
const uintptr_t sp_page = (uintptr_t)sp / pagesize;
|
|
const uintptr_t bp_page = (uintptr_t)bp / pagesize;
|
|
const uintptr_t fault_page = addr / pagesize;
|
|
|
|
/* SP in ucontext is not decremented yet when `push` failed, so
|
|
* the fault page can be the next. */
|
|
if (sp_page == fault_page || sp_page == fault_page + 1 ||
|
|
sp_page <= fault_page && fault_page <= bp_page) {
|
|
rb_execution_context_t *ec = GET_EC();
|
|
int crit = FALSE;
|
|
if ((uintptr_t)ec->tag->buf / pagesize <= fault_page + 1) {
|
|
/* drop the last tag if it is close to the fault,
|
|
* otherwise it can cause stack overflow again at the same
|
|
* place. */
|
|
ec->tag = ec->tag->prev;
|
|
crit = TRUE;
|
|
}
|
|
reset_sigmask(sig);
|
|
rb_ec_stack_overflow(ec, crit);
|
|
}
|
|
}
|
|
# else
|
|
static void
|
|
check_stack_overflow(int sig, const void *addr)
|
|
{
|
|
int ruby_stack_overflowed_p(const rb_thread_t *, const void *);
|
|
rb_thread_t *th = GET_THREAD();
|
|
if (ruby_stack_overflowed_p(th, addr)) {
|
|
reset_sigmask(sig);
|
|
rb_ec_stack_overflow(th->ec, FALSE);
|
|
}
|
|
}
|
|
# endif
|
|
# ifdef _WIN32
|
|
# define CHECK_STACK_OVERFLOW() check_stack_overflow(sig, 0)
|
|
# else
|
|
# define FAULT_ADDRESS info->si_addr
|
|
# ifdef USE_UCONTEXT_REG
|
|
# define CHECK_STACK_OVERFLOW() check_stack_overflow(sig, (uintptr_t)FAULT_ADDRESS, ctx)
|
|
# else
|
|
# define CHECK_STACK_OVERFLOW() check_stack_overflow(sig, FAULT_ADDRESS)
|
|
# endif
|
|
# define MESSAGE_FAULT_ADDRESS " at %p", FAULT_ADDRESS
|
|
# endif
|
|
#else
|
|
# define CHECK_STACK_OVERFLOW() (void)0
|
|
#endif
|
|
#ifndef MESSAGE_FAULT_ADDRESS
|
|
# define MESSAGE_FAULT_ADDRESS
|
|
#endif
|
|
|
|
#if defined SIGSEGV || defined SIGBUS || defined SIGILL || defined SIGFPE
|
|
NOINLINE(static void check_reserved_signal_(const char *name, size_t name_len));
|
|
/* noinine to reduce stack usage in signal handers */
|
|
|
|
#define check_reserved_signal(name) check_reserved_signal_(name, sizeof(name)-1)
|
|
|
|
#ifdef SIGBUS
|
|
|
|
NORETURN(static ruby_sigaction_t sigbus);
|
|
|
|
static RETSIGTYPE
|
|
sigbus(int sig SIGINFO_ARG)
|
|
{
|
|
check_reserved_signal("BUS");
|
|
/*
|
|
* Mac OS X makes KERN_PROTECTION_FAILURE when thread touch guard page.
|
|
* and it's delivered as SIGBUS instead of SIGSEGV to userland. It's crazy
|
|
* wrong IMHO. but anyway we have to care it. Sigh.
|
|
*/
|
|
/* Seems Linux also delivers SIGBUS. */
|
|
#if defined __APPLE__ || defined __linux__
|
|
CHECK_STACK_OVERFLOW();
|
|
#endif
|
|
rb_bug_context(SIGINFO_CTX, "Bus Error" MESSAGE_FAULT_ADDRESS);
|
|
}
|
|
#endif
|
|
|
|
NORETURN(static void ruby_abort(void));
|
|
|
|
static void
|
|
ruby_abort(void)
|
|
{
|
|
#ifdef __sun
|
|
/* Solaris's abort() is async signal unsafe. Of course, it is not
|
|
* POSIX compliant.
