1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * linux/kernel/signal.c
4 *
5 * Copyright (C) 1991, 1992 Linus Torvalds
6 *
7 * 1997-11-02 Modified for POSIX.1b signals by Richard Henderson
8 *
9 * 2003-06-02 Jim Houston - Concurrent Computer Corp.
10 * Changes to use preallocated sigqueue structures
11 * to allow signals to be sent reliably.
12 */
13
14 #include <linux/slab.h>
15 #include <linux/export.h>
16 #include <linux/init.h>
17 #include <linux/sched/mm.h>
18 #include <linux/sched/user.h>
19 #include <linux/sched/debug.h>
20 #include <linux/sched/task.h>
21 #include <linux/sched/task_stack.h>
22 #include <linux/sched/cputime.h>
23 #include <linux/file.h>
24 #include <linux/fs.h>
25 #include <linux/mm.h>
26 #include <linux/proc_fs.h>
27 #include <linux/tty.h>
28 #include <linux/binfmts.h>
29 #include <linux/coredump.h>
30 #include <linux/security.h>
31 #include <linux/syscalls.h>
32 #include <linux/ptrace.h>
33 #include <linux/signal.h>
34 #include <linux/signalfd.h>
35 #include <linux/ratelimit.h>
36 #include <linux/task_work.h>
37 #include <linux/capability.h>
38 #include <linux/freezer.h>
39 #include <linux/pid_namespace.h>
40 #include <linux/nsproxy.h>
41 #include <linux/user_namespace.h>
42 #include <linux/uprobes.h>
43 #include <linux/compat.h>
44 #include <linux/cn_proc.h>
45 #include <linux/compiler.h>
46 #include <linux/posix-timers.h>
47 #include <linux/cgroup.h>
48 #include <linux/audit.h>
49 #include <linux/sysctl.h>
50 #include <uapi/linux/pidfd.h>
51
52 #define CREATE_TRACE_POINTS
53 #include <trace/events/signal.h>
54
55 #include <asm/param.h>
56 #include <linux/uaccess.h>
57 #include <asm/unistd.h>
58 #include <asm/siginfo.h>
59 #include <asm/cacheflush.h>
60 #include <asm/syscall.h> /* for syscall_get_* */
61
62 /*
63 * SLAB caches for signal bits.
64 */
65
66 static struct kmem_cache *sigqueue_cachep;
67
68 int print_fatal_signals __read_mostly;
69
sig_handler(struct task_struct * t,int sig)70 static void __user *sig_handler(struct task_struct *t, int sig)
71 {
72 return t->sighand->action[sig - 1].sa.sa_handler;
73 }
74
sig_handler_ignored(void __user * handler,int sig)75 static inline bool sig_handler_ignored(void __user *handler, int sig)
76 {
77 /* Is it explicitly or implicitly ignored? */
78 return handler == SIG_IGN ||
79 (handler == SIG_DFL && sig_kernel_ignore(sig));
80 }
81
sig_task_ignored(struct task_struct * t,int sig,bool force)82 static bool sig_task_ignored(struct task_struct *t, int sig, bool force)
83 {
84 void __user *handler;
85
86 handler = sig_handler(t, sig);
87
88 /* SIGKILL and SIGSTOP may not be sent to the global init */
89 if (unlikely(is_global_init(t) && sig_kernel_only(sig)))
90 return true;
91
92 if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
93 handler == SIG_DFL && !(force && sig_kernel_only(sig)))
94 return true;
95
96 /* Only allow kernel generated signals to this kthread */
97 if (unlikely((t->flags & PF_KTHREAD) &&
98 (handler == SIG_KTHREAD_KERNEL) && !force))
99 return true;
100
101 return sig_handler_ignored(handler, sig);
102 }
103
sig_ignored(struct task_struct * t,int sig,bool force)104 static bool sig_ignored(struct task_struct *t, int sig, bool force)
105 {
106 /*
107 * Blocked signals are never ignored, since the
108 * signal handler may change by the time it is
109 * unblocked.
110 */
111 if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
112 return false;
113
114 /*
115 * Tracers may want to know about even ignored signal unless it
116 * is SIGKILL which can't be reported anyway but can be ignored
117 * by SIGNAL_UNKILLABLE task.
118 */
119 if (t->ptrace && sig != SIGKILL)
120 return false;
121
122 return sig_task_ignored(t, sig, force);
123 }
124
125 /*
126 * Re-calculate pending state from the set of locally pending
127 * signals, globally pending signals, and blocked signals.
128 */
has_pending_signals(sigset_t * signal,sigset_t * blocked)129 static inline bool has_pending_signals(sigset_t *signal, sigset_t *blocked)
130 {
131 unsigned long ready;
132 long i;
133
134 switch (_NSIG_WORDS) {
135 default:
136 for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
137 ready |= signal->sig[i] &~ blocked->sig[i];
138 break;
139
140 case 4: ready = signal->sig[3] &~ blocked->sig[3];
141 ready |= signal->sig[2] &~ blocked->sig[2];
142 ready |= signal->sig[1] &~ blocked->sig[1];
143 ready |= signal->sig[0] &~ blocked->sig[0];
144 break;
145
146 case 2: ready = signal->sig[1] &~ blocked->sig[1];
147 ready |= signal->sig[0] &~ blocked->sig[0];
148 break;
149
150 case 1: ready = signal->sig[0] &~ blocked->sig[0];
151 }
152 return ready != 0;
153 }
154
155 #define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
156
recalc_sigpending_tsk(struct task_struct * t)157 static bool recalc_sigpending_tsk(struct task_struct *t)
158 {
159 if ((t->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) ||
160 PENDING(&t->pending, &t->blocked) ||
161 PENDING(&t->signal->shared_pending, &t->blocked) ||
162 cgroup_task_frozen(t)) {
163 set_tsk_thread_flag(t, TIF_SIGPENDING);
164 return true;
165 }
166
167 /*
168 * We must never clear the flag in another thread, or in current
169 * when it's possible the current syscall is returning -ERESTART*.
170 * So we don't clear it here, and only callers who know they should do.
171 */
172 return false;
173 }
174
recalc_sigpending(void)175 void recalc_sigpending(void)
176 {
177 if (!recalc_sigpending_tsk(current) && !freezing(current))
178 clear_thread_flag(TIF_SIGPENDING);
179
180 }
181 EXPORT_SYMBOL(recalc_sigpending);
182
calculate_sigpending(void)183 void calculate_sigpending(void)
184 {
185 /* Have any signals or users of TIF_SIGPENDING been delayed
186 * until after fork?
187 */
188 spin_lock_irq(¤t->sighand->siglock);
189 set_tsk_thread_flag(current, TIF_SIGPENDING);
190 recalc_sigpending();
191 spin_unlock_irq(¤t->sighand->siglock);
192 }
193
194 /* Given the mask, find the first available signal that should be serviced. */
195
196 #define SYNCHRONOUS_MASK \
197 (sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
198 sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
199
next_signal(struct sigpending * pending,sigset_t * mask)200 int next_signal(struct sigpending *pending, sigset_t *mask)
201 {
202 unsigned long i, *s, *m, x;
203 int sig = 0;
204
205 s = pending->signal.sig;
206 m = mask->sig;
207
208 /*
209 * Handle the first word specially: it contains the
210 * synchronous signals that need to be dequeued first.
211 */
212 x = *s &~ *m;
213 if (x) {
214 if (x & SYNCHRONOUS_MASK)
215 x &= SYNCHRONOUS_MASK;
216 sig = ffz(~x) + 1;
217 return sig;
218 }
219
220 switch (_NSIG_WORDS) {
221 default:
222 for (i = 1; i < _NSIG_WORDS; ++i) {
223 x = *++s &~ *++m;
224 if (!x)
225 continue;
226 sig = ffz(~x) + i*_NSIG_BPW + 1;
227 break;
228 }
229 break;
230
231 case 2:
232 x = s[1] &~ m[1];
233 if (!x)
234 break;
235 sig = ffz(~x) + _NSIG_BPW + 1;
236 break;
237
238 case 1:
239 /* Nothing to do */
240 break;
241 }
242
243 return sig;
244 }
245
print_dropped_signal(int sig)246 static inline void print_dropped_signal(int sig)
247 {
248 static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
249
250 if (!print_fatal_signals)
251 return;
252
253 if (!__ratelimit(&ratelimit_state))
254 return;
255
256 pr_info("%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
257 current->comm, current->pid, sig);
258 }
259
260 /**
261 * task_set_jobctl_pending - set jobctl pending bits
262 * @task: target task
263 * @mask: pending bits to set
264 *
265 * Clear @mask from @task->jobctl. @mask must be subset of
266 * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
267 * %JOBCTL_TRAPPING. If stop signo is being set, the existing signo is
268 * cleared. If @task is already being killed or exiting, this function
269 * becomes noop.
270 *
271 * CONTEXT:
272 * Must be called with @task->sighand->siglock held.
273 *
274 * RETURNS:
275 * %true if @mask is set, %false if made noop because @task was dying.
276 */
task_set_jobctl_pending(struct task_struct * task,unsigned long mask)277 bool task_set_jobctl_pending(struct task_struct *task, unsigned long mask)
278 {
279 BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
280 JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
281 BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
282
283 if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
284 return false;
285
286 if (mask & JOBCTL_STOP_SIGMASK)
287 task->jobctl &= ~JOBCTL_STOP_SIGMASK;
288
289 task->jobctl |= mask;
290 return true;
291 }
292
293 /**
294 * task_clear_jobctl_trapping - clear jobctl trapping bit
295 * @task: target task
296 *
297 * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
298 * Clear it and wake up the ptracer. Note that we don't need any further
299 * locking. @task->siglock guarantees that @task->parent points to the
300 * ptracer.
301 *
302 * CONTEXT:
303 * Must be called with @task->sighand->siglock held.
304 */
task_clear_jobctl_trapping(struct task_struct * task)305 void task_clear_jobctl_trapping(struct task_struct *task)
306 {
307 if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
308 task->jobctl &= ~JOBCTL_TRAPPING;
309 smp_mb(); /* advised by wake_up_bit() */
310 wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
311 }
312 }
313
314 /**
315 * task_clear_jobctl_pending - clear jobctl pending bits
316 * @task: target task
317 * @mask: pending bits to clear
318 *
319 * Clear @mask from @task->jobctl. @mask must be subset of
320 * %JOBCTL_PENDING_MASK. If %JOBCTL_STOP_PENDING is being cleared, other
321 * STOP bits are cleared together.
322 *
323 * If clearing of @mask leaves no stop or trap pending, this function calls
324 * task_clear_jobctl_trapping().
325 *
326 * CONTEXT:
327 * Must be called with @task->sighand->siglock held.
328 */
task_clear_jobctl_pending(struct task_struct * task,unsigned long mask)329 void task_clear_jobctl_pending(struct task_struct *task, unsigned long mask)
330 {
331 BUG_ON(mask & ~JOBCTL_PENDING_MASK);
332
333 if (mask & JOBCTL_STOP_PENDING)
334 mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
335
336 task->jobctl &= ~mask;
337
338 if (!(task->jobctl & JOBCTL_PENDING_MASK))
339 task_clear_jobctl_trapping(task);
340 }
341
342 /**
343 * task_participate_group_stop - participate in a group stop
344 * @task: task participating in a group stop
345 *
346 * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
347 * Group stop states are cleared and the group stop count is consumed if
348 * %JOBCTL_STOP_CONSUME was set. If the consumption completes the group
349 * stop, the appropriate `SIGNAL_*` flags are set.
350 *
351 * CONTEXT:
352 * Must be called with @task->sighand->siglock held.
353 *
354 * RETURNS:
355 * %true if group stop completion should be notified to the parent, %false
356 * otherwise.
357 */
task_participate_group_stop(struct task_struct * task)358 static bool task_participate_group_stop(struct task_struct *task)
359 {
360 struct signal_struct *sig = task->signal;
361 bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
362
363 WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
364
365 task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
366
367 if (!consume)
368 return false;
369
370 if (!WARN_ON_ONCE(sig->group_stop_count == 0))
371 sig->group_stop_count--;
372
373 /*
374 * Tell the caller to notify completion iff we are entering into a
375 * fresh group stop. Read comment in do_signal_stop() for details.
376 */
377 if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
378 signal_set_stop_flags(sig, SIGNAL_STOP_STOPPED);
379 return true;
380 }
381 return false;
382 }
383
task_join_group_stop(struct task_struct * task)384 void task_join_group_stop(struct task_struct *task)
385 {
386 unsigned long mask = current->jobctl & JOBCTL_STOP_SIGMASK;
387 struct signal_struct *sig = current->signal;
388
389 if (sig->group_stop_count) {
390 sig->group_stop_count++;
391 mask |= JOBCTL_STOP_CONSUME;
392 } else if (!(sig->flags & SIGNAL_STOP_STOPPED))
393 return;
394
395 /* Have the new thread join an on-going signal group stop */
396 task_set_jobctl_pending(task, mask | JOBCTL_STOP_PENDING);
397 }
398
399 /*
400 * allocate a new signal queue record
401 * - this may be called without locks if and only if t == current, otherwise an
402 * appropriate lock must be held to stop the target task from exiting
403 */
404 static struct sigqueue *
__sigqueue_alloc(int sig,struct task_struct * t,gfp_t gfp_flags,int override_rlimit,const unsigned int sigqueue_flags)405 __sigqueue_alloc(int sig, struct task_struct *t, gfp_t gfp_flags,
406 int override_rlimit, const unsigned int sigqueue_flags)
407 {
408 struct sigqueue *q = NULL;
409 struct ucounts *ucounts;
410 long sigpending;
411
412 /*
413 * Protect access to @t credentials. This can go away when all
414 * callers hold rcu read lock.
415 *
416 * NOTE! A pending signal will hold on to the user refcount,
417 * and we get/put the refcount only when the sigpending count
418 * changes from/to zero.
419 */
420 rcu_read_lock();
421 ucounts = task_ucounts(t);
422 sigpending = inc_rlimit_get_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING,
423 override_rlimit);
424 rcu_read_unlock();
425 if (!sigpending)
426 return NULL;
427
428 if (override_rlimit || likely(sigpending <= task_rlimit(t, RLIMIT_SIGPENDING))) {
429 q = kmem_cache_alloc(sigqueue_cachep, gfp_flags);
430 } else {
431 print_dropped_signal(sig);
432 }
433
434 if (unlikely(q == NULL)) {
435 dec_rlimit_put_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
436 } else {
437 INIT_LIST_HEAD(&q->list);
438 q->flags = sigqueue_flags;
439 q->ucounts = ucounts;
440 }
441 return q;
442 }
443
__sigqueue_free(struct sigqueue * q)444 static void __sigqueue_free(struct sigqueue *q)
445 {
446 if (q->flags & SIGQUEUE_PREALLOC)
447 return;
448 if (q->ucounts) {
449 dec_rlimit_put_ucounts(q->ucounts, UCOUNT_RLIMIT_SIGPENDING);
450 q->ucounts = NULL;
451 }
452 kmem_cache_free(sigqueue_cachep, q);
453 }
454
flush_sigqueue(struct sigpending * queue)455 void flush_sigqueue(struct sigpending *queue)
456 {
457 struct sigqueue *q;
458
459 sigemptyset(&queue->signal);
460 while (!list_empty(&queue->list)) {
461 q = list_entry(queue->list.next, struct sigqueue , list);
462 list_del_init(&q->list);
463 __sigqueue_free(q);
464 }
465 }
466
467 /*
468 * Flush all pending signals for this kthread.
469 */
flush_signals(struct task_struct * t)470 void flush_signals(struct task_struct *t)
471 {
472 unsigned long flags;
473
474 spin_lock_irqsave(&t->sighand->siglock, flags);
475 clear_tsk_thread_flag(t, TIF_SIGPENDING);
476 flush_sigqueue(&t->pending);
477 flush_sigqueue(&t->signal->shared_pending);
478 spin_unlock_irqrestore(&t->sighand->siglock, flags);
479 }
480 EXPORT_SYMBOL(flush_signals);
481
482 #ifdef CONFIG_POSIX_TIMERS
__flush_itimer_signals(struct sigpending * pending)483 static void __flush_itimer_signals(struct sigpending *pending)
484 {
485 sigset_t signal, retain;
486 struct sigqueue *q, *n;
487
488 signal = pending->signal;
489 sigemptyset(&retain);
490
491 list_for_each_entry_safe(q, n, &pending->list, list) {
492 int sig = q->info.si_signo;
493
494 if (likely(q->info.si_code != SI_TIMER)) {
495 sigaddset(&retain, sig);
496 } else {
497 sigdelset(&signal, sig);
498 list_del_init(&q->list);
499 __sigqueue_free(q);
500 }
501 }
502
503 sigorsets(&pending->signal, &signal, &retain);
504 }
505
flush_itimer_signals(void)506 void flush_itimer_signals(void)
507 {
508 struct task_struct *tsk = current;
509 unsigned long flags;
510
511 spin_lock_irqsave(&tsk->sighand->siglock, flags);
512 __flush_itimer_signals(&tsk->pending);
513 __flush_itimer_signals(&tsk->signal->shared_pending);
514 spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
515 }
516 #endif
517
ignore_signals(struct task_struct * t)518 void ignore_signals(struct task_struct *t)
519 {
520 int i;
521
522 for (i = 0; i < _NSIG; ++i)
523 t->sighand->action[i].sa.sa_handler = SIG_IGN;
524
525 flush_signals(t);
526 }
527
528 /*
529 * Flush all handlers for a task.
530 */
531
532 void
flush_signal_handlers(struct task_struct * t,int force_default)533 flush_signal_handlers(struct task_struct *t, int force_default)
534 {
535 int i;
536 struct k_sigaction *ka = &t->sighand->action[0];
537 for (i = _NSIG ; i != 0 ; i--) {
538 if (force_default || ka->sa.sa_handler != SIG_IGN)
539 ka->sa.sa_handler = SIG_DFL;
540 ka->sa.sa_flags = 0;
541 #ifdef __ARCH_HAS_SA_RESTORER
542 ka->sa.sa_restorer = NULL;
543 #endif
544 sigemptyset(&ka->sa.sa_mask);
545 ka++;
546 }
547 }
548
unhandled_signal(struct task_struct * tsk,int sig)549 bool unhandled_signal(struct task_struct *tsk, int sig)
550 {
551 void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
552 if (is_global_init(tsk))
553 return true;
554
555 if (handler != SIG_IGN && handler != SIG_DFL)
556 return false;
557
558 /* If dying, we handle all new signals by ignoring them */
559 if (fatal_signal_pending(tsk))
560 return false;
561
562 /* if ptraced, let the tracer determine */
563 return !tsk->ptrace;
564 }
565
collect_signal(int sig,struct sigpending * list,kernel_siginfo_t * info,bool * resched_timer)566 static void collect_signal(int sig, struct sigpending *list, kernel_siginfo_t *info,
567 bool *resched_timer)
568 {
569 struct sigqueue *q, *first = NULL;
570
571 /*
572 * Collect the siginfo appropriate to this signal. Check if
573 * there is another siginfo for the same signal.
574 */
575 list_for_each_entry(q, &list->list, list) {
576 if (q->info.si_signo == sig) {
577 if (first)
578 goto still_pending;
579 first = q;
580 }
581 }
582
583 sigdelset(&list->signal, sig);
584
585 if (first) {
586 still_pending:
587 list_del_init(&first->list);
588 copy_siginfo(info, &first->info);
589
590 *resched_timer =
591 (first->flags & SIGQUEUE_PREALLOC) &&
592 (info->si_code == SI_TIMER) &&
593 (info->si_sys_private);
594
595 __sigqueue_free(first);
596 } else {
597 /*
598 * Ok, it wasn't in the queue. This must be
599 * a fast-pathed signal or we must have been
600 * out of queue space. So zero out the info.
601 */
602 clear_siginfo(info);
603 info->si_signo = sig;
604 info->si_errno = 0;
605 info->si_code = SI_USER;
606 info->si_pid = 0;
607 info->si_uid = 0;
608 }
609 }
610
__dequeue_signal(struct sigpending * pending,sigset_t * mask,kernel_siginfo_t * info,bool * resched_timer)611 static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
612 kernel_siginfo_t *info, bool *resched_timer)
613 {
614 int sig = next_signal(pending, mask);
615
616 if (sig)
617 collect_signal(sig, pending, info, resched_timer);
618 return sig;
619 }
620
621 /*
622 * Try to dequeue a signal. If a deliverable signal is found fill in the
623 * caller provided siginfo and return the signal number. Otherwise return
624 * 0.
625 */
dequeue_signal(sigset_t * mask,kernel_siginfo_t * info,enum pid_type * type)626 int dequeue_signal(sigset_t *mask, kernel_siginfo_t *info, enum pid_type *type)
627 {
628 struct task_struct *tsk = current;
629 bool resched_timer = false;
630 int signr;
631
632 lockdep_assert_held(&tsk->sighand->siglock);
633
634 *type = PIDTYPE_PID;
635 signr = __dequeue_signal(&tsk->pending, mask, info, &resched_timer);
636 if (!signr) {
637 *type = PIDTYPE_TGID;
638 signr = __dequeue_signal(&tsk->signal->shared_pending,
639 mask, info, &resched_timer);
640 #ifdef CONFIG_POSIX_TIMERS
641 /*
642 * itimer signal ?