|
|
*/
|
|
raise(SIGABRT);
|
|
#else
|
|
abort();
|
|
#endif
|
|
|
|
}
|
|
|
|
#ifdef SIGSEGV
|
|
|
|
NORETURN(static ruby_sigaction_t sigsegv);
|
|
|
|
static RETSIGTYPE
|
|
sigsegv(int sig SIGINFO_ARG)
|
|
{
|
|
check_reserved_signal("SEGV");
|
|
CHECK_STACK_OVERFLOW();
|
|
rb_bug_context(SIGINFO_CTX, "Segmentation fault" MESSAGE_FAULT_ADDRESS);
|
|
}
|
|
#endif
|
|
|
|
#ifdef SIGILL
|
|
|
|
NORETURN(static ruby_sigaction_t sigill);
|
|
|
|
static RETSIGTYPE
|
|
sigill(int sig SIGINFO_ARG)
|
|
{
|
|
check_reserved_signal("ILL");
|
|
#if defined __APPLE__
|
|
CHECK_STACK_OVERFLOW();
|
|
#endif
|
|
rb_bug_context(SIGINFO_CTX, "Illegal instruction" MESSAGE_FAULT_ADDRESS);
|
|
}
|
|
#endif
|
|
|
|
static void
|
|
check_reserved_signal_(const char *name, size_t name_len)
|
|
{
|
|
const char *prev = ATOMIC_PTR_EXCHANGE(received_signal, name);
|
|
|
|
if (prev) {
|
|
ssize_t RB_UNUSED_VAR(err);
|
|
#define NOZ(name, str) name[sizeof(str)-1] = str
|
|
static const char NOZ(msg1, " received in ");
|
|
static const char NOZ(msg2, " handler\n");
|
|
|
|
#ifdef HAVE_WRITEV
|
|
struct iovec iov[4];
|
|
|
|
iov[0].iov_base = (void *)name;
|
|
iov[0].iov_len = name_len;
|
|
iov[1].iov_base = (void *)msg1;
|
|
iov[1].iov_len = sizeof(msg1);
|
|
iov[2].iov_base = (void *)prev;
|
|
iov[2].iov_len = strlen(prev);
|
|
iov[3].iov_base = (void *)msg2;
|
|
iov[3].iov_len = sizeof(msg2);
|
|
err = writev(2, iov, 4);
|
|
#else
|
|
err = write(2, name, name_len);
|
|
err = write(2, msg1, sizeof(msg1));
|
|
err = write(2, prev, strlen(prev));
|
|
err = write(2, msg2, sizeof(msg2));
|
|
#endif
|
|
ruby_abort();
|
|
}
|
|
|
|
ruby_disable_gc = 1;
|
|
}
|
|
#endif
|
|
|
|
#if defined SIGPIPE || defined SIGSYS
|
|
static RETSIGTYPE
|
|
sig_do_nothing(int sig)
|
|
{
|
|
}
|
|
#endif
|
|
|
|
static void
|
|
signal_exec(VALUE cmd, int safe, int sig)
|
|
{
|
|
rb_execution_context_t *ec = GET_EC();
|
|
volatile rb_atomic_t old_interrupt_mask = ec->interrupt_mask;
|
|
enum ruby_tag_type state;
|
|
|
|
/*
|
|
* workaround the following race:
|
|
* 1. signal_enque queues signal for execution
|
|
* 2. user calls trap(sig, "IGNORE"), setting SIG_IGN
|
|
* 3. rb_signal_exec runs on queued signal
|
|
*/
|
|
if (IMMEDIATE_P(cmd))
|
|
return;
|
|
|
|
ec->interrupt_mask |= TRAP_INTERRUPT_MASK;
|
|
EC_PUSH_TAG(ec);
|
|
if ((state = EC_EXEC_TAG()) == TAG_NONE) {
|
|
VALUE signum = INT2NUM(sig);
|
|
rb_eval_cmd(cmd, rb_ary_new3(1, signum), safe);
|
|
}
|
|
EC_POP_TAG();
|
|
ec = GET_EC();
|
|
ec->interrupt_mask = old_interrupt_mask;
|
|
|
|
if (state) {
|
|
/* XXX: should be replaced with rb_threadptr_pending_interrupt_enque() */
|
|
EC_JUMP_TAG(ec, state);
|
|
}
|
|
}
|
|
|
|
void
|
|
rb_trap_exit(void)
|
|
{
|
|
rb_vm_t *vm = GET_VM();
|
|
VALUE trap_exit = vm->trap_list.cmd[0];
|
|
|
|
if (trap_exit) {
|
|
vm->trap_list.cmd[0] = 0;
|
|
signal_exec(trap_exit, vm->trap_list.safe[0], 0);
|
|
}
|
|
}
|
|
|
|
void ruby_waitpid_all(rb_vm_t *); /* process.c */
|
|
|
|
void
|
|
ruby_sigchld_handler(rb_vm_t *vm)
|
|
{
|
|
if (SIGCHLD_LOSSY || ATOMIC_EXCHANGE(sigchld_hit, 0)) {
|
|
ruby_waitpid_all(vm);
|
|
}
|
|
}
|
|
|
|
void
|
|
rb_signal_exec(rb_thread_t *th, int sig)
|
|
{
|
|
rb_vm_t *vm = GET_VM();
|
|
VALUE cmd = vm->trap_list.cmd[sig];
|
|
int safe = vm->trap_list.safe[sig];
|
|
|
|
if (cmd == 0) {
|
|
switch (sig) {
|
|
case SIGINT:
|
|
rb_interrupt();
|
|
break;
|
|
#ifdef SIGHUP
|
|
case SIGHUP:
|
|
#endif
|
|
#ifdef SIGQUIT
|
|
case SIGQUIT:
|
|
#endif
|
|
#ifdef SIGTERM
|
|
case SIGTERM:
|
|
#endif
|
|
#ifdef SIGALRM
|
|
case SIGALRM:
|
|
#endif
|
|
#ifdef SIGUSR1
|
|
case SIGUSR1:
|
|
#endif
|
|
#ifdef SIGUSR2
|
|
case SIGUSR2:
|
|
#endif
|
|
rb_threadptr_signal_raise(th, sig);
|
|
break;
|
|
}
|
|
}
|
|
else if (cmd == Qundef) {
|
|
rb_threadptr_signal_exit(th);
|
|
}
|
|
else {
|
|
signal_exec(cmd, safe, sig);
|
|
}
|
|
}
|
|
|
|
static sighandler_t
|
|
default_handler(int sig)
|
|
{
|
|
sighandler_t func;
|
|
switch (sig) {
|
|
case SIGINT:
|
|
#ifdef SIGHUP
|
|
case SIGHUP:
|
|
#endif
|
|
#ifdef SIGQUIT
|
|
case SIGQUIT:
|
|
#endif
|
|
#ifdef SIGTERM
|
|
case SIGTERM:
|
|
#endif
|
|
#ifdef SIGALRM
|
|
case SIGALRM:
|
|
#endif
|
|
#ifdef SIGUSR1
|
|
case SIGUSR1:
|
|
#endif
|
|
#ifdef SIGUSR2
|
|
case SIGUSR2:
|
|
#endif
|
|
#if RUBY_SIGCHLD
|
|
case RUBY_SIGCHLD:
|
|
#endif
|
|
func = sighandler;
|
|
break;
|
|
#ifdef SIGBUS
|
|
case SIGBUS:
|
|
func = (sighandler_t)sigbus;
|
|
break;
|
|
#endif
|
|
#ifdef SIGSEGV
|
|
case SIGSEGV:
|
|
func = (sighandler_t)sigsegv;
|
|
break;
|
|
#endif
|
|
#ifdef SIGPIPE
|
|
case SIGPIPE:
|
|
func = sig_do_nothing;
|
|
break;
|
|
#endif
|
|
#ifdef SIGSYS
|
|
case SIGSYS:
|
|
func = sig_do_nothing;
|
|
break;
|
|
#endif
|
|
default:
|
|
func = SIG_DFL;
|
|
break;
|
|
}
|
|
|
|
return func;
|
|
}
|
|
|
|
static sighandler_t
|
|
trap_handler(VALUE *cmd, int sig)
|
|
{
|
|
sighandler_t func = sighandler;
|
|
VALUE command;
|
|
|
|
if (NIL_P(*cmd)) {
|
|
func = SIG_IGN;
|
|
}
|
|
else {
|
|
command = rb_check_string_type(*cmd);
|
|
if (NIL_P(command) && SYMBOL_P(*cmd)) {
|
|
command = rb_sym2str(*cmd);
|
|
if (!command) rb_raise(rb_eArgError, "bad handler");
|
|
}
|
|
if (!NIL_P(command)) {
|
|
const char *cptr;
|
|
long len;
|
|
SafeStringValue(command); /* taint check */
|
|
*cmd = command;
|
|
RSTRING_GETMEM(command, cptr, len);
|
|
switch (len) {
|
|
case 0:
|
|
goto sig_ign;
|
|
break;
|
|
case 14:
|
|
if (memcmp(cptr, "SYSTEM_DEFAULT", 14) == 0) {
|
|
if (sig == RUBY_SIGCHLD) {
|
|
goto sig_dfl;
|
|
}
|
|
func = SIG_DFL;
|
|
*cmd = 0;
|
|
}
|
|
break;
|
|
case 7:
|
|
if (memcmp(cptr, "SIG_IGN", 7) == 0) {
|
|
sig_ign:
|
|
func = SIG_IGN;
|
|
*cmd = Qtrue;
|
|
}
|
|