643 *
644 * itimers are process shared and we restart periodic
645 * itimers in the signal delivery path to prevent DoS
646 * attacks in the high resolution timer case. This is
647 * compliant with the old way of self-restarting
648 * itimers, as the SIGALRM is a legacy signal and only
649 * queued once. Changing the restart behaviour to
650 * restart the timer in the signal dequeue path is
651 * reducing the timer noise on heavy loaded !highres
652 * systems too.
653 */
654 if (unlikely(signr == SIGALRM)) {
655 struct hrtimer *tmr = &tsk->signal->real_timer;
656
657 if (!hrtimer_is_queued(tmr) &&
658 tsk->signal->it_real_incr != 0) {
659 hrtimer_forward(tmr, tmr->base->get_time(),
660 tsk->signal->it_real_incr);
661 hrtimer_restart(tmr);
662 }
663 }
664 #endif
665 }
666
667 recalc_sigpending();
668 if (!signr)
669 return 0;
670
671 if (unlikely(sig_kernel_stop(signr))) {
672 /*
673 * Set a marker that we have dequeued a stop signal. Our
674 * caller might release the siglock and then the pending
675 * stop signal it is about to process is no longer in the
676 * pending bitmasks, but must still be cleared by a SIGCONT
677 * (and overruled by a SIGKILL). So those cases clear this
678 * shared flag after we've set it. Note that this flag may
679 * remain set after the signal we return is ignored or
680 * handled. That doesn't matter because its only purpose
681 * is to alert stop-signal processing code when another
682 * processor has come along and cleared the flag.
683 */
684 current->jobctl |= JOBCTL_STOP_DEQUEUED;
685 }
686 #ifdef CONFIG_POSIX_TIMERS
687 if (resched_timer) {
688 /*
689 * Release the siglock to ensure proper locking order
690 * of timer locks outside of siglocks. Note, we leave
691 * irqs disabled here, since the posix-timers code is
692 * about to disable them again anyway.
693 */
694 spin_unlock(&tsk->sighand->siglock);
695 posixtimer_rearm(info);
696 spin_lock(&tsk->sighand->siglock);
697
698 /* Don't expose the si_sys_private value to userspace */
699 info->si_sys_private = 0;
700 }
701 #endif
702 return signr;
703 }
704 EXPORT_SYMBOL_GPL(dequeue_signal);
705
dequeue_synchronous_signal(kernel_siginfo_t * info)706 static int dequeue_synchronous_signal(kernel_siginfo_t *info)
707 {
708 struct task_struct *tsk = current;
709 struct sigpending *pending = &tsk->pending;
710 struct sigqueue *q, *sync = NULL;
711
712 /*
713 * Might a synchronous signal be in the queue?
714 */
715 if (!((pending->signal.sig[0] & ~tsk->blocked.sig[0]) & SYNCHRONOUS_MASK))
716 return 0;
717
718 /*
719 * Return the first synchronous signal in the queue.
720 */
721 list_for_each_entry(q, &pending->list, list) {
722 /* Synchronous signals have a positive si_code */
723 if ((q->info.si_code > SI_USER) &&
724 (sigmask(q->info.si_signo) & SYNCHRONOUS_MASK)) {
725 sync = q;
726 goto next;
727 }
728 }
729 return 0;
730 next:
731 /*
732 * Check if there is another siginfo for the same signal.
733 */
734 list_for_each_entry_continue(q, &pending->list, list) {
735 if (q->info.si_signo == sync->info.si_signo)
736 goto still_pending;
737 }
738
739 sigdelset(&pending->signal, sync->info.si_signo);
740 recalc_sigpending();
741 still_pending:
742 list_del_init(&sync->list);
743 copy_siginfo(info, &sync->info);
744 __sigqueue_free(sync);
745 return info->si_signo;
746 }
747
748 /*
749 * Tell a process that it has a new active signal..
750 *
751 * NOTE! we rely on the previous spin_lock to
752 * lock interrupts for us! We can only be called with
753 * "siglock" held, and the local interrupt must
754 * have been disabled when that got acquired!
755 *
756 * No need to set need_resched since signal event passing
757 * goes through ->blocked
758 */
signal_wake_up_state(struct task_struct * t,unsigned int state)759 void signal_wake_up_state(struct task_struct *t, unsigned int state)
760 {
761 lockdep_assert_held(&t->sighand->siglock);
762
763 set_tsk_thread_flag(t, TIF_SIGPENDING);
764
765 /*
766 * TASK_WAKEKILL also means wake it up in the stopped/traced/killable
767 * case. We don't check t->state here because there is a race with it
768 * executing another processor and just now entering stopped state.
769 * By using wake_up_state, we ensure the process will wake up and
770 * handle its death signal.
771 */
772 if (!wake_up_state(t, state | TASK_INTERRUPTIBLE))
773 kick_process(t);
774 }
775
776 /*
777 * Remove signals in mask from the pending set and queue.
778 * Returns 1 if any signals were found.
779 *
780 * All callers must be holding the siglock.
781 */
flush_sigqueue_mask(sigset_t * mask,struct sigpending * s)782 static void flush_sigqueue_mask(sigset_t *mask, struct sigpending *s)
783 {
784 struct sigqueue *q, *n;
785 sigset_t m;
786
787 sigandsets(&m, mask, &s->signal);
788 if (sigisemptyset(&m))
789 return;
790
791 sigandnsets(&s->signal, &s->signal, mask);
792 list_for_each_entry_safe(q, n, &s->list, list) {
793 if (sigismember(mask, q->info.si_signo)) {
794 list_del_init(&q->list);
795 __sigqueue_free(q);
796 }
797 }
798 }
799
is_si_special(const struct kernel_siginfo * info)800 static inline int is_si_special(const struct kernel_siginfo *info)
801 {
802 return info <= SEND_SIG_PRIV;
803 }
804
si_fromuser(const struct kernel_siginfo * info)805 static inline bool si_fromuser(const struct kernel_siginfo *info)
806 {
807 return info == SEND_SIG_NOINFO ||
808 (!is_si_special(info) && SI_FROMUSER(info));
809 }
810
811 /*
812 * called with RCU read lock from check_kill_permission()
813 */
kill_ok_by_cred(struct task_struct * t)814 static bool kill_ok_by_cred(struct task_struct *t)
815 {
816 const struct cred *cred = current_cred();
817 const struct cred *tcred = __task_cred(t);
818
819 return uid_eq(cred->euid, tcred->suid) ||
820 uid_eq(cred->euid, tcred->uid) ||
821 uid_eq(cred->uid, tcred->suid) ||
822 uid_eq(cred->uid, tcred->uid) ||
823 ns_capable(tcred->user_ns, CAP_KILL);
824 }
825
826 /*
827 * Bad permissions for sending the signal
828 * - the caller must hold the RCU read lock
829 */
check_kill_permission(int sig,struct kernel_siginfo * info,struct task_struct * t)830 static int check_kill_permission(int sig, struct kernel_siginfo *info,
831 struct task_struct *t)
832 {
833 struct pid *sid;
834 int error;
835
836 if (!valid_signal(sig))
837 return -EINVAL;
838
839 if (!si_fromuser(info))
840 return 0;
841
842 error = audit_signal_info(sig, t); /* Let audit system see the signal */
843 if (error)
844 return error;
845
846 if (!same_thread_group(current, t) &&
847 !kill_ok_by_cred(t)) {
848 switch (sig) {
849 case SIGCONT:
850 sid = task_session(t);
851 /*
852 * We don't return the error if sid == NULL. The
853 * task was unhashed, the caller must notice this.
854 */
855 if (!sid || sid == task_session(current))
856 break;
857 fallthrough;
858 default:
859 return -EPERM;
860 }
861 }
862
863 return security_task_kill(t, info, sig, NULL);
864 }
865
866 /**
867 * ptrace_trap_notify - schedule trap to notify ptracer
868 * @t: tracee wanting to notify tracer
869 *
870 * This function schedules sticky ptrace trap which is cleared on the next
871 * TRAP_STOP to notify ptracer of an event. @t must have been seized by
872 * ptracer.
873 *
874 * If @t is running, STOP trap will be taken. If trapped for STOP and
875 * ptracer is listening for events, tracee is woken up so that it can
876 * re-trap for the new event. If trapped otherwise, STOP trap will be
877 * eventually taken without returning to userland after the existing traps
878 * are finished by PTRACE_CONT.
879 *
880 * CONTEXT:
881 * Must be called with @task->sighand->siglock held.
882 */
ptrace_trap_notify(struct task_struct * t)883 static void ptrace_trap_notify(struct task_struct *t)
884 {
885 WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
886 lockdep_assert_held(&t->sighand->siglock);
887
888 task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
889 ptrace_signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
890 }
891
892 /*
893 * Handle magic process-wide effects of stop/continue signals. Unlike
894 * the signal actions, these happen immediately at signal-generation
895 * time regardless of blocking, ignoring, or handling. This does the
896 * actual continuing for SIGCONT, but not the actual stopping for stop
897 * signals. The process stop is done as a signal action for SIG_DFL.
898 *
899 * Returns true if the signal should be actually delivered, otherwise
900 * it should be dropped.
901 */
prepare_signal(int sig,struct task_struct * p,bool force)902 static bool prepare_signal(int sig, struct task_struct *p, bool force)
903 {
904 struct signal_struct *signal = p->signal;
905 struct task_struct *t;
906 sigset_t flush;
907
908 if (signal->flags & SIGNAL_GROUP_EXIT) {
909 if (signal->core_state)
910 return sig == SIGKILL;
911 /*
912 * The process is in the middle of dying, drop the signal.
913 */
914 return false;
915 } else if (sig_kernel_stop(sig)) {
916 /*
917 * This is a stop signal. Remove SIGCONT from all queues.
918 */
919 siginitset(&flush, sigmask(SIGCONT));
920 flush_sigqueue_mask(&flush, &signal->shared_pending);
921 for_each_thread(p, t)
922 flush_sigqueue_mask(&flush, &t->pending);
923 } else if (sig == SIGCONT) {
924 unsigned int why;
925 /*
926 * Remove all stop signals from all queues, wake all threads.
927 */
928 siginitset(&flush, SIG_KERNEL_STOP_MASK);
929 flush_sigqueue_mask(&flush, &signal->shared_pending);
930 for_each_thread(p, t) {
931 flush_sigqueue_mask(&flush, &t->pending);
932 task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
933 if (likely(!(t->ptrace & PT_SEIZED))) {
934 t->jobctl &= ~JOBCTL_STOPPED;
935 wake_up_state(t, __TASK_STOPPED);
936 } else
937 ptrace_trap_notify(t);
938 }
939
940 /*
941 * Notify the parent with CLD_CONTINUED if we were stopped.
942 *
943 * If we were in the middle of a group stop, we pretend it
944 * was already finished, and then continued. Since SIGCHLD
945 * doesn't queue we report only CLD_STOPPED, as if the next
946 * CLD_CONTINUED was dropped.
947 */
948 why = 0;
949 if (signal->flags & SIGNAL_STOP_STOPPED)
950 why |= SIGNAL_CLD_CONTINUED;
951 else if (signal->group_stop_count)
952 why |= SIGNAL_CLD_STOPPED;
953
954 if (why) {
955 /*
956 * The first thread which returns from do_signal_stop()
957 * will take ->siglock, notice SIGNAL_CLD_MASK, and
958 * notify its parent. See get_signal().
959 */
960 signal_set_stop_flags(signal, why | SIGNAL_STOP_CONTINUED);
961 signal->group_stop_count = 0;
962 signal->group_exit_code = 0;
963 }
964 }
965
966 return !sig_ignored(p, sig, force);
967 }
968
969 /*
970 * Test if P wants to take SIG. After we've checked all threads with this,
971 * it's equivalent to finding no threads not blocking SIG. Any threads not
972 * blocking SIG were ruled out because they are not running and already
973 * have pending signals. Such threads will dequeue from the shared queue
974 * as soon as they're available, so putting the signal on the shared queue
975 * will be equivalent to sending it to one such thread.
976 */
wants_signal(int sig,struct task_struct * p)977 static inline bool wants_signal(int sig, struct task_struct *p)
978 {
979 if (sigismember(&p->blocked, sig))
980 return false;
981
982 if (p->flags & PF_EXITING)
983 return false;
984
985 if (sig == SIGKILL)
986 return true;
987
988 if (task_is_stopped_or_traced(p))
989 return false;
990
991 return task_curr(p) || !task_sigpending(p);
992 }
993
complete_signal(int sig,struct task_struct * p,enum pid_type type)994 static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
995 {
996 struct signal_struct *signal = p->signal;
997 struct task_struct *t;
998
999 /*
1000 * Now find a thread we can wake up to take the signal off the queue.
1001 *
1002 * Try the suggested task first (may or may not be the main thread).
1003 */
1004 if (wants_signal(sig, p))
1005 t = p;
1006 else if ((type == PIDTYPE_PID) || thread_group_empty(p))
1007 /*
1008 * There is just one thread and it does not need to be woken.
1009 * It will dequeue unblocked signals before it runs again.
1010 */
1011 return;
1012 else {
1013 /*
1014 * Otherwise try to find a suitable thread.
1015 */
1016 t = signal->curr_target;
1017 while (!wants_signal(sig, t)) {
1018 t = next_thread(t);
1019 if (t == signal->curr_target)
1020 /*
1021 * No thread needs to be woken.
1022 * Any eligible threads will see
1023 * the signal in the queue soon.
1024 */
1025 return;
1026 }
1027 signal->curr_target = t;
1028 }
1029
1030 /*
1031 * Found a killable thread. If the signal will be fatal,
1032 * then start taking the whole group down immediately.
1033 */
1034 if (sig_fatal(p, sig) &&
1035 (signal->core_state || !(signal->flags & SIGNAL_GROUP_EXIT)) &&
1036 !sigismember(&t->real_blocked, sig) &&
1037 (sig == SIGKILL || !p->ptrace)) {
1038 /*
1039 * This signal will be fatal to the whole group.
1040 */
1041 if (!sig_kernel_coredump(sig)) {
1042 /*
1043 * Start a group exit and wake everybody up.
1044 * This way we don't have other threads
1045 * running and doing things after a slower
1046 * thread has the fatal signal pending.
1047 */
1048 signal->flags = SIGNAL_GROUP_EXIT;
1049 signal->group_exit_code = sig;
1050 signal->group_stop_count = 0;
1051 __for_each_thread(signal, t) {
1052 task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1053 sigaddset(&t->pending.signal, SIGKILL);
1054 signal_wake_up(t, 1);
1055 }
1056 return;
1057 }
1058 }
1059
1060 /*
1061 * The signal is already in the shared-pending queue.
1062 * Tell the chosen thread to wake up and dequeue it.
1063 */
1064 signal_wake_up(t, sig == SIGKILL);
1065 return;
1066 }
1067
legacy_queue(struct sigpending * signals,int sig)1068 static inline bool legacy_queue(struct sigpending *signals, int sig)
1069 {
1070 return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
1071 }
1072
__send_signal_locked(int sig,struct kernel_siginfo * info,struct task_struct * t,enum pid_type type,bool force)1073 static int __send_signal_locked(int sig, struct kernel_siginfo *info,
1074 struct task_struct *t, enum pid_type type, bool force)
1075 {
1076 struct sigpending *pending;
1077 struct sigqueue *q;
1078 int override_rlimit;
1079 int ret = 0, result;
1080
1081 lockdep_assert_held(&t->sighand->siglock);
1082
1083 result = TRACE_SIGNAL_IGNORED;
1084 if (!prepare_signal(sig, t, force))
1085 goto ret;
1086
1087 pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1088 /*
1089 * Short-circuit ignored signals and support queuing
1090 * exactly one non-rt signal, so that we can get more
1091 * detailed information about the cause of the signal.
1092 */
1093 result = TRACE_SIGNAL_ALREADY_PENDING;
1094 if (legacy_queue(pending, sig))
1095 goto ret;
1096
1097 result = TRACE_SIGNAL_DELIVERED;
1098 /*
1099 * Skip useless siginfo allocation for SIGKILL and kernel threads.
1100 */
1101 if ((sig == SIGKILL) || (t->flags & PF_KTHREAD))
1102 goto out_set;
1103
1104 /*
1105 * Real-time signals must be queued if sent by sigqueue, or
1106 * some other real-time mechanism. It is implementation
1107 * defined whether kill() does so. We attempt to do so, on
1108 * the principle of least surprise, but since kill is not
1109 * allowed to fail with EAGAIN when low on memory we just
1110 * make sure at least one signal gets delivered and don't
1111 * pass on the info struct.
1112 */
1113 if (sig < SIGRTMIN)
1114 override_rlimit = (is_si_special(info) || info->si_code >= 0);
1115 else
1116 override_rlimit = 0;
1117
1118 q = __sigqueue_alloc(sig, t, GFP_ATOMIC, override_rlimit, 0);
1119
1120 if (q) {
1121 list_add_tail(&q->list, &pending->list);
1122 switch ((unsigned long) info) {
1123 case (unsigned long) SEND_SIG_NOINFO:
1124 clear_siginfo(&q->info);
1125 q->info.si_signo = sig;
1126 q->info.si_errno = 0;
1127 q->info.si_code = SI_USER;
1128 q->info.si_pid = task_tgid_nr_ns(current,
1129 task_active_pid_ns(t));
1130 rcu_read_lock();
1131 q->info.si_uid =
1132 from_kuid_munged(task_cred_xxx(t, user_ns),
1133 current_uid());
1134 rcu_read_unlock();
1135 break;
1136 case (unsigned long) SEND_SIG_PRIV:
1137 clear_siginfo(&q->info);
1138 q->info.si_signo = sig;
1139 q->info.si_errno = 0;
1140 q->info.si_code = SI_KERNEL;
1141 q->info.si_pid = 0;
1142 q->info.si_uid = 0;
1143 break;
1144 default:
1145 copy_siginfo(&q->info, info);
1146 break;
1147 }
1148 } else if (!is_si_special(info) &&
1149 sig >= SIGRTMIN && info->si_code != SI_USER) {
1150 /*
1151 * Queue overflow, abort. We may abort if the
1152 * signal was rt and sent by user using something
1153 * other than kill().
1154 */
1155 result = TRACE_SIGNAL_OVERFLOW_FAIL;
1156 ret = -EAGAIN;
1157 goto ret;
1158 } else {
1159 /*
1160 * This is a silent loss of information. We still
1161 * send the signal, but the *info bits are lost.