else if (memcmp(cptr, "SIG_DFL", 7) == 0) {
|
|
sig_dfl:
|
|
func = default_handler(sig);
|
|
*cmd = 0;
|
|
}
|
|
else if (memcmp(cptr, "DEFAULT", 7) == 0) {
|
|
goto sig_dfl;
|
|
}
|
|
break;
|
|
case 6:
|
|
if (memcmp(cptr, "IGNORE", 6) == 0) {
|
|
goto sig_ign;
|
|
}
|
|
break;
|
|
case 4:
|
|
if (memcmp(cptr, "EXIT", 4) == 0) {
|
|
*cmd = Qundef;
|
|
}
|
|
break;
|
|
}
|
|
}
|
|
else {
|
|
rb_proc_t *proc;
|
|
GetProcPtr(*cmd, proc);
|
|
(void)proc;
|
|
}
|
|
}
|
|
|
|
return func;
|
|
}
|
|
|
|
static int
|
|
trap_signm(VALUE vsig)
|
|
{
|
|
int sig = -1;
|
|
|
|
if (FIXNUM_P(vsig)) {
|
|
sig = FIX2INT(vsig);
|
|
if (sig < 0 || sig >= NSIG) {
|
|
rb_raise(rb_eArgError, "invalid signal number (%d)", sig);
|
|
}
|
|
}
|
|
else {
|
|
sig = signm2signo(&vsig, FALSE, TRUE, NULL);
|
|
}
|
|
return sig;
|
|
}
|
|
|
|
static VALUE
|
|
trap(int sig, sighandler_t func, VALUE command)
|
|
{
|
|
sighandler_t oldfunc;
|
|
VALUE oldcmd;
|
|
rb_vm_t *vm = GET_VM();
|
|
|
|
/*
|
|
* Be careful. ruby_signal() and trap_list.cmd[sig] must be changed
|
|
* atomically. In current implementation, we only need to don't call
|
|
* RUBY_VM_CHECK_INTS().
|
|
*/
|
|
if (sig == 0) {
|
|
oldfunc = SIG_ERR;
|
|
}
|
|
else {
|
|
oldfunc = ruby_signal(sig, func);
|
|
if (oldfunc == SIG_ERR) rb_sys_fail_str(rb_signo2signm(sig));
|
|
}
|
|
oldcmd = vm->trap_list.cmd[sig];
|
|
switch (oldcmd) {
|
|
case 0:
|
|
case Qtrue:
|
|
if (oldfunc == SIG_IGN) oldcmd = rb_str_new2("IGNORE");
|
|
else if (oldfunc == SIG_DFL) oldcmd = rb_str_new2("SYSTEM_DEFAULT");
|
|
else if (oldfunc == sighandler) oldcmd = rb_str_new2("DEFAULT");
|
|
else oldcmd = Qnil;
|
|
break;
|
|
case Qnil:
|
|
break;
|
|
case Qundef:
|
|
oldcmd = rb_str_new2("EXIT");
|
|
break;
|
|
}
|
|
|
|
ACCESS_ONCE(VALUE, vm->trap_list.cmd[sig]) = command;
|
|
vm->trap_list.safe[sig] = rb_safe_level();
|
|
|
|
return oldcmd;
|
|
}
|
|
|
|
static int
|
|
reserved_signal_p(int signo)
|
|
{
|
|
/* Synchronous signal can't deliver to main thread */
|
|
#ifdef SIGSEGV
|
|
if (signo == SIGSEGV)
|
|
return 1;
|
|
#endif
|
|
#ifdef SIGBUS
|
|
if (signo == SIGBUS)
|
|
return 1;
|
|
#endif
|
|
#ifdef SIGILL
|
|
if (signo == SIGILL)
|
|
return 1;
|
|
#endif
|
|
#ifdef SIGFPE
|
|
if (signo == SIGFPE)
|
|
return 1;
|
|
#endif
|
|
|
|
/* used ubf internal see thread_pthread.c. */
|
|
#ifdef SIGVTALRM
|
|
if (signo == SIGVTALRM)
|
|
return 1;
|
|
#endif
|
|
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* Signal.trap( signal, command ) -> obj
|
|
* Signal.trap( signal ) {| | block } -> obj
|
|
*
|
|
* Specifies the handling of signals. The first parameter is a signal
|
|
* name (a string such as ``SIGALRM'', ``SIGUSR1'', and so on) or a
|
|
* signal number. The characters ``SIG'' may be omitted from the
|
|
* signal name. The command or block specifies code to be run when the
|
|
* signal is raised.
|
|
* If the command is the string ``IGNORE'' or ``SIG_IGN'', the signal
|
|
* will be ignored.
|
|
* If the command is ``DEFAULT'' or ``SIG_DFL'', the Ruby's default handler
|
|
* will be invoked.
|
|
* If the command is ``EXIT'', the script will be terminated by the signal.
|
|
* If the command is ``SYSTEM_DEFAULT'', the operating system's default
|
|
* handler will be invoked.
|
|
* Otherwise, the given command or block will be run.
|
|
* The special signal name ``EXIT'' or signal number zero will be
|
|
* invoked just prior to program termination.
|
|
* trap returns the previous handler for the given signal.
|
|
*
|
|
* Signal.trap(0, proc { puts "Terminating: #{$$}" })
|
|
* Signal.trap("CLD") { puts "Child died" }
|
|
* fork && Process.wait
|
|
*
|
|
* produces:
|
|
* Terminating: 27461
|
|
* Child died
|
|
* Terminating: 27460
|
|
*/
|
|
static VALUE
|
|
sig_trap(int argc, VALUE *argv)
|
|
{
|
|
int sig;
|
|
sighandler_t func;
|
|
VALUE cmd;
|
|
|
|
rb_check_arity(argc, 1, 2);
|
|
|
|
sig = trap_signm(argv[0]);
|
|
if (reserved_signal_p(sig)) {
|
|
const char *name = signo2signm(sig);
|
|
if (name)
|
|
rb_raise(rb_eArgError, "can't trap reserved signal: SIG%s", name);
|
|
else
|
|
rb_raise(rb_eArgError, "can't trap reserved signal: %d", sig);
|
|
}
|
|
|
|
if (argc == 1) {
|
|
cmd = rb_block_proc();
|
|
func = sighandler;
|
|
}
|
|
else {
|
|
cmd = argv[1];
|
|
func = trap_handler(&cmd, sig);
|
|
}
|
|
|
|
if (OBJ_TAINTED(cmd)) {
|
|
rb_raise(rb_eSecurityError, "Insecure: tainted signal trap");
|
|
}
|
|
|
|
return trap(sig, func, cmd);
|
|
}
|
|
|
|
/*
|
|
* call-seq:
|
|
* Signal.list -> a_hash
|
|
*
|
|
* Returns a list of signal names mapped to the corresponding
|
|
* underlying signal numbers.