1162 */
1163 result = TRACE_SIGNAL_LOSE_INFO;
1164 }
1165
1166 out_set:
1167 signalfd_notify(t, sig);
1168 sigaddset(&pending->signal, sig);
1169
1170 /* Let multiprocess signals appear after on-going forks */
1171 if (type > PIDTYPE_TGID) {
1172 struct multiprocess_signals *delayed;
1173 hlist_for_each_entry(delayed, &t->signal->multiprocess, node) {
1174 sigset_t *signal = &delayed->signal;
1175 /* Can't queue both a stop and a continue signal */
1176 if (sig == SIGCONT)
1177 sigdelsetmask(signal, SIG_KERNEL_STOP_MASK);
1178 else if (sig_kernel_stop(sig))
1179 sigdelset(signal, SIGCONT);
1180 sigaddset(signal, sig);
1181 }
1182 }
1183
1184 complete_signal(sig, t, type);
1185 ret:
1186 trace_signal_generate(sig, info, t, type != PIDTYPE_PID, result);
1187 return ret;
1188 }
1189
has_si_pid_and_uid(struct kernel_siginfo * info)1190 static inline bool has_si_pid_and_uid(struct kernel_siginfo *info)
1191 {
1192 bool ret = false;
1193 switch (siginfo_layout(info->si_signo, info->si_code)) {
1194 case SIL_KILL:
1195 case SIL_CHLD:
1196 case SIL_RT:
1197 ret = true;
1198 break;
1199 case SIL_TIMER:
1200 case SIL_POLL:
1201 case SIL_FAULT:
1202 case SIL_FAULT_TRAPNO:
1203 case SIL_FAULT_MCEERR:
1204 case SIL_FAULT_BNDERR:
1205 case SIL_FAULT_PKUERR:
1206 case SIL_FAULT_PERF_EVENT:
1207 case SIL_SYS:
1208 ret = false;
1209 break;
1210 }
1211 return ret;
1212 }
1213
send_signal_locked(int sig,struct kernel_siginfo * info,struct task_struct * t,enum pid_type type)1214 int send_signal_locked(int sig, struct kernel_siginfo *info,
1215 struct task_struct *t, enum pid_type type)
1216 {
1217 /* Should SIGKILL or SIGSTOP be received by a pid namespace init? */
1218 bool force = false;
1219
1220 if (info == SEND_SIG_NOINFO) {
1221 /* Force if sent from an ancestor pid namespace */
1222 force = !task_pid_nr_ns(current, task_active_pid_ns(t));
1223 } else if (info == SEND_SIG_PRIV) {
1224 /* Don't ignore kernel generated signals */
1225 force = true;
1226 } else if (has_si_pid_and_uid(info)) {
1227 /* SIGKILL and SIGSTOP is special or has ids */
1228 struct user_namespace *t_user_ns;
1229
1230 rcu_read_lock();
1231 t_user_ns = task_cred_xxx(t, user_ns);
1232 if (current_user_ns() != t_user_ns) {
1233 kuid_t uid = make_kuid(current_user_ns(), info->si_uid);
1234 info->si_uid = from_kuid_munged(t_user_ns, uid);
1235 }
1236 rcu_read_unlock();
1237
1238 /* A kernel generated signal? */
1239 force = (info->si_code == SI_KERNEL);
1240
1241 /* From an ancestor pid namespace? */
1242 if (!task_pid_nr_ns(current, task_active_pid_ns(t))) {
1243 info->si_pid = 0;
1244 force = true;
1245 }
1246 }
1247 return __send_signal_locked(sig, info, t, type, force);
1248 }
1249
print_fatal_signal(int signr)1250 static void print_fatal_signal(int signr)
1251 {
1252 struct pt_regs *regs = task_pt_regs(current);
1253 struct file *exe_file;
1254
1255 exe_file = get_task_exe_file(current);
1256 if (exe_file) {
1257 pr_info("%pD: %s: potentially unexpected fatal signal %d.\n",
1258 exe_file, current->comm, signr);
1259 fput(exe_file);
1260 } else {
1261 pr_info("%s: potentially unexpected fatal signal %d.\n",
1262 current->comm, signr);
1263 }
1264
1265 #if defined(__i386__) && !defined(__arch_um__)
1266 pr_info("code at %08lx: ", regs->ip);
1267 {
1268 int i;
1269 for (i = 0; i < 16; i++) {
1270 unsigned char insn;
1271
1272 if (get_user(insn, (unsigned char *)(regs->ip + i)))
1273 break;
1274 pr_cont("%02x ", insn);
1275 }
1276 }
1277 pr_cont("\n");
1278 #endif
1279 preempt_disable();
1280 show_regs(regs);
1281 preempt_enable();
1282 }
1283
setup_print_fatal_signals(char * str)1284 static int __init setup_print_fatal_signals(char *str)
1285 {
1286 get_option (&str, &print_fatal_signals);
1287
1288 return 1;
1289 }
1290
1291 __setup("print-fatal-signals=", setup_print_fatal_signals);
1292
do_send_sig_info(int sig,struct kernel_siginfo * info,struct task_struct * p,enum pid_type type)1293 int do_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p,
1294 enum pid_type type)
1295 {
1296 unsigned long flags;
1297 int ret = -ESRCH;
1298
1299 if (lock_task_sighand(p, &flags)) {
1300 ret = send_signal_locked(sig, info, p, type);
1301 unlock_task_sighand(p, &flags);
1302 }
1303
1304 return ret;
1305 }
1306
1307 enum sig_handler {
1308 HANDLER_CURRENT, /* If reachable use the current handler */
1309 HANDLER_SIG_DFL, /* Always use SIG_DFL handler semantics */
1310 HANDLER_EXIT, /* Only visible as the process exit code */
1311 };
1312
1313 /*
1314 * Force a signal that the process can't ignore: if necessary
1315 * we unblock the signal and change any SIG_IGN to SIG_DFL.
1316 *
1317 * Note: If we unblock the signal, we always reset it to SIG_DFL,
1318 * since we do not want to have a signal handler that was blocked
1319 * be invoked when user space had explicitly blocked it.
1320 *
1321 * We don't want to have recursive SIGSEGV's etc, for example,
1322 * that is why we also clear SIGNAL_UNKILLABLE.
1323 */
1324 static int
force_sig_info_to_task(struct kernel_siginfo * info,struct task_struct * t,enum sig_handler handler)1325 force_sig_info_to_task(struct kernel_siginfo *info, struct task_struct *t,
1326 enum sig_handler handler)
1327 {
1328 unsigned long int flags;
1329 int ret, blocked, ignored;
1330 struct k_sigaction *action;
1331 int sig = info->si_signo;
1332
1333 spin_lock_irqsave(&t->sighand->siglock, flags);
1334 action = &t->sighand->action[sig-1];
1335 ignored = action->sa.sa_handler == SIG_IGN;
1336 blocked = sigismember(&t->blocked, sig);
1337 if (blocked || ignored || (handler != HANDLER_CURRENT)) {
1338 action->sa.sa_handler = SIG_DFL;
1339 if (handler == HANDLER_EXIT)
1340 action->sa.sa_flags |= SA_IMMUTABLE;
1341 if (blocked)
1342 sigdelset(&t->blocked, sig);
1343 }
1344 /*
1345 * Don't clear SIGNAL_UNKILLABLE for traced tasks, users won't expect
1346 * debugging to leave init killable. But HANDLER_EXIT is always fatal.
1347 */
1348 if (action->sa.sa_handler == SIG_DFL &&
1349 (!t->ptrace || (handler == HANDLER_EXIT)))
1350 t->signal->flags &= ~SIGNAL_UNKILLABLE;
1351 ret = send_signal_locked(sig, info, t, PIDTYPE_PID);
1352 /* This can happen if the signal was already pending and blocked */
1353 if (!task_sigpending(t))
1354 signal_wake_up(t, 0);
1355 spin_unlock_irqrestore(&t->sighand->siglock, flags);
1356
1357 return ret;
1358 }
1359
force_sig_info(struct kernel_siginfo * info)1360 int force_sig_info(struct kernel_siginfo *info)
1361 {
1362 return force_sig_info_to_task(info, current, HANDLER_CURRENT);
1363 }
1364
1365 /*
1366 * Nuke all other threads in the group.
1367 */
zap_other_threads(struct task_struct * p)1368 int zap_other_threads(struct task_struct *p)
1369 {
1370 struct task_struct *t;
1371 int count = 0;
1372
1373 p->signal->group_stop_count = 0;
1374
1375 for_other_threads(p, t) {
1376 task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1377 count++;
1378
1379 /* Don't bother with already dead threads */
1380 if (t->exit_state)
1381 continue;
1382 sigaddset(&t->pending.signal, SIGKILL);
1383 signal_wake_up(t, 1);
1384 }
1385
1386 return count;
1387 }
1388
__lock_task_sighand(struct task_struct * tsk,unsigned long * flags)1389 struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1390 unsigned long *flags)
1391 {
1392 struct sighand_struct *sighand;
1393
1394 rcu_read_lock();
1395 for (;;) {
1396 sighand = rcu_dereference(tsk->sighand);
1397 if (unlikely(sighand == NULL))
1398 break;
1399
1400 /*
1401 * This sighand can be already freed and even reused, but
1402 * we rely on SLAB_TYPESAFE_BY_RCU and sighand_ctor() which
1403 * initializes ->siglock: this slab can't go away, it has
1404 * the same object type, ->siglock can't be reinitialized.
1405 *
1406 * We need to ensure that tsk->sighand is still the same
1407 * after we take the lock, we can race with de_thread() or
1408 * __exit_signal(). In the latter case the next iteration
1409 * must see ->sighand == NULL.
1410 */
1411 spin_lock_irqsave(&sighand->siglock, *flags);
1412 if (likely(sighand == rcu_access_pointer(tsk->sighand)))
1413 break;
1414 spin_unlock_irqrestore(&sighand->siglock, *flags);
1415 }
1416 rcu_read_unlock();
1417
1418 return sighand;
1419 }
1420
1421 #ifdef CONFIG_LOCKDEP
lockdep_assert_task_sighand_held(struct task_struct * task)1422 void lockdep_assert_task_sighand_held(struct task_struct *task)
1423 {
1424 struct sighand_struct *sighand;
1425
1426 rcu_read_lock();
1427 sighand = rcu_dereference(task->sighand);
1428 if (sighand)
1429 lockdep_assert_held(&sighand->siglock);
1430 else
1431 WARN_ON_ONCE(1);
1432 rcu_read_unlock();
1433 }
1434 #endif
1435
1436 /*
1437 * send signal info to all the members of a thread group or to the
1438 * individual thread if type == PIDTYPE_PID.
1439 */
group_send_sig_info(int sig,struct kernel_siginfo * info,struct task_struct * p,enum pid_type type)1440 int group_send_sig_info(int sig, struct kernel_siginfo *info,
1441 struct task_struct *p, enum pid_type type)
1442 {
1443 int ret;
1444
1445 rcu_read_lock();
1446 ret = check_kill_permission(sig, info, p);
1447 rcu_read_unlock();
1448
1449 if (!ret && sig)
1450 ret = do_send_sig_info(sig, info, p, type);
1451
1452 return ret;
1453 }
1454
1455 /*
1456 * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1457 * control characters do (^C, ^Z etc)
1458 * - the caller must hold at least a readlock on tasklist_lock
1459 */
__kill_pgrp_info(int sig,struct kernel_siginfo * info,struct pid * pgrp)1460 int __kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1461 {
1462 struct task_struct *p = NULL;
1463 int ret = -ESRCH;
1464
1465 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1466 int err = group_send_sig_info(sig, info, p, PIDTYPE_PGID);
1467 /*
1468 * If group_send_sig_info() succeeds at least once ret
1469 * becomes 0 and after that the code below has no effect.
1470 * Otherwise we return the last err or -ESRCH if this
1471 * process group is empty.
1472 */
1473 if (ret)
1474 ret = err;
1475 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1476
1477 return ret;
1478 }
1479
kill_pid_info_type(int sig,struct kernel_siginfo * info,struct pid * pid,enum pid_type type)1480 static int kill_pid_info_type(int sig, struct kernel_siginfo *info,
1481 struct pid *pid, enum pid_type type)
1482 {
1483 int error = -ESRCH;
1484 struct task_struct *p;
1485
1486 for (;;) {
1487 rcu_read_lock();
1488 p = pid_task(pid, PIDTYPE_PID);
1489 if (p)
1490 error = group_send_sig_info(sig, info, p, type);
1491 rcu_read_unlock();
1492 if (likely(!p || error != -ESRCH))
1493 return error;
1494 /*
1495 * The task was unhashed in between, try again. If it
1496 * is dead, pid_task() will return NULL, if we race with
1497 * de_thread() it will find the new leader.
1498 */
1499 }
1500 }
1501
kill_pid_info(int sig,struct kernel_siginfo * info,struct pid * pid)1502 int kill_pid_info(int sig, struct kernel_siginfo *info, struct pid *pid)
1503 {
1504 return kill_pid_info_type(sig, info, pid, PIDTYPE_TGID);
1505 }
1506
kill_proc_info(int sig,struct kernel_siginfo * info,pid_t pid)1507 static int kill_proc_info(int sig, struct kernel_siginfo *info, pid_t pid)
1508 {
1509 int error;
1510 rcu_read_lock();
1511 error = kill_pid_info(sig, info, find_vpid(pid));
1512 rcu_read_unlock();
1513 return error;
1514 }
1515
kill_as_cred_perm(const struct cred * cred,struct task_struct * target)1516 static inline bool kill_as_cred_perm(const struct cred *cred,
1517 struct task_struct *target)
1518 {
1519 const struct cred *pcred = __task_cred(target);
1520
1521 return uid_eq(cred->euid, pcred->suid) ||
1522 uid_eq(cred->euid, pcred->uid) ||
1523 uid_eq(cred->uid, pcred->suid) ||
1524 uid_eq(cred->uid, pcred->uid);
1525 }
1526
1527 /*
1528 * The usb asyncio usage of siginfo is wrong. The glibc support
1529 * for asyncio which uses SI_ASYNCIO assumes the layout is SIL_RT.
1530 * AKA after the generic fields:
1531 * kernel_pid_t si_pid;
1532 * kernel_uid32_t si_uid;
1533 * sigval_t si_value;
1534 *
1535 * Unfortunately when usb generates SI_ASYNCIO it assumes the layout
1536 * after the generic fields is:
1537 * void __user *si_addr;
1538 *
1539 * This is a practical problem when there is a 64bit big endian kernel
1540 * and a 32bit userspace. As the 32bit address will encoded in the low
1541 * 32bits of the pointer. Those low 32bits will be stored at higher
1542 * address than appear in a 32 bit pointer. So userspace will not
1543 * see the address it was expecting for it's completions.
1544 *
1545 * There is nothing in the encoding that can allow
1546 * copy_siginfo_to_user32 to detect this confusion of formats, so
1547 * handle this by requiring the caller of kill_pid_usb_asyncio to
1548 * notice when this situration takes place and to store the 32bit
1549 * pointer in sival_int, instead of sival_addr of the sigval_t addr
1550 * parameter.
1551 */
kill_pid_usb_asyncio(int sig,int errno,sigval_t addr,struct pid * pid,const struct cred * cred)1552 int kill_pid_usb_asyncio(int sig, int errno, sigval_t addr,
1553 struct pid *pid, const struct cred *cred)
1554 {
1555 struct kernel_siginfo info;
1556 struct task_struct *p;
1557 unsigned long flags;
1558 int ret = -EINVAL;
1559
1560 if (!valid_signal(sig))
1561 return ret;
1562
1563 clear_siginfo(&info);
1564 info.si_signo = sig;
1565 info.si_errno = errno;
1566 info.si_code = SI_ASYNCIO;
1567 *((sigval_t *)&info.si_pid) = addr;
1568
1569 rcu_read_lock();
1570 p = pid_task(pid, PIDTYPE_PID);
1571 if (!p) {
1572 ret = -ESRCH;
1573 goto out_unlock;
1574 }
1575 if (!kill_as_cred_perm(cred, p)) {
1576 ret = -EPERM;
1577 goto out_unlock;
1578 }
1579 ret = security_task_kill(p, &info, sig, cred);
1580 if (ret)
1581 goto out_unlock;
1582
1583 if (sig) {
1584 if (lock_task_sighand(p, &flags)) {
1585 ret = __send_signal_locked(sig, &info, p, PIDTYPE_TGID, false);
1586 unlock_task_sighand(p, &flags);
1587 } else
1588 ret = -ESRCH;
1589 }
1590 out_unlock:
1591 rcu_read_unlock();
1592 return ret;
1593 }
1594 EXPORT_SYMBOL_GPL(kill_pid_usb_asyncio);
1595
1596 /*
1597 * kill_something_info() interprets pid in interesting ways just like kill(2).
1598 *
1599 * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1600 * is probably wrong. Should make it like BSD or SYSV.
1601 */
1602
kill_something_info(int sig,struct kernel_siginfo * info,pid_t pid)1603 static int kill_something_info(int sig, struct kernel_siginfo *info, pid_t pid)
1604 {
1605 int ret;
1606
1607 if (pid > 0)
1608 return kill_proc_info(sig, info, pid);
1609
1610 /* -INT_MIN is undefined. Exclude this case to avoid a UBSAN warning */
1611 if (pid == INT_MIN)
1612 return -ESRCH;
1613
1614 read_lock(&tasklist_lock);
1615 if (pid != -1) {
1616 ret = __kill_pgrp_info(sig, info,
1617 pid ? find_vpid(-pid) : task_pgrp(current));
1618 } else {
1619 int retval = 0, count = 0;
1620 struct task_struct * p;
1621
1622 for_each_process(p) {
1623 if (task_pid_vnr(p) > 1 &&
1624 !same_thread_group(p, current)) {
1625 int err = group_send_sig_info(sig, info, p,
1626 PIDTYPE_MAX);
1627 ++count;
1628 if (err != -EPERM)
1629 retval = err;
1630 }
1631 }
1632 ret = count ? retval : -ESRCH;
1633 }
1634 read_unlock(&tasklist_lock);
1635
1636 return ret;
1637 }
1638
1639 /*
1640 * These are for backward compatibility with the rest of the kernel source.
1641 */
1642
send_sig_info(int sig,struct kernel_siginfo * info,struct task_struct * p)1643 int send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1644 {
1645 /*
1646 * Make sure legacy kernel users don't send in bad values
1647 * (normal paths check this in check_kill_permission).
1648 */
1649 if (!valid_signal(sig))
1650 return -EINVAL;
1651
1652 return do_send_sig_info(sig, info, p, PIDTYPE_PID);
1653 }
1654 EXPORT_SYMBOL(send_sig_info);
1655
1656 #define __si_special(priv) \
1657 ((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1658
1659 int
send_sig(int sig,struct task_struct * p,int priv)1660 send_sig(int sig, struct task_struct *p, int priv)
1661 {
1662 return send_sig_info(sig, __si_special(priv), p);
1663 }
1664 EXPORT_SYMBOL(send_sig);
1665
force_sig(int sig)1666 void force_sig(int sig)
1667 {
1668 struct kernel_siginfo info;
1669
1670 clear_siginfo(&info);
1671 info.si_signo = sig;
1672 info.si_errno = 0;
1673 info.si_code = SI_KERNEL;
1674 info.si_pid = 0;
1675 info.si_uid = 0;
1676 force_sig_info(&info);
1677 }
1678 EXPORT_SYMBOL(force_sig);
1679
force_fatal_sig(int sig)1680 void force_fatal_sig(int sig)
1681 {
1682 struct kernel_siginfo info;
1683
1684 clear_siginfo(&info);
1685 info.si_signo = sig;
1686 info.si_errno = 0;
1687 info.si_code = SI_KERNEL;
1688 info.si_pid = 0;
1689 info.si_uid = 0;
1690 force_sig_info_to_task(&info, current, HANDLER_SIG_DFL);
1691 }
1692
force_exit_sig(int sig)1693 void force_exit_sig(int sig)
1694 {
1695 struct kernel_siginfo info;
1696
1697 clear_siginfo(&info);
1698 info.si_signo = sig;
1699 info.si_errno = 0;
1700 info.si_code = SI_KERNEL;
1701 info.si_pid = 0;
1702 info.si_uid = 0;
1703 force_sig_info_to_task(&info, current, HANDLER_EXIT);
1704 }
1705
1706 /*
1707 * When things go south during signal handling, we
1708 * will force a SIGSEGV. And if the signal that caused
1709 * the problem was already a SIGSEGV, we'll want to
1710 * make sure we don't even try to deliver the signal..
1711 */
force_sigsegv(int sig)1712 void force_sigsegv(int sig)
1713 {
1714 if (sig == SIGSEGV)
1715 force_fatal_sig(SIGSEGV);
1716 else
1717 force_sig(SIGSEGV);
1718 }
1719
force_sig_fault_to_task(int sig,int code,void __user * addr,struct task_struct * t)1720 int force_sig_fault_to_task(int sig, int code, void __user *addr,
1721 struct task_struct *t)
1722 {
1723 struct kernel_siginfo info;
1724
1725 clear_siginfo(&info);
1726 info.si_signo = sig;
1727 info.si_errno = 0;
1728 info.si_code = code;
1729 info.si_addr = addr;
1730 return force_sig_info_to_task(&info, t, HANDLER_CURRENT);
1731 }
1732
force_sig_fault(int sig,int code,void __user * addr)1733 int force_sig_fault(int sig, int code, void __user *addr)
1734 {
1735 return force_sig_fault_to_task(sig, code, addr, current);
1736 }
1737
send_sig_fault(int sig,int code,void __user * addr,struct task_struct * t)1738 int send_sig_fault(int sig, int code, void __user *addr, struct task_struct *t)
1739 {
1740 struct kernel_siginfo info;
1741
1742 clear_siginfo(&info);
1743 info.si_signo = sig;
1744 info.si_errno = 0;
1745 info.si_code = code;
1746 info.si_addr = addr;
1747 return send_sig_info(info.si_signo, &info, t);
1748 }
1749
force_sig_mceerr(int code,void __user * addr,short lsb)1750 int force_sig_mceerr(int code, void __user *addr, short lsb)
1751 {
1752 struct kernel_siginfo info;
1753
1754 WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1755 clear_siginfo(&info);
1756 info.si_signo = SIGBUS;
1757 info.si_errno = 0;
1758 info.si_code = code;
1759 info.si_addr = addr;
1760 info.si_addr_lsb = lsb;
1761 return force_sig_info(&info);
1762 }
1763
send_sig_mceerr(int code,void __user * addr,short lsb,struct task_struct * t)1764 int send_sig_mceerr(int code, void __user *addr, short lsb, struct task_struct *t)
1765 {
1766 struct kernel_siginfo info;
1767
1768 WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1769 clear_siginfo(&info);
1770 info.si_signo = SIGBUS;
1771 info.si_errno = 0;
1772 info.si_code = code;
1773 info.si_addr = addr;
1774 info.si_addr_lsb = lsb;
1775 return send_sig_info(info.si_signo, &info, t);
1776 }
1777 EXPORT_SYMBOL(send_sig_mceerr);
1778
force_sig_bnderr(void __user * addr,void __user * lower,void __user * upper)1779 int force_sig_bnderr(void __user *addr, void __user *lower, void __user *upper)
1780 {
1781 struct kernel_siginfo info;
1782
1783 clear_siginfo(&info);
1784 info.si_signo = SIGSEGV;
1785 info.si_errno = 0;
1786 info.si_code = SEGV_BNDERR;
1787 info.si_addr = addr;
1788 info.si_lower = lower;
1789 info.si_upper = upper;
1790 return force_sig_info(&info);
1791 }
1792
1793 #ifdef SEGV_PKUERR
force_sig_pkuerr(void __user * addr,u32 pkey)1794 int force_sig_pkuerr(void __user *addr, u32 pkey)
1795 {
1796 struct kernel_siginfo info;
1797
1798 clear_siginfo(&info);
1799 info.si_signo = SIGSEGV;
1800 info.si_errno = 0;
1801 info.si_code = SEGV_PKUERR;
1802 info.si_addr = addr;
1803 info.si_pkey = pkey;
1804 return force_sig_info(&info);
1805 }
1806 #endif
1807
send_sig_perf(void __user * addr,u32 type,u64 sig_data)1808 int send_sig_perf(void __user *addr, u32 type, u64 sig_data)
1809 {
1810 struct kernel_siginfo info;
1811
1812 clear_siginfo(&info);
1813 info.si_signo = SIGTRAP;
1814 info.si_errno = 0;
1815 info.si_code = TRAP_PERF;
1816 info.si_addr = addr;
1817 info.si_perf_data = sig_data;
1818 info.si_perf_type = type;
1819
1820 /*
1821 * Signals generated by perf events should not terminate the whole
1822 * process if SIGTRAP is blocked, however, delivering the signal
1823 * asynchronously is better than not delivering at all. But tell user
1824 * space if the signal was asynchronous, so it can clearly be
1825 * distinguished from normal synchronous ones.
1826 */
1827 info.si_perf_flags = sigismember(¤t->blocked, info.si_signo) ?
1828 TRAP_PERF_FLAG_ASYNC :
1829 0;
1830
1831 return send_sig_info(info.si_signo, &info, current);
1832 }
1833
1834 /**
1835 * force_sig_seccomp - signals the task to allow in-process syscall emulation
1836 * @syscall: syscall number to send to userland
1837 * @reason: filter-supplied reason code to send to userland (via si_errno)
1838 * @force_coredump: true to trigger a coredump
1839 *
1840 * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info.
1841 */
force_sig_seccomp(int syscall,int reason,bool force_coredump)1842 int force_sig_seccomp(int syscall, int reason, bool force_coredump)
1843 {
1844 struct kernel_siginfo info;
1845
1846 clear_siginfo(&info);
1847 info.si_signo = SIGSYS;
1848 info.si_code = SYS_SECCOMP;
1849 info.si_call_addr = (void __user *)KSTK_EIP(current);
1850 info.si_errno = reason;
1851 info.si_arch = syscall_get_arch(current);
1852 info.si_syscall = syscall;
1853 return force_sig_info_to_task(&info, current,
1854 force_coredump ? HANDLER_EXIT : HANDLER_CURRENT);
1855 }
1856
1857 /* For the crazy architectures that include trap information in
1858 * the errno field, instead of an actual errno value.
1859 */
force_sig_ptrace_errno_trap(int errno,void __user * addr)1860 int force_sig_ptrace_errno_trap(int errno, void __user *addr)
1861 {
1862 struct kernel_siginfo info;
1863
1864 clear_siginfo(&info);
1865 info.si_signo = SIGTRAP;
1866 info.si_errno = errno;
1867 info.si_code = TRAP_HWBKPT;
1868 info.si_addr = addr;
1869 return force_sig_info(&info);
1870 }
1871
1872 /* For the rare architectures that include trap information using
1873 * si_trapno.
1874 */
force_sig_fault_trapno(int sig,int code,void __user * addr,int trapno)1875 int force_sig_fault_trapno(int sig, int code, void __user *addr, int trapno)
1876 {
1877 struct kernel_siginfo info;
1878
1879 clear_siginfo(&info);
1880 info.si_signo = sig;
1881 info.si_errno = 0;
1882 info.si_code = code;
1883 info.si_addr = addr;
1884 info.si_trapno = trapno;
1885 return force_sig_info(&info);
1886 }
1887
1888 /* For the rare architectures that include trap information using
1889 * si_trapno.
1890 */
send_sig_fault_trapno(int sig,int code,void __user * addr,int trapno,struct task_struct * t)1891 int send_sig_fault_trapno(int sig, int code, void __user *addr, int trapno,
1892 struct task_struct *t)
1893 {
1894 struct kernel_siginfo info;
1895
1896 clear_siginfo(&info);
1897 info.si_signo = sig;
1898 info.si_errno = 0;
1899 info.si_code = code;
1900 info.si_addr = addr;
1901 info.si_trapno = trapno;
1902 return send_sig_info(info.si_signo, &info, t);
1903 }
1904
kill_pgrp_info(int sig,struct kernel_siginfo * info,struct pid * pgrp)1905 static int kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1906 {
1907 int ret;
1908 read_lock(&tasklist_lock);
1909 ret = __kill_pgrp_info(sig, info, pgrp);
1910 read_unlock(&tasklist_lock);
1911 return ret;
1912 }
1913
kill_pgrp(struct pid * pid,int sig,int priv)1914 int kill_pgrp(struct pid *pid, int sig, int priv)
1915 {
1916 return kill_pgrp_info(sig, __si_special(priv), pid);
1917 }
1918 EXPORT_SYMBOL(kill_pgrp);
1919
kill_pid(struct pid * pid,int sig,int priv)1920 int kill_pid(struct pid *pid, int sig, int priv)
1921 {
1922 return kill_pid_info(sig, __si_special(priv), pid);
1923 }
1924 EXPORT_SYMBOL(kill_pid);
1925
1926 /*
1927 * These functions support sending signals using preallocated sigqueue
1928 * structures. This is needed "because realtime applications cannot
1929 * afford to lose notifications of asynchronous events, like timer
1930 * expirations or I/O completions". In the case of POSIX Timers
1931 * we allocate the sigqueue structure from the timer_create. If this
1932 * allocation fails we are able to report the failure to the application
1933 * with an EAGAIN error.
1934 */
sigqueue_alloc(void)1935 struct sigqueue *sigqueue_alloc(void)
1936 {
1937 return __sigqueue_alloc(-1, current, GFP_KERNEL, 0, SIGQUEUE_PREALLOC);
1938 }
1939
sigqueue_free(struct sigqueue * q)1940 void sigqueue_free(struct sigqueue *q)
1941 {
1942 spinlock_t *lock = ¤t->sighand->siglock;
1943 unsigned long flags;
1944
1945 if (WARN_ON_ONCE(!(q->flags & SIGQUEUE_PREALLOC)))
1946 return;
1947 /*
1948 * We must hold ->siglock while testing q->list
1949 * to serialize with collect_signal() or with
1950 * __exit_signal()->flush_sigqueue().
1951 */
1952 spin_lock_irqsave(lock, flags);
1953 q->flags &= ~SIGQUEUE_PREALLOC;
1954 /*
1955 * If it is queued it will be freed when dequeued,
1956 * like the "regular" sigqueue.
1957 */
1958 if (!list_empty(&q->list))
1959 q = NULL;
1960 spin_unlock_irqrestore(lock, flags);
1961
1962 if (q)
1963 __sigqueue_free(q);
1964 }
1965
send_sigqueue(struct sigqueue * q,struct pid * pid,enum pid_type type)1966 int send_sigqueue(struct sigqueue *q, struct pid *pid, enum pid_type type)
1967 {
1968 int sig = q->info.si_signo;
1969 struct sigpending *pending;
1970 struct task_struct *t;
1971 unsigned long flags;
1972 int ret, result;
1973
1974 if (WARN_ON_ONCE(!(q->flags & SIGQUEUE_PREALLOC)))
1975 return 0;
1976 if (WARN_ON_ONCE(q->info.si_code != SI_TIMER))
1977 return 0;
1978
1979 ret = -1;
1980 rcu_read_lock();
1981
1982 /*
1983 * This function is used by POSIX timers to deliver a timer signal.
1984 * Where type is PIDTYPE_PID (such as for timers with SIGEV_THREAD_ID
1985 * set), the signal must be delivered to the specific thread (queues
1986 * into t->pending).
1987 *
1988 * Where type is not PIDTYPE_PID, signals must be delivered to the
1989 * process. In this case, prefer to deliver to current if it is in
1990 * the same thread group as the target process, which avoids
1991 * unnecessarily waking up a potentially idle task.
1992 */
1993 t = pid_task(pid, type);
1994 if (!t)
1995 goto ret;
1996 if (type != PIDTYPE_PID && same_thread_group(t, current))
1997 t = current;
1998 if (!likely(lock_task_sighand(t, &flags)))
1999 goto ret;
2000
2001 ret = 1; /* the signal is ignored */
2002 result = TRACE_SIGNAL_IGNORED;
2003 if (!prepare_signal(sig, t, false))
2004 goto out;
2005
2006 ret = 0;
2007 if (unlikely(!list_empty(&q->list))) {
2008 /*
2009 * If an SI_TIMER entry is already queue just increment
2010 * the overrun count.
2011 */
2012 q->info.si_overrun++;
2013 result = TRACE_SIGNAL_ALREADY_PENDING;
2014 goto out;
2015 }
2016 q->info.si_overrun = 0;
2017
2018 signalfd_notify(t, sig);
2019 pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
2020 list_add_tail(&q->list, &pending->list);
2021 sigaddset(&pending->signal, sig);
2022 complete_signal(sig, t, type);
2023 result = TRACE_SIGNAL_DELIVERED;
2024 out:
2025 trace_signal_generate(sig, &q->info, t, type != PIDTYPE_PID, result);
2026 unlock_task_sighand(t, &flags);
2027 ret:
2028 rcu_read_unlock();
2029 return ret;
2030 }
2031
do_notify_pidfd(struct task_struct * task)2032 void do_notify_pidfd(struct task_struct *task)
2033 {
2034 struct pid *pid = task_pid(task);
2035
2036 WARN_ON(task->exit_state == 0);
2037
2038 __wake_up(&pid->wait_pidfd, TASK_NORMAL, 0,
2039 poll_to_key(EPOLLIN | EPOLLRDNORM));
2040 }
2041
2042 /*
2043 * Let a parent know about the death of a child.
2044 * For a stopped/continued status change, use do_notify_parent_cldstop instead.
2045 *
2046 * Returns true if our parent ignored us and so we've switched to
2047 * self-reaping.
2048 */
do_notify_parent(struct task_struct * tsk,int sig)2049 bool do_notify_parent(struct task_struct *tsk, int sig)
2050 {
2051 struct kernel_siginfo info;
2052 unsigned long flags;
2053 struct sighand_struct *psig;
2054 bool autoreap = false;
2055 u64 utime, stime;
2056
2057 WARN_ON_ONCE(sig == -1);
2058
2059 /* do_notify_parent_cldstop should have been called instead. */
2060 WARN_ON_ONCE(task_is_stopped_or_traced(tsk));
2061
2062 WARN_ON_ONCE(!tsk->ptrace &&
2063 (tsk->group_leader != tsk || !thread_group_empty(tsk)));
2064 /*
2065 * tsk is a group leader and has no threads, wake up the
2066 * non-PIDFD_THREAD waiters.
2067 */
2068 if (thread_group_empty(tsk))
2069 do_notify_pidfd(tsk);
2070
2071 if (sig != SIGCHLD) {
2072 /*
2073 * This is only possible if parent == real_parent.
2074 * Check if it has changed security domain.
2075 */
2076 if (tsk->parent_exec_id != READ_ONCE(tsk->parent->self_exec_id))
2077 sig = SIGCHLD;
2078 }
2079
2080 clear_siginfo(&info);
2081 info.si_signo = sig;
2082 info.si_errno = 0;
2083 /*
2084 * We are under tasklist_lock here so our parent is tied to
2085 * us and cannot change.
2086 *
2087 * task_active_pid_ns will always return the same pid namespace
2088 * until a task passes through release_task.
2089 *
2090 * write_lock() currently calls preempt_disable() which is the
2091 * same as rcu_read_lock(), but according to Oleg, this is not
2092 * correct to rely on this
2093 */
2094 rcu_read_lock();
2095 info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
2096 info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
2097 task_uid(tsk));
2098 rcu_read_unlock();
2099
2100 task_cputime(tsk, &utime, &stime);
2101 info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
2102 info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
2103
2104 info.si_status = tsk->exit_code & 0x7f;
2105 if (tsk->exit_code & 0x80)
2106 info.si_code = CLD_DUMPED;
2107 else if (tsk->exit_code & 0x7f)
2108 info.si_code = CLD_KILLED;
2109 else {
2110 info.si_code = CLD_EXITED;
2111 info.si_status = tsk->exit_code >> 8;
2112 }
2113
2114 psig = tsk->parent->sighand;
2115 spin_lock_irqsave(&psig->siglock, flags);
2116 if (!tsk->ptrace && sig == SIGCHLD &&
2117 (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
2118 (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
2119 /*
2120 * We are exiting and our parent doesn't care. POSIX.1
2121 * defines special semantics for setting SIGCHLD to SIG_IGN
2122 * or setting the SA_NOCLDWAIT flag: we should be reaped
2123 * automatically and not left for our parent's wait4 call.
2124 * Rather than having the parent do it as a magic kind of
2125 * signal handler, we just set this to tell do_exit that we
2126 * can be cleaned up without becoming a zombie. Note that
2127 * we still call __wake_up_parent in this case, because a
2128 * blocked sys_wait4 might now return -ECHILD.
2129 *
2130 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
2131 * is implementation-defined: we do (if you don't want
2132 * it, just use SIG_IGN instead).
2133 */
2134 autoreap = true;
2135 if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
2136 sig = 0;
2137 }
2138 /*
2139 * Send with __send_signal as si_pid and si_uid are in the
2140 * parent's namespaces.
2141 */
2142 if (valid_signal(sig) && sig)
2143 __send_signal_locked(sig, &info, tsk->parent, PIDTYPE_TGID, false);
2144 __wake_up_parent(tsk, tsk->parent);
2145 spin_unlock_irqrestore(&psig->siglock, flags);
2146
2147 return autoreap;
2148 }
2149
2150 /**
2151 * do_notify_parent_cldstop - notify parent of stopped/continued state change
2152 * @tsk: task reporting the state change
2153 * @for_ptracer: the notification is for ptracer
2154 * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
2155 *
2156 * Notify @tsk's parent that the stopped/continued state has changed. If
2157 * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
2158 * If %true, @tsk reports to @tsk->parent which should be the ptracer.
2159 *
2160 * CONTEXT:
2161 * Must be called with tasklist_lock at least read locked.
2162 */
do_notify_parent_cldstop(struct task_struct * tsk,bool for_ptracer,int why)2163 static void do_notify_parent_cldstop(struct task_struct *tsk,
2164 bool for_ptracer, int why)
2165 {
2166 struct kernel_siginfo info;
2167 unsigned long flags;
2168 struct task_struct *parent;
2169 struct sighand_struct *sighand;
2170 u64 utime, stime;
2171
2172 if (for_ptracer) {
2173 parent = tsk->parent;
2174 } else {
2175 tsk = tsk->group_leader;
2176 parent = tsk->real_parent;
2177 }
2178
2179 clear_siginfo(&info);
2180 info.si_signo = SIGCHLD;
2181 info.si_errno = 0;
2182 /*
2183 * see comment in do_notify_parent() about the following 4 lines
2184 */
2185 rcu_read_lock();
2186 info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
2187 info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
2188 rcu_read_unlock();
2189
2190 task_cputime(tsk, &utime, &stime);
2191 info.si_utime = nsec_to_clock_t(utime);
2192 info.si_stime = nsec_to_clock_t(stime);
2193
2194 info.si_code = why;
2195 switch (why) {
2196 case CLD_CONTINUED:
2197 info.si_status = SIGCONT;
2198 break;
2199 case CLD_STOPPED:
2200 info.si_status = tsk->signal->group_exit_code & 0x7f;
2201 break;
2202 case CLD_TRAPPED:
2203 info.si_status = tsk->exit_code & 0x7f;
2204 break;
2205 default:
2206 BUG();
2207 }
2208
2209 sighand = parent->sighand;
2210 spin_lock_irqsave(&sighand->siglock, flags);
2211 if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
2212 !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
2213 send_signal_locked(SIGCHLD, &info, parent, PIDTYPE_TGID);
2214 /*
2215 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
2216 */
2217 __wake_up_parent(tsk, parent);
2218 spin_unlock_irqrestore(&sighand->siglock, flags);
2219 }
2220
2221 /*
2222 * This must be called with current->sighand->siglock held.
2223 *
2224 * This should be the path for all ptrace stops.
2225 * We always set current->last_siginfo while stopped here.
2226 * That makes it a way to test a stopped process for
2227 * being ptrace-stopped vs being job-control-stopped.
2228 *
2229 * Returns the signal the ptracer requested the code resume
2230 * with. If the code did not stop because the tracer is gone,
2231 * the stop signal remains unchanged unless clear_code.
2232 */
ptrace_stop(int exit_code,int why,unsigned long message,kernel_siginfo_t * info)2233 static int ptrace_stop(int exit_code, int why, unsigned long message,
2234 kernel_siginfo_t *info)
2235 __releases(¤t->sighand->siglock)
2236 __acquires(¤t->sighand->siglock)
2237 {
2238 bool gstop_done = false;
2239
2240 if (arch_ptrace_stop_needed()) {
2241 /*
2242 * The arch code has something special to do before a
2243 * ptrace stop. This is allowed to block, e.g. for faults
2244 * on user stack pages. We can't keep the siglock while
2245 * calling arch_ptrace_stop, so we must release it now.
2246 * To preserve proper semantics, we must do this before
2247 * any signal bookkeeping like checking group_stop_count.
2248 */
2249 spin_unlock_irq(¤t->sighand->siglock);
2250 arch_ptrace_stop();
2251 spin_lock_irq(¤t->sighand->siglock);
2252 }
2253
2254 /*
2255 * After this point ptrace_signal_wake_up or signal_wake_up
2256 * will clear TASK_TRACED if ptrace_unlink happens or a fatal
2257 * signal comes in. Handle previous ptrace_unlinks and fatal
2258 * signals here to prevent ptrace_stop sleeping in schedule.
2259 */
2260 if (!current->ptrace || __fatal_signal_pending(current))
2261 return exit_code;
2262
2263 set_special_state(TASK_TRACED);
2264 current->jobctl |= JOBCTL_TRACED;
2265
2266 /*
2267 * We're committing to trapping. TRACED should be visible before
2268 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
2269 * Also, transition to TRACED and updates to ->jobctl should be
2270 * atomic with respect to siglock and should be done after the arch
2271 * hook as siglock is released and regrabbed across it.
2272 *
2273 * TRACER TRACEE
2274 *
2275 * ptrace_attach()
2276 * [L] wait_on_bit(JOBCTL_TRAPPING) [S] set_special_state(TRACED)
2277 * do_wait()
2278 * set_current_state() smp_wmb();
2279 * ptrace_do_wait()
2280 * wait_task_stopped()
2281 * task_stopped_code()
2282 * [L] task_is_traced() [S] task_clear_jobctl_trapping();
2283 */
2284 smp_wmb();
2285
2286 current->ptrace_message = message;
2287 current->last_siginfo = info;
2288 current->exit_code = exit_code;
2289
2290 /*
2291 * If @why is CLD_STOPPED, we're trapping to participate in a group
2292 * stop. Do the bookkeeping. Note that if SIGCONT was delievered
2293 * across siglock relocks since INTERRUPT was scheduled, PENDING
2294 * could be clear now. We act as if SIGCONT is received after
2295 * TASK_TRACED is entered - ignore it.
2296 */
2297 if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
2298 gstop_done = task_participate_group_stop(current);
2299
2300 /* any trap clears pending STOP trap, STOP trap clears NOTIFY */
2301 task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
2302 if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
2303 task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
2304
2305 /* entering a trap, clear TRAPPING */
2306 task_clear_jobctl_trapping(current);
2307
2308 spin_unlock_irq(¤t->sighand->siglock);
2309 read_lock(&tasklist_lock);
2310 /*
2311 * Notify parents of the stop.
2312 *
2313 * While ptraced, there are two parents - the ptracer and
2314 * the real_parent of the group_leader. The ptracer should
2315 * know about every stop while the real parent is only
2316 * interested in the completion of group stop. The states
2317 * for the two don't interact with each other. Notify
2318 * separately unless they're gonna be duplicates.
2319 */
2320 if (current->ptrace)
2321 do_notify_parent_cldstop(current, true, why);
2322 if (gstop_done && (!current->ptrace || ptrace_reparented(current)))
2323 do_notify_parent_cldstop(current, false, why);
2324
2325 /*
2326 * The previous do_notify_parent_cldstop() invocation woke ptracer.
2327 * One a PREEMPTION kernel this can result in preemption requirement
2328 * which will be fulfilled after read_unlock() and the ptracer will be
2329 * put on the CPU.
2330 * The ptracer is in wait_task_inactive(, __TASK_TRACED) waiting for
2331 * this task wait in schedule(). If this task gets preempted then it
2332 * remains enqueued on the runqueue. The ptracer will observe this and
2333 * then sleep for a delay of one HZ tick. In the meantime this task
2334 * gets scheduled, enters schedule() and will wait for the ptracer.
2335 *
2336 * This preemption point is not bad from a correctness point of
2337 * view but extends the runtime by one HZ tick time due to the
2338 * ptracer's sleep. The preempt-disable section ensures that there
2339 * will be no preemption between unlock and schedule() and so
2340 * improving the performance since the ptracer will observe that
2341 * the tracee is scheduled out once it gets on the CPU.