|
|
*
|
|
* Signal.list #=> {"EXIT"=>0, "HUP"=>1, "INT"=>2, "QUIT"=>3, "ILL"=>4, "TRAP"=>5, "IOT"=>6, "ABRT"=>6, "FPE"=>8, "KILL"=>9, "BUS"=>7, "SEGV"=>11, "SYS"=>31, "PIPE"=>13, "ALRM"=>14, "TERM"=>15, "URG"=>23, "STOP"=>19, "TSTP"=>20, "CONT"=>18, "CHLD"=>17, "CLD"=>17, "TTIN"=>21, "TTOU"=>22, "IO"=>29, "XCPU"=>24, "XFSZ"=>25, "VTALRM"=>26, "PROF"=>27, "WINCH"=>28, "USR1"=>10, "USR2"=>12, "PWR"=>30, "POLL"=>29}
|
|
*/
|
|
static VALUE
|
|
sig_list(void)
|
|
{
|
|
VALUE h = rb_hash_new();
|
|
const struct signals *sigs;
|
|
|
|
FOREACH_SIGNAL(sigs, 0) {
|
|
rb_hash_aset(h, rb_fstring_cstr(sigs->signm), INT2FIX(sigs->signo));
|
|
}
|
|
return h;
|
|
}
|
|
|
|
#define INSTALL_SIGHANDLER(cond, signame, signum) do { \
|
|
static const char failed[] = "failed to install "signame" handler"; \
|
|
if (!(cond)) break; \
|
|
if (reserved_signal_p(signum)) rb_bug(failed); \
|
|
perror(failed); \
|
|
} while (0)
|
|
static int
|
|
install_sighandler(int signum, sighandler_t handler)
|
|
{
|
|
sighandler_t old;
|
|
|
|
old = ruby_signal(signum, handler);
|
|
if (old == SIG_ERR) return -1;
|
|
/* signal handler should be inherited during exec. */
|
|
if (old != SIG_DFL) {
|
|
ruby_signal(signum, old);
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
# define install_sighandler(signum, handler) \
|
|
INSTALL_SIGHANDLER(install_sighandler(signum, handler), #signum, signum)
|
|
|
|
#if RUBY_SIGCHLD
|
|
static int
|
|
init_sigchld(int sig)
|
|
{
|
|
sighandler_t oldfunc;
|
|
sighandler_t func = sighandler;
|
|
|
|
oldfunc = ruby_signal(sig, SIG_DFL);
|
|
if (oldfunc == SIG_ERR) return -1;
|
|
ruby_signal(sig, func);
|
|
ACCESS_ONCE(VALUE, GET_VM()->trap_list.cmd[sig]) = 0;
|
|
|
|
return 0;
|
|
}
|
|
|
|
# define init_sigchld(signum) \
|
|
INSTALL_SIGHANDLER(init_sigchld(signum), #signum, signum)
|
|
#endif
|
|
|
|
void
|
|
ruby_sig_finalize(void)
|
|
{
|
|
sighandler_t oldfunc;
|
|
|
|
oldfunc = ruby_signal(SIGINT, SIG_IGN);
|
|
if (oldfunc == sighandler) {
|
|
ruby_signal(SIGINT, SIG_DFL);
|
|
}
|
|
}
|
|
|
|
|
|
int ruby_enable_coredump = 0;
|
|
|
|
/*
|
|
* Many operating systems allow signals to be sent to running
|
|
* processes. Some signals have a defined effect on the process, while
|
|
* others may be trapped at the code level and acted upon. For
|
|
* example, your process may trap the USR1 signal and use it to toggle
|
|
* debugging, and may use TERM to initiate a controlled shutdown.
|
|
*
|
|
* pid = fork do
|
|
* Signal.trap("USR1") do
|
|
* $debug = !$debug
|
|
* puts "Debug now: #$debug"
|
|
* end
|
|
* Signal.trap("TERM") do
|
|
* puts "Terminating..."
|
|
* shutdown()
|
|
* end
|
|
* # . . . do some work . . .
|
|
* end
|
|
*
|
|
* Process.detach(pid)
|
|
*
|
|
* # Controlling program:
|
|
* Process.kill("USR1", pid)
|
|
* # ...
|
|
* Process.kill("USR1", pid)
|
|
* # ...
|
|
* Process.kill("TERM", pid)
|
|
*
|
|
* produces:
|
|
* Debug now: true
|
|
* Debug now: false
|
|
* Terminating...
|
|
*
|
|
* The list of available signal names and their interpretation is
|
|
* system dependent. Signal delivery semantics may also vary between
|
|
* systems; in particular signal delivery may not always be reliable.