2342 *
2343 * On PREEMPT_RT locking tasklist_lock does not disable preemption.
2344 * Therefore the task can be preempted after do_notify_parent_cldstop()
2345 * before unlocking tasklist_lock so there is no benefit in doing this.
2346 *
2347 * In fact disabling preemption is harmful on PREEMPT_RT because
2348 * the spinlock_t in cgroup_enter_frozen() must not be acquired
2349 * with preemption disabled due to the 'sleeping' spinlock
2350 * substitution of RT.
2351 */
2352 if (!IS_ENABLED(CONFIG_PREEMPT_RT))
2353 preempt_disable();
2354 read_unlock(&tasklist_lock);
2355 cgroup_enter_frozen();
2356 if (!IS_ENABLED(CONFIG_PREEMPT_RT))
2357 preempt_enable_no_resched();
2358 schedule();
2359 cgroup_leave_frozen(true);
2360
2361 /*
2362 * We are back. Now reacquire the siglock before touching
2363 * last_siginfo, so that we are sure to have synchronized with
2364 * any signal-sending on another CPU that wants to examine it.
2365 */
2366 spin_lock_irq(¤t->sighand->siglock);
2367 exit_code = current->exit_code;
2368 current->last_siginfo = NULL;
2369 current->ptrace_message = 0;
2370 current->exit_code = 0;
2371
2372 /* LISTENING can be set only during STOP traps, clear it */
2373 current->jobctl &= ~(JOBCTL_LISTENING | JOBCTL_PTRACE_FROZEN);
2374
2375 /*
2376 * Queued signals ignored us while we were stopped for tracing.
2377 * So check for any that we should take before resuming user mode.
2378 * This sets TIF_SIGPENDING, but never clears it.
2379 */
2380 recalc_sigpending_tsk(current);
2381 return exit_code;
2382 }
2383
ptrace_do_notify(int signr,int exit_code,int why,unsigned long message)2384 static int ptrace_do_notify(int signr, int exit_code, int why, unsigned long message)
2385 {
2386 kernel_siginfo_t info;
2387
2388 clear_siginfo(&info);
2389 info.si_signo = signr;
2390 info.si_code = exit_code;
2391 info.si_pid = task_pid_vnr(current);
2392 info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2393
2394 /* Let the debugger run. */
2395 return ptrace_stop(exit_code, why, message, &info);
2396 }
2397
ptrace_notify(int exit_code,unsigned long message)2398 int ptrace_notify(int exit_code, unsigned long message)
2399 {
2400 int signr;
2401
2402 BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
2403 if (unlikely(task_work_pending(current)))
2404 task_work_run();
2405
2406 spin_lock_irq(¤t->sighand->siglock);
2407 signr = ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED, message);
2408 spin_unlock_irq(¤t->sighand->siglock);
2409 return signr;
2410 }
2411
2412 /**
2413 * do_signal_stop - handle group stop for SIGSTOP and other stop signals
2414 * @signr: signr causing group stop if initiating
2415 *
2416 * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
2417 * and participate in it. If already set, participate in the existing
2418 * group stop. If participated in a group stop (and thus slept), %true is
2419 * returned with siglock released.
2420 *
2421 * If ptraced, this function doesn't handle stop itself. Instead,
2422 * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
2423 * untouched. The caller must ensure that INTERRUPT trap handling takes
2424 * places afterwards.
2425 *
2426 * CONTEXT:
2427 * Must be called with @current->sighand->siglock held, which is released
2428 * on %true return.
2429 *
2430 * RETURNS:
2431 * %false if group stop is already cancelled or ptrace trap is scheduled.
2432 * %true if participated in group stop.
2433 */
do_signal_stop(int signr)2434 static bool do_signal_stop(int signr)
2435 __releases(¤t->sighand->siglock)
2436 {
2437 struct signal_struct *sig = current->signal;
2438
2439 if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2440 unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2441 struct task_struct *t;
2442
2443 /* signr will be recorded in task->jobctl for retries */
2444 WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2445
2446 if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2447 unlikely(sig->flags & SIGNAL_GROUP_EXIT) ||
2448 unlikely(sig->group_exec_task))
2449 return false;
2450 /*
2451 * There is no group stop already in progress. We must
2452 * initiate one now.
2453 *
2454 * While ptraced, a task may be resumed while group stop is
2455 * still in effect and then receive a stop signal and
2456 * initiate another group stop. This deviates from the
2457 * usual behavior as two consecutive stop signals can't
2458 * cause two group stops when !ptraced. That is why we
2459 * also check !task_is_stopped(t) below.
2460 *
2461 * The condition can be distinguished by testing whether
2462 * SIGNAL_STOP_STOPPED is already set. Don't generate
2463 * group_exit_code in such case.
2464 *
2465 * This is not necessary for SIGNAL_STOP_CONTINUED because
2466 * an intervening stop signal is required to cause two
2467 * continued events regardless of ptrace.
2468 */
2469 if (!(sig->flags & SIGNAL_STOP_STOPPED))
2470 sig->group_exit_code = signr;
2471
2472 sig->group_stop_count = 0;
2473 if (task_set_jobctl_pending(current, signr | gstop))
2474 sig->group_stop_count++;
2475
2476 for_other_threads(current, t) {
2477 /*
2478 * Setting state to TASK_STOPPED for a group
2479 * stop is always done with the siglock held,
2480 * so this check has no races.
2481 */
2482 if (!task_is_stopped(t) &&
2483 task_set_jobctl_pending(t, signr | gstop)) {
2484 sig->group_stop_count++;
2485 if (likely(!(t->ptrace & PT_SEIZED)))
2486 signal_wake_up(t, 0);
2487 else
2488 ptrace_trap_notify(t);
2489 }
2490 }
2491 }
2492
2493 if (likely(!current->ptrace)) {
2494 int notify = 0;
2495
2496 /*
2497 * If there are no other threads in the group, or if there
2498 * is a group stop in progress and we are the last to stop,
2499 * report to the parent.
2500 */
2501 if (task_participate_group_stop(current))
2502 notify = CLD_STOPPED;
2503
2504 current->jobctl |= JOBCTL_STOPPED;
2505 set_special_state(TASK_STOPPED);
2506 spin_unlock_irq(¤t->sighand->siglock);
2507
2508 /*
2509 * Notify the parent of the group stop completion. Because
2510 * we're not holding either the siglock or tasklist_lock
2511 * here, ptracer may attach inbetween; however, this is for
2512 * group stop and should always be delivered to the real
2513 * parent of the group leader. The new ptracer will get
2514 * its notification when this task transitions into
2515 * TASK_TRACED.
2516 */
2517 if (notify) {
2518 read_lock(&tasklist_lock);
2519 do_notify_parent_cldstop(current, false, notify);
2520 read_unlock(&tasklist_lock);
2521 }
2522
2523 /* Now we don't run again until woken by SIGCONT or SIGKILL */
2524 cgroup_enter_frozen();
2525 schedule();
2526 return true;
2527 } else {
2528 /*
2529 * While ptraced, group stop is handled by STOP trap.
2530 * Schedule it and let the caller deal with it.
2531 */
2532 task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2533 return false;
2534 }
2535 }
2536
2537 /**
2538 * do_jobctl_trap - take care of ptrace jobctl traps
2539 *
2540 * When PT_SEIZED, it's used for both group stop and explicit
2541 * SEIZE/INTERRUPT traps. Both generate PTRACE_EVENT_STOP trap with
2542 * accompanying siginfo. If stopped, lower eight bits of exit_code contain
2543 * the stop signal; otherwise, %SIGTRAP.
2544 *
2545 * When !PT_SEIZED, it's used only for group stop trap with stop signal
2546 * number as exit_code and no siginfo.
2547 *
2548 * CONTEXT:
2549 * Must be called with @current->sighand->siglock held, which may be
2550 * released and re-acquired before returning with intervening sleep.
2551 */
do_jobctl_trap(void)2552 static void do_jobctl_trap(void)
2553 {
2554 struct signal_struct *signal = current->signal;
2555 int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2556
2557 if (current->ptrace & PT_SEIZED) {
2558 if (!signal->group_stop_count &&
2559 !(signal->flags & SIGNAL_STOP_STOPPED))
2560 signr = SIGTRAP;
2561 WARN_ON_ONCE(!signr);
2562 ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2563 CLD_STOPPED, 0);
2564 } else {
2565 WARN_ON_ONCE(!signr);
2566 ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2567 }
2568 }
2569
2570 /**
2571 * do_freezer_trap - handle the freezer jobctl trap
2572 *
2573 * Puts the task into frozen state, if only the task is not about to quit.
2574 * In this case it drops JOBCTL_TRAP_FREEZE.
2575 *
2576 * CONTEXT:
2577 * Must be called with @current->sighand->siglock held,
2578 * which is always released before returning.
2579 */
do_freezer_trap(void)2580 static void do_freezer_trap(void)
2581 __releases(¤t->sighand->siglock)
2582 {
2583 /*
2584 * If there are other trap bits pending except JOBCTL_TRAP_FREEZE,
2585 * let's make another loop to give it a chance to be handled.
2586 * In any case, we'll return back.
2587 */
2588 if ((current->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) !=
2589 JOBCTL_TRAP_FREEZE) {
2590 spin_unlock_irq(¤t->sighand->siglock);
2591 return;
2592 }
2593
2594 /*
2595 * Now we're sure that there is no pending fatal signal and no
2596 * pending traps. Clear TIF_SIGPENDING to not get out of schedule()
2597 * immediately (if there is a non-fatal signal pending), and
2598 * put the task into sleep.
2599 */
2600 __set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
2601 clear_thread_flag(TIF_SIGPENDING);
2602 spin_unlock_irq(¤t->sighand->siglock);
2603 cgroup_enter_frozen();
2604 schedule();
2605
2606 /*
2607 * We could've been woken by task_work, run it to clear
2608 * TIF_NOTIFY_SIGNAL. The caller will retry if necessary.
2609 */
2610 clear_notify_signal();
2611 if (unlikely(task_work_pending(current)))
2612 task_work_run();
2613 }
2614
ptrace_signal(int signr,kernel_siginfo_t * info,enum pid_type type)2615 static int ptrace_signal(int signr, kernel_siginfo_t *info, enum pid_type type)
2616 {
2617 /*
2618 * We do not check sig_kernel_stop(signr) but set this marker
2619 * unconditionally because we do not know whether debugger will
2620 * change signr. This flag has no meaning unless we are going
2621 * to stop after return from ptrace_stop(). In this case it will
2622 * be checked in do_signal_stop(), we should only stop if it was
2623 * not cleared by SIGCONT while we were sleeping. See also the
2624 * comment in dequeue_signal().
2625 */
2626 current->jobctl |= JOBCTL_STOP_DEQUEUED;
2627 signr = ptrace_stop(signr, CLD_TRAPPED, 0, info);
2628
2629 /* We're back. Did the debugger cancel the sig? */
2630 if (signr == 0)
2631 return signr;
2632
2633 /*
2634 * Update the siginfo structure if the signal has
2635 * changed. If the debugger wanted something
2636 * specific in the siginfo structure then it should
2637 * have updated *info via PTRACE_SETSIGINFO.
2638 */
2639 if (signr != info->si_signo) {
2640 clear_siginfo(info);
2641 info->si_signo = signr;
2642 info->si_errno = 0;
2643 info->si_code = SI_USER;
2644 rcu_read_lock();
2645 info->si_pid = task_pid_vnr(current->parent);
2646 info->si_uid = from_kuid_munged(current_user_ns(),
2647 task_uid(current->parent));
2648 rcu_read_unlock();
2649 }
2650
2651 /* If the (new) signal is now blocked, requeue it. */
2652 if (sigismember(¤t->blocked, signr) ||
2653 fatal_signal_pending(current)) {
2654 send_signal_locked(signr, info, current, type);
2655 signr = 0;
2656 }
2657
2658 return signr;
2659 }
2660
hide_si_addr_tag_bits(struct ksignal * ksig)2661 static void hide_si_addr_tag_bits(struct ksignal *ksig)
2662 {
2663 switch (siginfo_layout(ksig->sig, ksig->info.si_code)) {
2664 case SIL_FAULT:
2665 case SIL_FAULT_TRAPNO:
2666 case SIL_FAULT_MCEERR:
2667 case SIL_FAULT_BNDERR:
2668 case SIL_FAULT_PKUERR:
2669 case SIL_FAULT_PERF_EVENT:
2670 ksig->info.si_addr = arch_untagged_si_addr(
2671 ksig->info.si_addr, ksig->sig, ksig->info.si_code);
2672 break;
2673 case SIL_KILL:
2674 case SIL_TIMER:
2675 case SIL_POLL:
2676 case SIL_CHLD:
2677 case SIL_RT:
2678 case SIL_SYS:
2679 break;
2680 }
2681 }
2682
get_signal(struct ksignal * ksig)2683 bool get_signal(struct ksignal *ksig)
2684 {
2685 struct sighand_struct *sighand = current->sighand;
2686 struct signal_struct *signal = current->signal;
2687 int signr;
2688
2689 clear_notify_signal();
2690 if (unlikely(task_work_pending(current)))
2691 task_work_run();
2692
2693 if (!task_sigpending(current))
2694 return false;
2695
2696 if (unlikely(uprobe_deny_signal()))
2697 return false;
2698
2699 /*
2700 * Do this once, we can't return to user-mode if freezing() == T.
2701 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2702 * thus do not need another check after return.
2703 */
2704 try_to_freeze();
2705
2706 relock:
2707 spin_lock_irq(&sighand->siglock);
2708
2709 /*
2710 * Every stopped thread goes here after wakeup. Check to see if
2711 * we should notify the parent, prepare_signal(SIGCONT) encodes
2712 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2713 */
2714 if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2715 int why;
2716
2717 if (signal->flags & SIGNAL_CLD_CONTINUED)
2718 why = CLD_CONTINUED;
2719 else
2720 why = CLD_STOPPED;
2721
2722 signal->flags &= ~SIGNAL_CLD_MASK;
2723
2724 spin_unlock_irq(&sighand->siglock);
2725
2726 /*
2727 * Notify the parent that we're continuing. This event is
2728 * always per-process and doesn't make whole lot of sense
2729 * for ptracers, who shouldn't consume the state via
2730 * wait(2) either, but, for backward compatibility, notify
2731 * the ptracer of the group leader too unless it's gonna be
2732 * a duplicate.
2733 */
2734 read_lock(&tasklist_lock);
2735 do_notify_parent_cldstop(current, false, why);
2736
2737 if (ptrace_reparented(current->group_leader))
2738 do_notify_parent_cldstop(current->group_leader,
2739 true, why);
2740 read_unlock(&tasklist_lock);
2741
2742 goto relock;
2743 }
2744
2745 for (;;) {
2746 struct k_sigaction *ka;
2747 enum pid_type type;
2748
2749 /* Has this task already been marked for death? */
2750 if ((signal->flags & SIGNAL_GROUP_EXIT) ||
2751 signal->group_exec_task) {
2752 signr = SIGKILL;
2753 sigdelset(¤t->pending.signal, SIGKILL);
2754 trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
2755 &sighand->action[SIGKILL-1]);
2756 recalc_sigpending();
2757 /*
2758 * implies do_group_exit() or return to PF_USER_WORKER,
2759 * no need to initialize ksig->info/etc.
2760 */
2761 goto fatal;
2762 }
2763
2764 if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2765 do_signal_stop(0))
2766 goto relock;
2767
2768 if (unlikely(current->jobctl &
2769 (JOBCTL_TRAP_MASK | JOBCTL_TRAP_FREEZE))) {
2770 if (current->jobctl & JOBCTL_TRAP_MASK) {
2771 do_jobctl_trap();
2772 spin_unlock_irq(&sighand->siglock);
2773 } else if (current->jobctl & JOBCTL_TRAP_FREEZE)
2774 do_freezer_trap();
2775
2776 goto relock;
2777 }
2778
2779 /*
2780 * If the task is leaving the frozen state, let's update
2781 * cgroup counters and reset the frozen bit.
2782 */
2783 if (unlikely(cgroup_task_frozen(current))) {
2784 spin_unlock_irq(&sighand->siglock);
2785 cgroup_leave_frozen(false);
2786 goto relock;
2787 }
2788
2789 /*
2790 * Signals generated by the execution of an instruction
2791 * need to be delivered before any other pending signals
2792 * so that the instruction pointer in the signal stack
2793 * frame points to the faulting instruction.
2794 */
2795 type = PIDTYPE_PID;
2796 signr = dequeue_synchronous_signal(&ksig->info);
2797 if (!signr)
2798 signr = dequeue_signal(¤t->blocked, &ksig->info, &type);
2799
2800 if (!signr)
2801 break; /* will return 0 */
2802
2803 if (unlikely(current->ptrace) && (signr != SIGKILL) &&
2804 !(sighand->action[signr -1].sa.sa_flags & SA_IMMUTABLE)) {
2805 signr = ptrace_signal(signr, &ksig->info, type);
2806 if (!signr)
2807 continue;
2808 }
2809
2810 ka = &sighand->action[signr-1];
2811
2812 /* Trace actually delivered signals. */
2813 trace_signal_deliver(signr, &ksig->info, ka);
2814
2815 if (ka->sa.sa_handler == SIG_IGN) /* Do nothing. */
2816 continue;
2817 if (ka->sa.sa_handler != SIG_DFL) {
2818 /* Run the handler. */
2819 ksig->ka = *ka;
2820
2821 if (ka->sa.sa_flags & SA_ONESHOT)
2822 ka->sa.sa_handler = SIG_DFL;
2823
2824 break; /* will return non-zero "signr" value */
2825 }
2826
2827 /*
2828 * Now we are doing the default action for this signal.
2829 */
2830 if (sig_kernel_ignore(signr)) /* Default is nothing. */
2831 continue;
2832
2833 /*
2834 * Global init gets no signals it doesn't want.
2835 * Container-init gets no signals it doesn't want from same
2836 * container.
2837 *
2838 * Note that if global/container-init sees a sig_kernel_only()
2839 * signal here, the signal must have been generated internally
2840 * or must have come from an ancestor namespace. In either
2841 * case, the signal cannot be dropped.
2842 */
2843 if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2844 !sig_kernel_only(signr))
2845 continue;
2846
2847 if (sig_kernel_stop(signr)) {
2848 /*
2849 * The default action is to stop all threads in
2850 * the thread group. The job control signals
2851 * do nothing in an orphaned pgrp, but SIGSTOP
2852 * always works. Note that siglock needs to be
2853 * dropped during the call to is_orphaned_pgrp()
2854 * because of lock ordering with tasklist_lock.
2855 * This allows an intervening SIGCONT to be posted.
2856 * We need to check for that and bail out if necessary.
2857 */
2858 if (signr != SIGSTOP) {
2859 spin_unlock_irq(&sighand->siglock);
2860
2861 /* signals can be posted during this window */
2862
2863 if (is_current_pgrp_orphaned())
2864 goto relock;
2865
2866 spin_lock_irq(&sighand->siglock);
2867 }
2868
2869 if (likely(do_signal_stop(signr))) {
2870 /* It released the siglock. */
2871 goto relock;
2872 }
2873
2874 /*
2875 * We didn't actually stop, due to a race
2876 * with SIGCONT or something like that.
2877 */
2878 continue;
2879 }
2880
2881 fatal:
2882 spin_unlock_irq(&sighand->siglock);
2883 if (unlikely(cgroup_task_frozen(current)))
2884 cgroup_leave_frozen(true);
2885
2886 /*
2887 * Anything else is fatal, maybe with a core dump.
2888 */
2889 current->flags |= PF_SIGNALED;
2890
2891 if (sig_kernel_coredump(signr)) {
2892 if (print_fatal_signals)
2893 print_fatal_signal(signr);
2894 proc_coredump_connector(current);
2895 /*
2896 * If it was able to dump core, this kills all
2897 * other threads in the group and synchronizes with
2898 * their demise. If we lost the race with another
2899 * thread getting here, it set group_exit_code
2900 * first and our do_group_exit call below will use
2901 * that value and ignore the one we pass it.
2902 */
2903 do_coredump(&ksig->info);
2904 }
2905
2906 /*
2907 * PF_USER_WORKER threads will catch and exit on fatal signals
2908 * themselves. They have cleanup that must be performed, so we
2909 * cannot call do_exit() on their behalf. Note that ksig won't
2910 * be properly initialized, PF_USER_WORKER's shouldn't use it.
2911 */
2912 if (current->flags & PF_USER_WORKER)
2913 goto out;
2914
2915 /*
2916 * Death signals, no core dump.
2917 */
2918 do_group_exit(signr);
2919 /* NOTREACHED */
2920 }
2921 spin_unlock_irq(&sighand->siglock);
2922
2923 ksig->sig = signr;
2924
2925 if (signr && !(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
2926 hide_si_addr_tag_bits(ksig);
2927 out:
2928 return signr > 0;
2929 }
2930
2931 /**
2932 * signal_delivered - called after signal delivery to update blocked signals
2933 * @ksig: kernel signal struct
2934 * @stepping: nonzero if debugger single-step or block-step in use
2935 *
2936 * This function should be called when a signal has successfully been
2937 * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
2938 * is always blocked), and the signal itself is blocked unless %SA_NODEFER
2939 * is set in @ksig->ka.sa.sa_flags. Tracing is notified.
2940 */
signal_delivered(struct ksignal * ksig,int stepping)2941 static void signal_delivered(struct ksignal *ksig, int stepping)
2942 {
2943 sigset_t blocked;
2944
2945 /* A signal was successfully delivered, and the
2946 saved sigmask was stored on the signal frame,
2947 and will be restored by sigreturn. So we can
2948 simply clear the restore sigmask flag. */
2949 clear_restore_sigmask();
2950
2951 sigorsets(&blocked, ¤t->blocked, &ksig->ka.sa.sa_mask);
2952 if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
2953 sigaddset(&blocked, ksig->sig);
2954 set_current_blocked(&blocked);
2955 if (current->sas_ss_flags & SS_AUTODISARM)
2956 sas_ss_reset(current);
2957 if (stepping)
2958 ptrace_notify(SIGTRAP, 0);
2959 }
2960
signal_setup_done(int failed,struct ksignal * ksig,int stepping)2961 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
2962 {
2963 if (failed)
2964 force_sigsegv(ksig->sig);
2965 else
2966 signal_delivered(ksig, stepping);
2967 }
2968
2969 /*
2970 * It could be that complete_signal() picked us to notify about the
2971 * group-wide signal. Other threads should be notified now to take
2972 * the shared signals in @which since we will not.
2973 */
retarget_shared_pending(struct task_struct * tsk,sigset_t * which)2974 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2975 {
2976 sigset_t retarget;
2977 struct task_struct *t;
2978
2979 sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2980 if (sigisemptyset(&retarget))
2981 return;
2982
2983 for_other_threads(tsk, t) {
2984 if (t->flags & PF_EXITING)
2985 continue;
2986
2987 if (!has_pending_signals(&retarget, &t->blocked))
2988 continue;
2989 /* Remove the signals this thread can handle. */
2990 sigandsets(&retarget, &retarget, &t->blocked);
2991
2992 if (!task_sigpending(t))
2993 signal_wake_up(t, 0);
2994
2995 if (sigisemptyset(&retarget))
2996 break;
2997 }
2998 }
2999
exit_signals(struct task_struct * tsk)3000 void exit_signals(struct task_struct *tsk)
3001 {
3002 int group_stop = 0;
3003 sigset_t unblocked;
3004
3005 /*
3006 * @tsk is about to have PF_EXITING set - lock out users which
3007 * expect stable threadgroup.
3008 */
3009 cgroup_threadgroup_change_begin(tsk);
3010
3011 if (thread_group_empty(tsk) || (tsk->signal->flags & SIGNAL_GROUP_EXIT)) {
3012 sched_mm_cid_exit_signals(tsk);
3013 tsk->flags |= PF_EXITING;
3014 cgroup_threadgroup_change_end(tsk);
3015 return;
3016 }
3017
3018 spin_lock_irq(&tsk->sighand->siglock);
3019 /*
3020 * From now this task is not visible for group-wide signals,
3021 * see wants_signal(), do_signal_stop().
3022 */
3023 sched_mm_cid_exit_signals(tsk);
3024 tsk->flags |= PF_EXITING;
3025
3026 cgroup_threadgroup_change_end(tsk);
3027
3028 if (!task_sigpending(tsk))
3029 goto out;
3030
3031 unblocked = tsk->blocked;
3032 signotset(&unblocked);
3033 retarget_shared_pending(tsk, &unblocked);
3034
3035 if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
3036 task_participate_group_stop(tsk))
3037 group_stop = CLD_STOPPED;
3038 out:
3039 spin_unlock_irq(&tsk->sighand->siglock);
3040
3041 /*
3042 * If group stop has completed, deliver the notification. This
3043 * should always go to the real parent of the group leader.
3044 */
3045 if (unlikely(group_stop)) {
3046 read_lock(&tasklist_lock);
3047 do_notify_parent_cldstop(tsk, false, group_stop);
3048 read_unlock(&tasklist_lock);
3049 }
3050 }
3051
3052 /*
3053 * System call entry points.
3054 */
3055
3056 /**
3057 * sys_restart_syscall - restart a system call
3058 */
SYSCALL_DEFINE0(restart_syscall)3059 SYSCALL_DEFINE0(restart_syscall)
3060 {
3061 struct restart_block *restart = ¤t->restart_block;
3062 return restart->fn(restart);
3063 }
3064
do_no_restart_syscall(struct restart_block * param)3065 long do_no_restart_syscall(struct restart_block *param)
3066 {
3067 return -EINTR;
3068 }
3069
__set_task_blocked(struct task_struct * tsk,const sigset_t * newset)3070 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
3071 {
3072 if (task_sigpending(tsk) && !thread_group_empty(tsk)) {
3073 sigset_t newblocked;
3074 /* A set of now blocked but previously unblocked signals. */
3075 sigandnsets(&newblocked, newset, ¤t->blocked);
3076 retarget_shared_pending(tsk, &newblocked);
3077 }
3078 tsk->blocked = *newset;
3079 recalc_sigpending();
3080 }
3081
3082 /**
3083 * set_current_blocked - change current->blocked mask
3084 * @newset: new mask
3085 *
3086 * It is wrong to change ->blocked directly, this helper should be used
3087 * to ensure the process can't miss a shared signal we are going to block.
3088 */
set_current_blocked(sigset_t * newset)3089 void set_current_blocked(sigset_t *newset)
3090 {
3091 sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
3092 __set_current_blocked(newset);
3093 }
3094
__set_current_blocked(const sigset_t * newset)3095 void __set_current_blocked(const sigset_t *newset)
3096 {
3097 struct task_struct *tsk = current;
3098
3099 /*
3100 * In case the signal mask hasn't changed, there is nothing we need
3101 * to do. The current->blocked shouldn't be modified by other task.
3102 */
3103 if (sigequalsets(&tsk->blocked, newset))
3104 return;
3105
3106 spin_lock_irq(&tsk->sighand->siglock);
3107 __set_task_blocked(tsk, newset);
3108 spin_unlock_irq(&tsk->sighand->siglock);
3109 }
3110
3111 /*
3112 * This is also useful for kernel threads that want to temporarily
3113 * (or permanently) block certain signals.
3114 *
3115 * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
3116 * interface happily blocks "unblockable" signals like SIGKILL
3117 * and friends.
3118 */
sigprocmask(int how,sigset_t * set,sigset_t * oldset)3119 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
3120 {
3121 struct task_struct *tsk = current;
3122 sigset_t newset;
3123
3124 /* Lockless, only current can change ->blocked, never from irq */
3125 if (oldset)
3126 *oldset = tsk->blocked;
3127
3128 switch (how) {
3129 case SIG_BLOCK:
3130 sigorsets(&newset, &tsk->blocked, set);
3131 break;
3132 case SIG_UNBLOCK:
3133 sigandnsets(&newset, &tsk->blocked, set);
3134 break;
3135 case SIG_SETMASK:
3136 newset = *set;
3137 break;
3138 default:
3139 return -EINVAL;
3140 }
3141
3142 __set_current_blocked(&newset);
3143 return 0;
3144 }
3145 EXPORT_SYMBOL(sigprocmask);
3146
3147 /*
3148 * The api helps set app-provided sigmasks.
3149 *
3150 * This is useful for syscalls such as ppoll, pselect, io_pgetevents and
3151 * epoll_pwait where a new sigmask is passed from userland for the syscalls.
3152 *
3153 * Note that it does set_restore_sigmask() in advance, so it must be always
3154 * paired with restore_saved_sigmask_unless() before return from syscall.
3155 */
set_user_sigmask(const sigset_t __user * umask,size_t sigsetsize)3156 int set_user_sigmask(const sigset_t __user *umask, size_t sigsetsize)
3157 {
3158 sigset_t kmask;
3159
3160 if (!umask)
3161 return 0;
3162 if (sigsetsize != sizeof(sigset_t))
3163 return -EINVAL;
3164 if (copy_from_user(&kmask, umask, sizeof(sigset_t)))
3165 return -EFAULT;
3166
3167 set_restore_sigmask();
3168 current->saved_sigmask = current->blocked;
3169 set_current_blocked(&kmask);
3170
3171 return 0;
3172 }
3173
3174 #ifdef CONFIG_COMPAT
set_compat_user_sigmask(const compat_sigset_t __user * umask,size_t sigsetsize)3175 int set_compat_user_sigmask(const compat_sigset_t __user *umask,
3176 size_t sigsetsize)
3177 {
3178 sigset_t kmask;
3179
3180 if (!umask)
3181 return 0;
3182 if (sigsetsize != sizeof(compat_sigset_t))
3183 return -EINVAL;
3184 if (get_compat_sigset(&kmask, umask))
3185 return -EFAULT;
3186
3187 set_restore_sigmask();
3188 current->saved_sigmask = current->blocked;
3189 set_current_blocked(&kmask);
3190
3191 return 0;
3192 }
3193 #endif
3194
3195 /**
3196 * sys_rt_sigprocmask - change the list of currently blocked signals
3197 * @how: whether to add, remove, or set signals
3198 * @nset: stores pending signals
3199 * @oset: previous value of signal mask if non-null
3200 * @sigsetsize: size of sigset_t type
3201 */
SYSCALL_DEFINE4(rt_sigprocmask,int,how,sigset_t __user *,nset,sigset_t __user *,oset,size_t,sigsetsize)3202 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
3203 sigset_t __user *, oset, size_t, sigsetsize)
3204 {
3205 sigset_t old_set, new_set;
3206 int error;
3207
3208 /* XXX: Don't preclude handling different sized sigset_t's. */
3209 if (sigsetsize != sizeof(sigset_t))
3210 return -EINVAL;
3211
3212 old_set = current->blocked;
3213
3214 if (nset) {
3215 if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
3216 return -EFAULT;
3217 sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3218
3219 error = sigprocmask(how, &new_set, NULL);
3220 if (error)
3221 return error;
3222 }
3223
3224 if (oset) {
3225 if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
3226 return -EFAULT;
3227 }
3228
3229 return 0;
3230 }
3231
3232 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(rt_sigprocmask,int,how,compat_sigset_t __user *,nset,compat_sigset_t __user *,oset,compat_size_t,sigsetsize)3233 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
3234 compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
3235 {
3236 sigset_t old_set = current->blocked;
3237
3238 /* XXX: Don't preclude handling different sized sigset_t's. */
3239 if (sigsetsize != sizeof(sigset_t))
3240 return -EINVAL;
3241
3242 if (nset) {
3243 sigset_t new_set;
3244 int error;
3245 if (get_compat_sigset(&new_set, nset))
3246 return -EFAULT;
3247 sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3248
3249 error = sigprocmask(how, &new_set, NULL);
3250 if (error)
3251 return error;
3252 }
3253 return oset ? put_compat_sigset(oset, &old_set, sizeof(*oset)) : 0;
3254 }
3255 #endif
3256
do_sigpending(sigset_t * set)3257 static void do_sigpending(sigset_t *set)
3258 {
3259 spin_lock_irq(¤t->sighand->siglock);
3260 sigorsets(set, ¤t->pending.signal,
3261 ¤t->signal->shared_pending.signal);
3262 spin_unlock_irq(¤t->sighand->siglock);
3263
3264 /* Outside the lock because only this thread touches it. */
3265 sigandsets(set, ¤t->blocked, set);
3266 }
3267
3268 /**
3269 * sys_rt_sigpending - examine a pending signal that has been raised
3270 * while blocked
3271 * @uset: stores pending signals
3272 * @sigsetsize: size of sigset_t type or larger
3273 */
SYSCALL_DEFINE2(rt_sigpending,sigset_t __user *,uset,size_t,sigsetsize)3274 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
3275 {
3276 sigset_t set;
3277
3278 if (sigsetsize > sizeof(*uset))
3279 return -EINVAL;
3280
3281 do_sigpending(&set);
3282
3283 if (copy_to_user(uset, &set, sigsetsize))
3284 return -EFAULT;
3285
3286 return 0;
3287 }
3288
3289 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(rt_sigpending,compat_sigset_t __user *,uset,compat_size_t,sigsetsize)3290 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
3291 compat_size_t, sigsetsize)
3292 {
3293 sigset_t set;
3294
3295 if (sigsetsize > sizeof(*uset))
3296 return -EINVAL;
3297
3298 do_sigpending(&set);
3299
3300 return put_compat_sigset(uset, &set, sigsetsize);
3301 }
3302 #endif
3303
3304 static const struct {
3305 unsigned char limit, layout;
3306 } sig_sicodes[] = {
3307 [SIGILL] = { NSIGILL, SIL_FAULT },
3308 [SIGFPE] = { NSIGFPE, SIL_FAULT },
3309 [SIGSEGV] = { NSIGSEGV, SIL_FAULT },
3310 [SIGBUS] = { NSIGBUS, SIL_FAULT },
3311 [SIGTRAP] = { NSIGTRAP, SIL_FAULT },
3312 #if defined(SIGEMT)
3313 [SIGEMT] = { NSIGEMT, SIL_FAULT },
3314 #endif
3315 [SIGCHLD] = { NSIGCHLD, SIL_CHLD },
3316 [SIGPOLL] = { NSIGPOLL, SIL_POLL },
3317 [SIGSYS] = { NSIGSYS, SIL_SYS },
3318 };
3319
known_siginfo_layout(unsigned sig,int si_code)3320 static bool known_siginfo_layout(unsigned sig, int si_code)
3321 {
3322 if (si_code == SI_KERNEL)
3323 return true;
3324 else if ((si_code > SI_USER)) {
3325 if (sig_specific_sicodes(sig)) {
3326 if (si_code <= sig_sicodes[sig].limit)
3327 return true;
3328 }
3329 else if (si_code <= NSIGPOLL)
3330 return true;
3331 }
3332 else if (si_code >= SI_DETHREAD)
3333 return true;
3334 else if (si_code == SI_ASYNCNL)
3335 return true;
3336 return false;
3337 }
3338
siginfo_layout(unsigned sig,int si_code)3339 enum siginfo_layout siginfo_layout(unsigned sig, int si_code)
3340 {
3341 enum siginfo_layout layout = SIL_KILL;
3342 if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
3343 if ((sig < ARRAY_SIZE(sig_sicodes)) &&
3344 (si_code <= sig_sicodes[sig].limit)) {
3345 layout = sig_sicodes[sig].layout;
3346 /* Handle the exceptions */
3347 if ((sig == SIGBUS) &&
3348 (si_code >= BUS_MCEERR_AR) && (si_code <= BUS_MCEERR_AO))
3349 layout = SIL_FAULT_MCEERR;
3350 else if ((sig == SIGSEGV) && (si_code == SEGV_BNDERR))
3351 layout = SIL_FAULT_BNDERR;
3352 #ifdef SEGV_PKUERR
3353 else if ((sig == SIGSEGV) && (si_code == SEGV_PKUERR))
3354 layout = SIL_FAULT_PKUERR;
3355 #endif
3356 else if ((sig == SIGTRAP) && (si_code == TRAP_PERF))
3357 layout = SIL_FAULT_PERF_EVENT;
3358 else if (IS_ENABLED(CONFIG_SPARC) &&
3359 (sig == SIGILL) && (si_code == ILL_ILLTRP))
3360 layout = SIL_FAULT_TRAPNO;
3361 else if (IS_ENABLED(CONFIG_ALPHA) &&
3362 ((sig == SIGFPE) ||
3363 ((sig == SIGTRAP) && (si_code == TRAP_UNK))))
3364 layout = SIL_FAULT_TRAPNO;
3365 }
3366 else if (si_code <= NSIGPOLL)
3367 layout = SIL_POLL;
3368 } else {
3369 if (si_code == SI_TIMER)
3370 layout = SIL_TIMER;
3371 else if (si_code == SI_SIGIO)
3372 layout = SIL_POLL;
3373 else if (si_code < 0)
3374 layout = SIL_RT;
3375 }
3376 return layout;
3377 }
3378
si_expansion(const siginfo_t __user * info)3379 static inline char __user *si_expansion(const siginfo_t __user *info)
3380 {
3381 return ((char __user *)info) + sizeof(struct kernel_siginfo);
3382 }
3383
copy_siginfo_to_user(siginfo_t __user * to,const kernel_siginfo_t * from)3384 int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from)
3385 {
3386 char __user *expansion = si_expansion(to);
3387 if (copy_to_user(to, from , sizeof(struct kernel_siginfo)))
3388 return -EFAULT;
3389 if (clear_user(expansion, SI_EXPANSION_SIZE))
3390 return -EFAULT;
3391 return 0;
3392 }
3393
post_copy_siginfo_from_user(kernel_siginfo_t * info,const siginfo_t __user * from)3394 static int post_copy_siginfo_from_user(kernel_siginfo_t *info,
3395 const siginfo_t __user *from)
3396 {
3397 if (unlikely(!known_siginfo_layout(info->si_signo, info->si_code))) {
3398 char __user *expansion = si_expansion(from);
3399 char buf[SI_EXPANSION_SIZE];
3400 int i;
3401 /*
3402 * An unknown si_code might need more than
3403 * sizeof(struct kernel_siginfo) bytes. Verify all of the
3404 * extra bytes are 0. This guarantees copy_siginfo_to_user
3405 * will return this data to userspace exactly.
3406 */
3407 if (copy_from_user(&buf, expansion, SI_EXPANSION_SIZE))
3408 return -EFAULT;
3409 for (i = 0; i < SI_EXPANSION_SIZE; i++) {
3410 if (buf[i] != 0)
3411 return -E2BIG;
3412 }
3413 }
3414 return 0;
3415 }
3416
__copy_siginfo_from_user(int signo,kernel_siginfo_t * to,const siginfo_t __user * from)3417 static int __copy_siginfo_from_user(int signo, kernel_siginfo_t *to,
3418 const siginfo_t __user *from)
3419 {
3420 if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3421 return -EFAULT;
3422 to->si_signo = signo;
3423 return post_copy_siginfo_from_user(to, from);
3424 }
3425
copy_siginfo_from_user(kernel_siginfo_t * to,const siginfo_t __user * from)3426 int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from)
3427 {
3428 if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3429 return -EFAULT;
3430 return post_copy_siginfo_from_user(to, from);
3431 }
3432
3433 #ifdef CONFIG_COMPAT
3434 /**
3435 * copy_siginfo_to_external32 - copy a kernel siginfo into a compat user siginfo
3436 * @to: compat siginfo destination
3437 * @from: kernel siginfo source
3438 *
3439 * Note: This function does not work properly for the SIGCHLD on x32, but
3440 * fortunately it doesn't have to. The only valid callers for this function are
3441 * copy_siginfo_to_user32, which is overriden for x32 and the coredump code.
3442 * The latter does not care because SIGCHLD will never cause a coredump.