|
|
*/
|
|
void
|
|
Init_signal(void)
|
|
{
|
|
VALUE mSignal = rb_define_module("Signal");
|
|
|
|
rb_define_global_function("trap", sig_trap, -1);
|
|
rb_define_module_function(mSignal, "trap", sig_trap, -1);
|
|
rb_define_module_function(mSignal, "list", sig_list, 0);
|
|
rb_define_module_function(mSignal, "signame", sig_signame, 1);
|
|
|
|
rb_define_method(rb_eSignal, "initialize", esignal_init, -1);
|
|
rb_define_method(rb_eSignal, "signo", esignal_signo, 0);
|
|
rb_alias(rb_eSignal, rb_intern_const("signm"), rb_intern_const("message"));
|
|
rb_define_method(rb_eInterrupt, "initialize", interrupt_init, -1);
|
|
|
|
/* At this time, there is no subthread. Then sigmask guarantee atomics. */
|
|
rb_disable_interrupt();
|
|
|
|
install_sighandler(SIGINT, sighandler);
|
|
#ifdef SIGHUP
|
|
install_sighandler(SIGHUP, sighandler);
|
|
#endif
|
|
#ifdef SIGQUIT
|
|
install_sighandler(SIGQUIT, sighandler);
|
|
#endif
|
|
#ifdef SIGTERM
|
|
install_sighandler(SIGTERM, sighandler);
|
|
#endif
|
|
#ifdef SIGALRM
|
|
install_sighandler(SIGALRM, sighandler);
|
|
#endif
|
|
#ifdef SIGUSR1
|
|
install_sighandler(SIGUSR1, sighandler);
|
|
#endif
|
|
#ifdef SIGUSR2
|
|
install_sighandler(SIGUSR2, sighandler);
|
|
#endif
|
|
|
|
if (!ruby_enable_coredump) {
|
|
#ifdef SIGBUS
|
|
install_sighandler(SIGBUS, (sighandler_t)sigbus);
|
|
#endif
|
|
#ifdef SIGILL
|
|
install_sighandler(SIGILL, (sighandler_t)sigill);
|
|
#endif
|
|
#ifdef SIGSEGV
|
|
RB_ALTSTACK_INIT(GET_VM()->main_altstack);
|
|
install_sighandler(SIGSEGV, (sighandler_t)sigsegv);
|
|
#endif
|
|
}
|
|
#ifdef SIGPIPE
|
|
install_sighandler(SIGPIPE, sig_do_nothing);
|
|
#endif
|
|
#ifdef SIGSYS
|
|
install_sighandler(SIGSYS, sig_do_nothing);
|
|
#endif
|
|
|
|
#if RUBY_SIGCHLD
|
|
init_sigchld(RUBY_SIGCHLD);
|
|
#endif
|
|
|
|
rb_enable_interrupt();
|
|
}
|
|
|
|
#if defined(HAVE_GRANTPT)
|
|
extern int grantpt(int);
|
|
#else
|
|
static int
|
|
fake_grantfd(int masterfd)
|
|
{
|
|
errno = ENOSYS;
|
|
return -1;
|
|
}
|
|
#define grantpt(fd) fake_grantfd(fd)
|
|
#endif
|
|
|
|
int
|
|
rb_grantpt(int masterfd)
|
|
{
|
|
if (RUBY_SIGCHLD) {
|
|
rb_vm_t *vm = GET_VM();
|
|
int ret, e;
|
|
|
|
/*
|
|
* Prevent waitpid calls from Ruby by taking waitpid_lock.
|
|
* Pedantically, grantpt(3) is undefined if a non-default
|
|
* SIGCHLD handler is defined, but preventing conflicting
|
|
* waitpid calls ought to be sufficient.
|
|
*
|
|
* We could install the default sighandler temporarily, but that
|
|
* could cause SIGCHLD to be missed by other threads. Blocking
|
|
* SIGCHLD won't work here, either, unless we stop and restart
|
|
* timer-thread (as only timer-thread sees SIGCHLD), but that
|
|
* seems like overkill.
|
|
*/
|
|
rb_nativethread_lock_lock(&vm->waitpid_lock);
|
|
{
|
|
ret = grantpt(masterfd); /* may spawn `pt_chown' and wait on it */
|
|
if (ret < 0) e = errno;
|
|
}
|
|
rb_nativethread_lock_unlock(&vm->waitpid_lock);
|
|
|
|
if (ret < 0) errno = e;
|
|
return ret;
|
|
}
|
|
else {
|
|
return grantpt(masterfd);
|
|
}
|
|
}
|