3443 */
copy_siginfo_to_external32(struct compat_siginfo * to,const struct kernel_siginfo * from)3444 void copy_siginfo_to_external32(struct compat_siginfo *to,
3445 const struct kernel_siginfo *from)
3446 {
3447 memset(to, 0, sizeof(*to));
3448
3449 to->si_signo = from->si_signo;
3450 to->si_errno = from->si_errno;
3451 to->si_code = from->si_code;
3452 switch(siginfo_layout(from->si_signo, from->si_code)) {
3453 case SIL_KILL:
3454 to->si_pid = from->si_pid;
3455 to->si_uid = from->si_uid;
3456 break;
3457 case SIL_TIMER:
3458 to->si_tid = from->si_tid;
3459 to->si_overrun = from->si_overrun;
3460 to->si_int = from->si_int;
3461 break;
3462 case SIL_POLL:
3463 to->si_band = from->si_band;
3464 to->si_fd = from->si_fd;
3465 break;
3466 case SIL_FAULT:
3467 to->si_addr = ptr_to_compat(from->si_addr);
3468 break;
3469 case SIL_FAULT_TRAPNO:
3470 to->si_addr = ptr_to_compat(from->si_addr);
3471 to->si_trapno = from->si_trapno;
3472 break;
3473 case SIL_FAULT_MCEERR:
3474 to->si_addr = ptr_to_compat(from->si_addr);
3475 to->si_addr_lsb = from->si_addr_lsb;
3476 break;
3477 case SIL_FAULT_BNDERR:
3478 to->si_addr = ptr_to_compat(from->si_addr);
3479 to->si_lower = ptr_to_compat(from->si_lower);
3480 to->si_upper = ptr_to_compat(from->si_upper);
3481 break;
3482 case SIL_FAULT_PKUERR:
3483 to->si_addr = ptr_to_compat(from->si_addr);
3484 to->si_pkey = from->si_pkey;
3485 break;
3486 case SIL_FAULT_PERF_EVENT:
3487 to->si_addr = ptr_to_compat(from->si_addr);
3488 to->si_perf_data = from->si_perf_data;
3489 to->si_perf_type = from->si_perf_type;
3490 to->si_perf_flags = from->si_perf_flags;
3491 break;
3492 case SIL_CHLD:
3493 to->si_pid = from->si_pid;
3494 to->si_uid = from->si_uid;
3495 to->si_status = from->si_status;
3496 to->si_utime = from->si_utime;
3497 to->si_stime = from->si_stime;
3498 break;
3499 case SIL_RT:
3500 to->si_pid = from->si_pid;
3501 to->si_uid = from->si_uid;
3502 to->si_int = from->si_int;
3503 break;
3504 case SIL_SYS:
3505 to->si_call_addr = ptr_to_compat(from->si_call_addr);
3506 to->si_syscall = from->si_syscall;
3507 to->si_arch = from->si_arch;
3508 break;
3509 }
3510 }
3511
__copy_siginfo_to_user32(struct compat_siginfo __user * to,const struct kernel_siginfo * from)3512 int __copy_siginfo_to_user32(struct compat_siginfo __user *to,
3513 const struct kernel_siginfo *from)
3514 {
3515 struct compat_siginfo new;
3516
3517 copy_siginfo_to_external32(&new, from);
3518 if (copy_to_user(to, &new, sizeof(struct compat_siginfo)))
3519 return -EFAULT;
3520 return 0;
3521 }
3522
post_copy_siginfo_from_user32(kernel_siginfo_t * to,const struct compat_siginfo * from)3523 static int post_copy_siginfo_from_user32(kernel_siginfo_t *to,
3524 const struct compat_siginfo *from)
3525 {
3526 clear_siginfo(to);
3527 to->si_signo = from->si_signo;
3528 to->si_errno = from->si_errno;
3529 to->si_code = from->si_code;
3530 switch(siginfo_layout(from->si_signo, from->si_code)) {
3531 case SIL_KILL:
3532 to->si_pid = from->si_pid;
3533 to->si_uid = from->si_uid;
3534 break;
3535 case SIL_TIMER:
3536 to->si_tid = from->si_tid;
3537 to->si_overrun = from->si_overrun;
3538 to->si_int = from->si_int;
3539 break;
3540 case SIL_POLL:
3541 to->si_band = from->si_band;
3542 to->si_fd = from->si_fd;
3543 break;
3544 case SIL_FAULT:
3545 to->si_addr = compat_ptr(from->si_addr);
3546 break;
3547 case SIL_FAULT_TRAPNO:
3548 to->si_addr = compat_ptr(from->si_addr);
3549 to->si_trapno = from->si_trapno;
3550 break;
3551 case SIL_FAULT_MCEERR:
3552 to->si_addr = compat_ptr(from->si_addr);
3553 to->si_addr_lsb = from->si_addr_lsb;
3554 break;
3555 case SIL_FAULT_BNDERR:
3556 to->si_addr = compat_ptr(from->si_addr);
3557 to->si_lower = compat_ptr(from->si_lower);
3558 to->si_upper = compat_ptr(from->si_upper);
3559 break;
3560 case SIL_FAULT_PKUERR:
3561 to->si_addr = compat_ptr(from->si_addr);
3562 to->si_pkey = from->si_pkey;
3563 break;
3564 case SIL_FAULT_PERF_EVENT:
3565 to->si_addr = compat_ptr(from->si_addr);
3566 to->si_perf_data = from->si_perf_data;
3567 to->si_perf_type = from->si_perf_type;
3568 to->si_perf_flags = from->si_perf_flags;
3569 break;
3570 case SIL_CHLD:
3571 to->si_pid = from->si_pid;
3572 to->si_uid = from->si_uid;
3573 to->si_status = from->si_status;
3574 #ifdef CONFIG_X86_X32_ABI
3575 if (in_x32_syscall()) {
3576 to->si_utime = from->_sifields._sigchld_x32._utime;
3577 to->si_stime = from->_sifields._sigchld_x32._stime;
3578 } else
3579 #endif
3580 {
3581 to->si_utime = from->si_utime;
3582 to->si_stime = from->si_stime;
3583 }
3584 break;
3585 case SIL_RT:
3586 to->si_pid = from->si_pid;
3587 to->si_uid = from->si_uid;
3588 to->si_int = from->si_int;
3589 break;
3590 case SIL_SYS:
3591 to->si_call_addr = compat_ptr(from->si_call_addr);
3592 to->si_syscall = from->si_syscall;
3593 to->si_arch = from->si_arch;
3594 break;
3595 }
3596 return 0;
3597 }
3598
__copy_siginfo_from_user32(int signo,struct kernel_siginfo * to,const struct compat_siginfo __user * ufrom)3599 static int __copy_siginfo_from_user32(int signo, struct kernel_siginfo *to,
3600 const struct compat_siginfo __user *ufrom)
3601 {
3602 struct compat_siginfo from;
3603
3604 if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3605 return -EFAULT;
3606
3607 from.si_signo = signo;
3608 return post_copy_siginfo_from_user32(to, &from);
3609 }
3610
copy_siginfo_from_user32(struct kernel_siginfo * to,const struct compat_siginfo __user * ufrom)3611 int copy_siginfo_from_user32(struct kernel_siginfo *to,
3612 const struct compat_siginfo __user *ufrom)
3613 {
3614 struct compat_siginfo from;
3615
3616 if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3617 return -EFAULT;
3618
3619 return post_copy_siginfo_from_user32(to, &from);
3620 }
3621 #endif /* CONFIG_COMPAT */
3622
3623 /**
3624 * do_sigtimedwait - wait for queued signals specified in @which
3625 * @which: queued signals to wait for
3626 * @info: if non-null, the signal's siginfo is returned here
3627 * @ts: upper bound on process time suspension
3628 */
do_sigtimedwait(const sigset_t * which,kernel_siginfo_t * info,const struct timespec64 * ts)3629 static int do_sigtimedwait(const sigset_t *which, kernel_siginfo_t *info,
3630 const struct timespec64 *ts)
3631 {
3632 ktime_t *to = NULL, timeout = KTIME_MAX;
3633 struct task_struct *tsk = current;
3634 sigset_t mask = *which;
3635 enum pid_type type;
3636 int sig, ret = 0;
3637
3638 if (ts) {
3639 if (!timespec64_valid(ts))
3640 return -EINVAL;
3641 timeout = timespec64_to_ktime(*ts);
3642 to = &timeout;
3643 }
3644
3645 /*
3646 * Invert the set of allowed signals to get those we want to block.
3647 */
3648 sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
3649 signotset(&mask);
3650
3651 spin_lock_irq(&tsk->sighand->siglock);
3652 sig = dequeue_signal(&mask, info, &type);
3653 if (!sig && timeout) {
3654 /*
3655 * None ready, temporarily unblock those we're interested
3656 * while we are sleeping in so that we'll be awakened when
3657 * they arrive. Unblocking is always fine, we can avoid
3658 * set_current_blocked().
3659 */
3660 tsk->real_blocked = tsk->blocked;
3661 sigandsets(&tsk->blocked, &tsk->blocked, &mask);
3662 recalc_sigpending();
3663 spin_unlock_irq(&tsk->sighand->siglock);
3664
3665 __set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
3666 ret = schedule_hrtimeout_range(to, tsk->timer_slack_ns,
3667 HRTIMER_MODE_REL);
3668 spin_lock_irq(&tsk->sighand->siglock);
3669 __set_task_blocked(tsk, &tsk->real_blocked);
3670 sigemptyset(&tsk->real_blocked);
3671 sig = dequeue_signal(&mask, info, &type);
3672 }
3673 spin_unlock_irq(&tsk->sighand->siglock);
3674
3675 if (sig)
3676 return sig;
3677 return ret ? -EINTR : -EAGAIN;
3678 }
3679
3680 /**
3681 * sys_rt_sigtimedwait - synchronously wait for queued signals specified
3682 * in @uthese
3683 * @uthese: queued signals to wait for
3684 * @uinfo: if non-null, the signal's siginfo is returned here
3685 * @uts: upper bound on process time suspension
3686 * @sigsetsize: size of sigset_t type
3687 */
SYSCALL_DEFINE4(rt_sigtimedwait,const sigset_t __user *,uthese,siginfo_t __user *,uinfo,const struct __kernel_timespec __user *,uts,size_t,sigsetsize)3688 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
3689 siginfo_t __user *, uinfo,
3690 const struct __kernel_timespec __user *, uts,
3691 size_t, sigsetsize)
3692 {
3693 sigset_t these;
3694 struct timespec64 ts;
3695 kernel_siginfo_t info;
3696 int ret;
3697
3698 /* XXX: Don't preclude handling different sized sigset_t's. */
3699 if (sigsetsize != sizeof(sigset_t))
3700 return -EINVAL;
3701
3702 if (copy_from_user(&these, uthese, sizeof(these)))
3703 return -EFAULT;
3704
3705 if (uts) {
3706 if (get_timespec64(&ts, uts))
3707 return -EFAULT;
3708 }
3709
3710 ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3711
3712 if (ret > 0 && uinfo) {
3713 if (copy_siginfo_to_user(uinfo, &info))
3714 ret = -EFAULT;
3715 }
3716
3717 return ret;
3718 }
3719
3720 #ifdef CONFIG_COMPAT_32BIT_TIME
SYSCALL_DEFINE4(rt_sigtimedwait_time32,const sigset_t __user *,uthese,siginfo_t __user *,uinfo,const struct old_timespec32 __user *,uts,size_t,sigsetsize)3721 SYSCALL_DEFINE4(rt_sigtimedwait_time32, const sigset_t __user *, uthese,
3722 siginfo_t __user *, uinfo,
3723 const struct old_timespec32 __user *, uts,
3724 size_t, sigsetsize)
3725 {
3726 sigset_t these;
3727 struct timespec64 ts;
3728 kernel_siginfo_t info;
3729 int ret;
3730
3731 if (sigsetsize != sizeof(sigset_t))
3732 return -EINVAL;
3733
3734 if (copy_from_user(&these, uthese, sizeof(these)))
3735 return -EFAULT;
3736
3737 if (uts) {
3738 if (get_old_timespec32(&ts, uts))
3739 return -EFAULT;
3740 }
3741
3742 ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3743
3744 if (ret > 0 && uinfo) {
3745 if (copy_siginfo_to_user(uinfo, &info))
3746 ret = -EFAULT;
3747 }
3748
3749 return ret;
3750 }
3751 #endif
3752
3753 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64,compat_sigset_t __user *,uthese,struct compat_siginfo __user *,uinfo,struct __kernel_timespec __user *,uts,compat_size_t,sigsetsize)3754 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64, compat_sigset_t __user *, uthese,
3755 struct compat_siginfo __user *, uinfo,
3756 struct __kernel_timespec __user *, uts, compat_size_t, sigsetsize)
3757 {
3758 sigset_t s;
3759 struct timespec64 t;
3760 kernel_siginfo_t info;
3761 long ret;
3762
3763 if (sigsetsize != sizeof(sigset_t))
3764 return -EINVAL;
3765
3766 if (get_compat_sigset(&s, uthese))
3767 return -EFAULT;
3768
3769 if (uts) {
3770 if (get_timespec64(&t, uts))
3771 return -EFAULT;
3772 }
3773
3774 ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3775
3776 if (ret > 0 && uinfo) {
3777 if (copy_siginfo_to_user32(uinfo, &info))
3778 ret = -EFAULT;
3779 }
3780
3781 return ret;
3782 }
3783
3784 #ifdef CONFIG_COMPAT_32BIT_TIME
COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32,compat_sigset_t __user *,uthese,struct compat_siginfo __user *,uinfo,struct old_timespec32 __user *,uts,compat_size_t,sigsetsize)3785 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32, compat_sigset_t __user *, uthese,
3786 struct compat_siginfo __user *, uinfo,
3787 struct old_timespec32 __user *, uts, compat_size_t, sigsetsize)
3788 {
3789 sigset_t s;
3790 struct timespec64 t;
3791 kernel_siginfo_t info;
3792 long ret;
3793
3794 if (sigsetsize != sizeof(sigset_t))
3795 return -EINVAL;
3796
3797 if (get_compat_sigset(&s, uthese))
3798 return -EFAULT;
3799
3800 if (uts) {
3801 if (get_old_timespec32(&t, uts))
3802 return -EFAULT;
3803 }
3804
3805 ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3806
3807 if (ret > 0 && uinfo) {
3808 if (copy_siginfo_to_user32(uinfo, &info))
3809 ret = -EFAULT;
3810 }
3811
3812 return ret;
3813 }
3814 #endif
3815 #endif
3816
prepare_kill_siginfo(int sig,struct kernel_siginfo * info,enum pid_type type)3817 static void prepare_kill_siginfo(int sig, struct kernel_siginfo *info,
3818 enum pid_type type)
3819 {
3820 clear_siginfo(info);
3821 info->si_signo = sig;
3822 info->si_errno = 0;
3823 info->si_code = (type == PIDTYPE_PID) ? SI_TKILL : SI_USER;
3824 info->si_pid = task_tgid_vnr(current);
3825 info->si_uid = from_kuid_munged(current_user_ns(), current_uid());
3826 }
3827
3828 /**
3829 * sys_kill - send a signal to a process
3830 * @pid: the PID of the process
3831 * @sig: signal to be sent
3832 */
SYSCALL_DEFINE2(kill,pid_t,pid,int,sig)3833 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
3834 {
3835 struct kernel_siginfo info;
3836
3837 prepare_kill_siginfo(sig, &info, PIDTYPE_TGID);
3838
3839 return kill_something_info(sig, &info, pid);
3840 }
3841
3842 /*
3843 * Verify that the signaler and signalee either are in the same pid namespace
3844 * or that the signaler's pid namespace is an ancestor of the signalee's pid
3845 * namespace.
3846 */
access_pidfd_pidns(struct pid * pid)3847 static bool access_pidfd_pidns(struct pid *pid)
3848 {
3849 struct pid_namespace *active = task_active_pid_ns(current);
3850 struct pid_namespace *p = ns_of_pid(pid);
3851
3852 for (;;) {
3853 if (!p)
3854 return false;
3855 if (p == active)
3856 break;
3857 p = p->parent;
3858 }
3859
3860 return true;
3861 }
3862
copy_siginfo_from_user_any(kernel_siginfo_t * kinfo,siginfo_t __user * info)3863 static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo,
3864 siginfo_t __user *info)
3865 {
3866 #ifdef CONFIG_COMPAT
3867 /*
3868 * Avoid hooking up compat syscalls and instead handle necessary
3869 * conversions here. Note, this is a stop-gap measure and should not be
3870 * considered a generic solution.
3871 */
3872 if (in_compat_syscall())
3873 return copy_siginfo_from_user32(
3874 kinfo, (struct compat_siginfo __user *)info);
3875 #endif
3876 return copy_siginfo_from_user(kinfo, info);
3877 }
3878
pidfd_to_pid(const struct file * file)3879 static struct pid *pidfd_to_pid(const struct file *file)
3880 {
3881 struct pid *pid;
3882
3883 pid = pidfd_pid(file);
3884 if (!IS_ERR(pid))
3885 return pid;
3886
3887 return tgid_pidfd_to_pid(file);
3888 }
3889
3890 #define PIDFD_SEND_SIGNAL_FLAGS \
3891 (PIDFD_SIGNAL_THREAD | PIDFD_SIGNAL_THREAD_GROUP | \
3892 PIDFD_SIGNAL_PROCESS_GROUP)
3893
3894 /**
3895 * sys_pidfd_send_signal - Signal a process through a pidfd
3896 * @pidfd: file descriptor of the process
3897 * @sig: signal to send
3898 * @info: signal info
3899 * @flags: future flags
3900 *
3901 * Send the signal to the thread group or to the individual thread depending
3902 * on PIDFD_THREAD.
3903 * In the future extension to @flags may be used to override the default scope
3904 * of @pidfd.
3905 *
3906 * Return: 0 on success, negative errno on failure
3907 */
SYSCALL_DEFINE4(pidfd_send_signal,int,pidfd,int,sig,siginfo_t __user *,info,unsigned int,flags)3908 SYSCALL_DEFINE4(pidfd_send_signal, int, pidfd, int, sig,
3909 siginfo_t __user *, info, unsigned int, flags)
3910 {
3911 int ret;
3912 struct fd f;
3913 struct pid *pid;
3914 kernel_siginfo_t kinfo;
3915 enum pid_type type;
3916
3917 /* Enforce flags be set to 0 until we add an extension. */
3918 if (flags & ~PIDFD_SEND_SIGNAL_FLAGS)
3919 return -EINVAL;
3920
3921 /* Ensure that only a single signal scope determining flag is set. */
3922 if (hweight32(flags & PIDFD_SEND_SIGNAL_FLAGS) > 1)
3923 return -EINVAL;
3924
3925 f = fdget(pidfd);
3926 if (!fd_file(f))
3927 return -EBADF;
3928
3929 /* Is this a pidfd? */
3930 pid = pidfd_to_pid(fd_file(f));
3931 if (IS_ERR(pid)) {
3932 ret = PTR_ERR(pid);
3933 goto err;
3934 }
3935
3936 ret = -EINVAL;
3937 if (!access_pidfd_pidns(pid))
3938 goto err;
3939
3940 switch (flags) {
3941 case 0:
3942 /* Infer scope from the type of pidfd. */
3943 if (fd_file(f)->f_flags & PIDFD_THREAD)
3944 type = PIDTYPE_PID;
3945 else
3946 type = PIDTYPE_TGID;
3947 break;
3948 case PIDFD_SIGNAL_THREAD:
3949 type = PIDTYPE_PID;
3950 break;
3951 case PIDFD_SIGNAL_THREAD_GROUP:
3952 type = PIDTYPE_TGID;
3953 break;
3954 case PIDFD_SIGNAL_PROCESS_GROUP:
3955 type = PIDTYPE_PGID;
3956 break;
3957 }
3958
3959 if (info) {
3960 ret = copy_siginfo_from_user_any(&kinfo, info);
3961 if (unlikely(ret))
3962 goto err;
3963
3964 ret = -EINVAL;
3965 if (unlikely(sig != kinfo.si_signo))
3966 goto err;
3967
3968 /* Only allow sending arbitrary signals to yourself. */
3969 ret = -EPERM;
3970 if ((task_pid(current) != pid || type > PIDTYPE_TGID) &&
3971 (kinfo.si_code >= 0 || kinfo.si_code == SI_TKILL))
3972 goto err;
3973 } else {
3974 prepare_kill_siginfo(sig, &kinfo, type);
3975 }
3976
3977 if (type == PIDTYPE_PGID)
3978 ret = kill_pgrp_info(sig, &kinfo, pid);
3979 else
3980 ret = kill_pid_info_type(sig, &kinfo, pid, type);
3981 err:
3982 fdput(f);
3983 return ret;
3984 }
3985
3986 static int
do_send_specific(pid_t tgid,pid_t pid,int sig,struct kernel_siginfo * info)3987 do_send_specific(pid_t tgid, pid_t pid, int sig, struct kernel_siginfo *info)
3988 {
3989 struct task_struct *p;
3990 int error = -ESRCH;
3991
3992 rcu_read_lock();
3993 p = find_task_by_vpid(pid);
3994 if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
3995 error = check_kill_permission(sig, info, p);
3996 /*
3997 * The null signal is a permissions and process existence
3998 * probe. No signal is actually delivered.
3999 */
4000 if (!error && sig) {
4001 error = do_send_sig_info(sig, info, p, PIDTYPE_PID);
4002 /*
4003 * If lock_task_sighand() failed we pretend the task
4004 * dies after receiving the signal. The window is tiny,
4005 * and the signal is private anyway.
4006 */
4007 if (unlikely(error == -ESRCH))
4008 error = 0;
4009 }
4010 }
4011 rcu_read_unlock();
4012
4013 return error;
4014 }
4015
do_tkill(pid_t tgid,pid_t pid,int sig)4016 static int do_tkill(pid_t tgid, pid_t pid, int sig)
4017 {
4018 struct kernel_siginfo info;
4019
4020 prepare_kill_siginfo(sig, &info, PIDTYPE_PID);
4021
4022 return do_send_specific(tgid, pid, sig, &info);
4023 }
4024
4025 /**
4026 * sys_tgkill - send signal to one specific thread
4027 * @tgid: the thread group ID of the thread
4028 * @pid: the PID of the thread
4029 * @sig: signal to be sent
4030 *
4031 * This syscall also checks the @tgid and returns -ESRCH even if the PID
4032 * exists but it's not belonging to the target process anymore. This
4033 * method solves the problem of threads exiting and PIDs getting reused.
4034 */
SYSCALL_DEFINE3(tgkill,pid_t,tgid,pid_t,pid,int,sig)4035 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
4036 {
4037 /* This is only valid for single tasks */
4038 if (pid <= 0 || tgid <= 0)
4039 return -EINVAL;
4040
4041 return do_tkill(tgid, pid, sig);
4042 }
4043
4044 /**
4045 * sys_tkill - send signal to one specific task
4046 * @pid: the PID of the task
4047 * @sig: signal to be sent
4048 *
4049 * Send a signal to only one task, even if it's a CLONE_THREAD task.
4050 */
SYSCALL_DEFINE2(tkill,pid_t,pid,int,sig)4051 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
4052 {
4053 /* This is only valid for single tasks */
4054 if (pid <= 0)
4055 return -EINVAL;
4056
4057 return do_tkill(0, pid, sig);
4058 }
4059
do_rt_sigqueueinfo(pid_t pid,int sig,kernel_siginfo_t * info)4060 static int do_rt_sigqueueinfo(pid_t pid, int sig, kernel_siginfo_t *info)
4061 {
4062 /* Not even root can pretend to send signals from the kernel.
4063 * Nor can they impersonate a kill()/tgkill(), which adds source info.
4064 */
4065 if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4066 (task_pid_vnr(current) != pid))
4067 return -EPERM;
4068
4069 /* POSIX.1b doesn't mention process groups. */
4070 return kill_proc_info(sig, info, pid);
4071 }
4072
4073 /**
4074 * sys_rt_sigqueueinfo - send signal information to a signal
4075 * @pid: the PID of the thread
4076 * @sig: signal to be sent
4077 * @uinfo: signal info to be sent
4078 */
SYSCALL_DEFINE3(rt_sigqueueinfo,pid_t,pid,int,sig,siginfo_t __user *,uinfo)4079 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
4080 siginfo_t __user *, uinfo)
4081 {
4082 kernel_siginfo_t info;
4083 int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4084 if (unlikely(ret))
4085 return ret;
4086 return do_rt_sigqueueinfo(pid, sig, &info);
4087 }
4088
4089 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,compat_pid_t,pid,int,sig,struct compat_siginfo __user *,uinfo)4090 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
4091 compat_pid_t, pid,
4092 int, sig,
4093 struct compat_siginfo __user *, uinfo)
4094 {
4095 kernel_siginfo_t info;
4096 int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4097 if (unlikely(ret))
4098 return ret;
4099 return do_rt_sigqueueinfo(pid, sig, &info);
4100 }
4101 #endif
4102
do_rt_tgsigqueueinfo(pid_t tgid,pid_t pid,int sig,kernel_siginfo_t * info)4103 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, kernel_siginfo_t *info)
4104 {
4105 /* This is only valid for single tasks */
4106 if (pid <= 0 || tgid <= 0)
4107 return -EINVAL;
4108
4109 /* Not even root can pretend to send signals from the kernel.
4110 * Nor can they impersonate a kill()/tgkill(), which adds source info.
4111 */
4112 if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4113 (task_pid_vnr(current) != pid))
4114 return -EPERM;
4115
4116 return do_send_specific(tgid, pid, sig, info);
4117 }
4118
SYSCALL_DEFINE4(rt_tgsigqueueinfo,pid_t,tgid,pid_t,pid,int,sig,siginfo_t __user *,uinfo)4119 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
4120 siginfo_t __user *, uinfo)
4121 {
4122 kernel_siginfo_t info;
4123 int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4124 if (unlikely(ret))
4125 return ret;
4126 return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4127 }
4128
4129 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,compat_pid_t,tgid,compat_pid_t,pid,int,sig,struct compat_siginfo __user *,uinfo)4130 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
4131 compat_pid_t, tgid,
4132 compat_pid_t, pid,
4133 int, sig,
4134 struct compat_siginfo __user *, uinfo)
4135 {
4136 kernel_siginfo_t info;
4137 int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4138 if (unlikely(ret))
4139 return ret;
4140 return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4141 }
4142 #endif
4143
4144 /*
4145 * For kthreads only, must not be used if cloned with CLONE_SIGHAND
4146 */
kernel_sigaction(int sig,__sighandler_t action)4147 void kernel_sigaction(int sig, __sighandler_t action)
4148 {
4149 spin_lock_irq(¤t->sighand->siglock);
4150 current->sighand->action[sig - 1].sa.sa_handler = action;
4151 if (action == SIG_IGN) {
4152 sigset_t mask;
4153
4154 sigemptyset(&mask);
4155 sigaddset(&mask, sig);
4156
4157 flush_sigqueue_mask(&mask, ¤t->signal->shared_pending);
4158 flush_sigqueue_mask(&mask, ¤t->pending);
4159 recalc_sigpending();
4160 }
4161 spin_unlock_irq(¤t->sighand->siglock);
4162 }
4163 EXPORT_SYMBOL(kernel_sigaction);
4164
sigaction_compat_abi(struct k_sigaction * act,struct k_sigaction * oact)4165 void __weak sigaction_compat_abi(struct k_sigaction *act,
4166 struct k_sigaction *oact)
4167 {
4168 }
4169
do_sigaction(int sig,struct k_sigaction * act,struct k_sigaction * oact)4170 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
4171 {
4172 struct task_struct *p = current, *t;
4173 struct k_sigaction *k;
4174 sigset_t mask;
4175
4176 if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
4177 return -EINVAL;
4178
4179 k = &p->sighand->action[sig-1];
4180
4181 spin_lock_irq(&p->sighand->siglock);
4182 if (k->sa.sa_flags & SA_IMMUTABLE) {
4183 spin_unlock_irq(&p->sighand->siglock);
4184 return -EINVAL;
4185 }
4186 if (oact)
4187 *oact = *k;
4188
4189 /*
4190 * Make sure that we never accidentally claim to support SA_UNSUPPORTED,
4191 * e.g. by having an architecture use the bit in their uapi.
4192 */
4193 BUILD_BUG_ON(UAPI_SA_FLAGS & SA_UNSUPPORTED);
4194
4195 /*
4196 * Clear unknown flag bits in order to allow userspace to detect missing
4197 * support for flag bits and to allow the kernel to use non-uapi bits
4198 * internally.
4199 */
4200 if (act)
4201 act->sa.sa_flags &= UAPI_SA_FLAGS;
4202 if (oact)
4203 oact->sa.sa_flags &= UAPI_SA_FLAGS;
4204
4205 sigaction_compat_abi(act, oact);
4206
4207 if (act) {
4208 sigdelsetmask(&act->sa.sa_mask,
4209 sigmask(SIGKILL) | sigmask(SIGSTOP));
4210 *k = *act;
4211 /*
4212 * POSIX 3.3.1.3:
4213 * "Setting a signal action to SIG_IGN for a signal that is
4214 * pending shall cause the pending signal to be discarded,
4215 * whether or not it is blocked."
4216 *
4217 * "Setting a signal action to SIG_DFL for a signal that is
4218 * pending and whose default action is to ignore the signal
4219 * (for example, SIGCHLD), shall cause the pending signal to
4220 * be discarded, whether or not it is blocked"
4221 */
4222 if (sig_handler_ignored(sig_handler(p, sig), sig)) {
4223 sigemptyset(&mask);
4224 sigaddset(&mask, sig);
4225 flush_sigqueue_mask(&mask, &p->signal->shared_pending);
4226 for_each_thread(p, t)
4227 flush_sigqueue_mask(&mask, &t->pending);
4228 }
4229 }
4230
4231 spin_unlock_irq(&p->sighand->siglock);
4232 return 0;
4233 }
4234
4235 #ifdef CONFIG_DYNAMIC_SIGFRAME
sigaltstack_lock(void)4236 static inline void sigaltstack_lock(void)
4237 __acquires(¤t->sighand->siglock)
4238 {
4239 spin_lock_irq(¤t->sighand->siglock);
4240 }
4241
sigaltstack_unlock(void)4242 static inline void sigaltstack_unlock(void)
4243 __releases(¤t->sighand->siglock)
4244 {
4245 spin_unlock_irq(¤t->sighand->siglock);
4246 }
4247 #else
sigaltstack_lock(void)4248 static inline void sigaltstack_lock(void) { }
sigaltstack_unlock(void)4249 static inline void sigaltstack_unlock(void) { }
4250 #endif
4251
4252 static int
do_sigaltstack(const stack_t * ss,stack_t * oss,unsigned long sp,size_t min_ss_size)4253 do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
4254 size_t min_ss_size)
4255 {
4256 struct task_struct *t = current;
4257 int ret = 0;
4258
4259 if (oss) {
4260 memset(oss, 0, sizeof(stack_t));
4261 oss->ss_sp = (void __user *) t->sas_ss_sp;
4262 oss->ss_size = t->sas_ss_size;
4263 oss->ss_flags = sas_ss_flags(sp) |
4264 (current->sas_ss_flags & SS_FLAG_BITS);
4265 }
4266
4267 if (ss) {
4268 void __user *ss_sp = ss->ss_sp;
4269 size_t ss_size = ss->ss_size;
4270 unsigned ss_flags = ss->ss_flags;
4271 int ss_mode;
4272
4273 if (unlikely(on_sig_stack(sp)))
4274 return -EPERM;
4275
4276 ss_mode = ss_flags & ~SS_FLAG_BITS;
4277 if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
4278 ss_mode != 0))
4279 return -EINVAL;
4280
4281 /*
4282 * Return before taking any locks if no actual
4283 * sigaltstack changes were requested.
4284 */
4285 if (t->sas_ss_sp == (unsigned long)ss_sp &&
4286 t->sas_ss_size == ss_size &&
4287 t->sas_ss_flags == ss_flags)
4288 return 0;
4289
4290 sigaltstack_lock();
4291 if (ss_mode == SS_DISABLE) {
4292 ss_size = 0;
4293 ss_sp = NULL;
4294 } else {
4295 if (unlikely(ss_size < min_ss_size))
4296 ret = -ENOMEM;
4297 if (!sigaltstack_size_valid(ss_size))
4298 ret = -ENOMEM;
4299 }
4300 if (!ret) {
4301 t->sas_ss_sp = (unsigned long) ss_sp;
4302 t->sas_ss_size = ss_size;
4303 t->sas_ss_flags = ss_flags;
4304 }
4305 sigaltstack_unlock();
4306 }
4307 return ret;
4308 }
4309
SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss,stack_t __user *,uoss)4310 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
4311 {
4312 stack_t new, old;
4313 int err;
4314 if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
4315 return -EFAULT;
4316 err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
4317 current_user_stack_pointer(),
4318 MINSIGSTKSZ);
4319 if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
4320 err = -EFAULT;
4321 return err;
4322 }
4323
restore_altstack(const stack_t __user * uss)4324 int restore_altstack(const stack_t __user *uss)
4325 {
4326 stack_t new;
4327 if (copy_from_user(&new, uss, sizeof(stack_t)))
4328 return -EFAULT;
4329 (void)do_sigaltstack(&new, NULL, current_user_stack_pointer(),
4330 MINSIGSTKSZ);
4331 /* squash all but EFAULT for now */
4332 return 0;
4333 }
4334
__save_altstack(stack_t __user * uss,unsigned long sp)4335 int __save_altstack(stack_t __user *uss, unsigned long sp)
4336 {
4337 struct task_struct *t = current;
4338 int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
4339 __put_user(t->sas_ss_flags, &uss->ss_flags) |
4340 __put_user(t->sas_ss_size, &uss->ss_size);
4341 return err;
4342 }
4343
4344 #ifdef CONFIG_COMPAT
do_compat_sigaltstack(const compat_stack_t __user * uss_ptr,compat_stack_t __user * uoss_ptr)4345 static int do_compat_sigaltstack(const compat_stack_t __user *uss_ptr,
4346 compat_stack_t __user *uoss_ptr)
4347 {
4348 stack_t uss, uoss;
4349 int ret;
4350
4351 if (uss_ptr) {
4352 compat_stack_t uss32;
4353 if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
4354 return -EFAULT;
4355 uss.ss_sp = compat_ptr(uss32.ss_sp);
4356 uss.ss_flags = uss32.ss_flags;
4357 uss.ss_size = uss32.ss_size;
4358 }
4359 ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
4360 compat_user_stack_pointer(),
4361 COMPAT_MINSIGSTKSZ);
4362 if (ret >= 0 && uoss_ptr) {
4363 compat_stack_t old;
4364 memset(&old, 0, sizeof(old));
4365 old.ss_sp = ptr_to_compat(uoss.ss_sp);
4366 old.ss_flags = uoss.ss_flags;
4367 old.ss_size = uoss.ss_size;
4368 if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
4369 ret = -EFAULT;
4370 }
4371 return ret;
4372 }
4373
COMPAT_SYSCALL_DEFINE2(sigaltstack,const compat_stack_t __user *,uss_ptr,compat_stack_t __user *,uoss_ptr)4374 COMPAT_SYSCALL_DEFINE2(sigaltstack,
4375 const compat_stack_t __user *, uss_ptr,
4376 compat_stack_t __user *, uoss_ptr)
4377 {
4378 return do_compat_sigaltstack(uss_ptr, uoss_ptr);
4379 }
4380
compat_restore_altstack(const compat_stack_t __user * uss)4381 int compat_restore_altstack(const compat_stack_t __user *uss)
4382 {
4383 int err = do_compat_sigaltstack(uss, NULL);
4384 /* squash all but -EFAULT for now */
4385 return err == -EFAULT ? err : 0;
4386 }
4387
__compat_save_altstack(compat_stack_t __user * uss,unsigned long sp)4388 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
4389 {
4390 int err;
4391 struct task_struct *t = current;
4392 err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
4393 &uss->ss_sp) |
4394 __put_user(t->sas_ss_flags, &uss->ss_flags) |
4395 __put_user(t->sas_ss_size, &uss->ss_size);
4396 return err;
4397 }
4398 #endif
4399
4400 #ifdef __ARCH_WANT_SYS_SIGPENDING
4401
4402 /**
4403 * sys_sigpending - examine pending signals
4404 * @uset: where mask of pending signal is returned
4405 */
SYSCALL_DEFINE1(sigpending,old_sigset_t __user *,uset)4406 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, uset)
4407 {
4408 sigset_t set;
4409
4410 if (sizeof(old_sigset_t) > sizeof(*uset))
4411 return -EINVAL;
4412
4413 do_sigpending(&set);
4414
4415 if (copy_to_user(uset, &set, sizeof(old_sigset_t)))
4416 return -EFAULT;
4417
4418 return 0;
4419 }
4420
4421 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE1(sigpending,compat_old_sigset_t __user *,set32)4422 COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
4423 {
4424 sigset_t set;
4425
4426 do_sigpending(&set);
4427
4428 return put_user(set.sig[0], set32);
4429 }
4430 #endif
4431
4432 #endif
4433
4434 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
4435 /**
4436 * sys_sigprocmask - examine and change blocked signals
4437 * @how: whether to add, remove, or set signals
4438 * @nset: signals to add or remove (if non-null)
4439 * @oset: previous value of signal mask if non-null
4440 *
4441 * Some platforms have their own version with special arguments;
4442 * others support only sys_rt_sigprocmask.
4443 */
4444
SYSCALL_DEFINE3(sigprocmask,int,how,old_sigset_t __user *,nset,old_sigset_t __user *,oset)4445 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
4446 old_sigset_t __user *, oset)
4447 {
4448 old_sigset_t old_set, new_set;
4449 sigset_t new_blocked;
4450
4451 old_set = current->blocked.sig[0];
4452
4453 if (nset) {
4454 if (copy_from_user(&new_set, nset, sizeof(*nset)))
4455 return -EFAULT;
4456
4457 new_blocked = current->blocked;
4458
4459 switch (how) {
4460 case SIG_BLOCK:
4461 sigaddsetmask(&new_blocked, new_set);
4462 break;
4463 case SIG_UNBLOCK:
4464 sigdelsetmask(&new_blocked, new_set);
4465 break;
4466 case SIG_SETMASK:
4467 new_blocked.sig[0] = new_set;
4468 break;
4469 default:
4470 return -EINVAL;
4471 }
4472
4473 set_current_blocked(&new_blocked);
4474 }
4475
4476 if (oset) {
4477 if (copy_to_user(oset, &old_set, sizeof(*oset)))
4478 return -EFAULT;
4479 }
4480
4481 return 0;
4482 }
4483 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
4484
4485 #ifndef CONFIG_ODD_RT_SIGACTION
4486 /**
4487 * sys_rt_sigaction - alter an action taken by a process
4488 * @sig: signal to be sent
4489 * @act: new sigaction
4490 * @oact: used to save the previous sigaction
4491 * @sigsetsize: size of sigset_t type
4492 */
SYSCALL_DEFINE4(rt_sigaction,int,sig,const struct sigaction __user *,act,struct sigaction __user *,oact,size_t,sigsetsize)4493 SYSCALL_DEFINE4(rt_sigaction, int, sig,
4494 const struct sigaction __user *, act,
4495 struct sigaction __user *, oact,
4496 size_t, sigsetsize)
4497 {
4498 struct k_sigaction new_sa, old_sa;
4499 int ret;
4500
4501 /* XXX: Don't preclude handling different sized sigset_t's. */
4502 if (sigsetsize != sizeof(sigset_t))
4503 return -EINVAL;
4504
4505 if (act && copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
4506 return -EFAULT;
4507
4508 ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
4509 if (ret)
4510 return ret;
4511
4512 if (oact && copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
4513 return -EFAULT;
4514
4515 return 0;
4516 }
4517 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(rt_sigaction,int,sig,const struct compat_sigaction __user *,act,struct compat_sigaction __user *,oact,compat_size_t,sigsetsize)4518 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
4519 const struct compat_sigaction __user *, act,
4520 struct compat_sigaction __user *, oact,
4521 compat_size_t, sigsetsize)
4522 {
4523 struct k_sigaction new_ka, old_ka;
4524 #ifdef __ARCH_HAS_SA_RESTORER
4525 compat_uptr_t restorer;
4526 #endif
4527 int ret;
4528
4529 /* XXX: Don't preclude handling different sized sigset_t's. */
4530 if (sigsetsize != sizeof(compat_sigset_t))
4531 return -EINVAL;
4532
4533 if (act) {
4534 compat_uptr_t handler;
4535 ret = get_user(handler, &act->sa_handler);
4536 new_ka.sa.sa_handler = compat_ptr(handler);
4537 #ifdef __ARCH_HAS_SA_RESTORER
4538 ret |= get_user(restorer, &act->sa_restorer);
4539 new_ka.sa.sa_restorer = compat_ptr(restorer);
4540 #endif
4541 ret |= get_compat_sigset(&new_ka.sa.sa_mask, &act->sa_mask);
4542 ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
4543 if (ret)
4544 return -EFAULT;
4545 }
4546
4547 ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4548 if (!ret && oact) {
4549 ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
4550 &oact->sa_handler);
4551 ret |= put_compat_sigset(&oact->sa_mask, &old_ka.sa.sa_mask,
4552 sizeof(oact->sa_mask));
4553 ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
4554 #ifdef __ARCH_HAS_SA_RESTORER
4555 ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4556 &oact->sa_restorer);
4557 #endif
4558 }
4559 return ret;
4560 }
4561 #endif
4562 #endif /* !CONFIG_ODD_RT_SIGACTION */
4563
4564 #ifdef CONFIG_OLD_SIGACTION
SYSCALL_DEFINE3(sigaction,int,sig,const struct old_sigaction __user *,act,struct old_sigaction __user *,oact)4565 SYSCALL_DEFINE3(sigaction, int, sig,
4566 const struct old_sigaction __user *, act,
4567 struct old_sigaction __user *, oact)
4568 {
4569 struct k_sigaction new_ka, old_ka;
4570 int ret;
4571
4572 if (act) {
4573 old_sigset_t mask;
4574 if (!access_ok(act, sizeof(*act)) ||
4575 __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
4576 __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
4577 __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4578 __get_user(mask, &act->sa_mask))
4579 return -EFAULT;
4580 #ifdef __ARCH_HAS_KA_RESTORER
4581 new_ka.ka_restorer = NULL;
4582 #endif
4583 siginitset(&new_ka.sa.sa_mask, mask);
4584 }
4585
4586 ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4587
4588 if (!ret && oact) {
4589 if (!access_ok(oact, sizeof(*oact)) ||
4590 __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
4591 __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
4592 __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4593 __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4594 return -EFAULT;
4595 }
4596
4597 return ret;
4598 }
4599 #endif
4600 #ifdef CONFIG_COMPAT_OLD_SIGACTION
COMPAT_SYSCALL_DEFINE3(sigaction,int,sig,const struct compat_old_sigaction __user *,act,struct compat_old_sigaction __user *,oact)4601 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
4602 const struct compat_old_sigaction __user *, act,
4603 struct compat_old_sigaction __user *, oact)
4604 {
4605 struct k_sigaction new_ka, old_ka;
4606 int ret;
4607 compat_old_sigset_t mask;
4608 compat_uptr_t handler, restorer;
4609
4610 if (act) {
4611 if (!access_ok(act, sizeof(*act)) ||
4612 __get_user(handler, &act->sa_handler) ||
4613 __get_user(restorer, &act->sa_restorer) ||
4614 __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4615 __get_user(mask, &act->sa_mask))
4616 return -EFAULT;
4617
4618 #ifdef __ARCH_HAS_KA_RESTORER
4619 new_ka.ka_restorer = NULL;
4620 #endif
4621 new_ka.sa.sa_handler = compat_ptr(handler);
4622 new_ka.sa.sa_restorer = compat_ptr(restorer);
4623 siginitset(&new_ka.sa.sa_mask, mask);
4624 }
4625
4626 ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4627
4628 if (!ret && oact) {
4629 if (!access_ok(oact, sizeof(*oact)) ||
4630 __put_user(ptr_to_compat(old_ka.sa.sa_handler),
4631 &oact->sa_handler) ||
4632 __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4633 &oact->sa_restorer) ||
4634 __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4635 __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4636 return -EFAULT;
4637 }
4638 return ret;
4639 }
4640 #endif
4641
4642 #ifdef CONFIG_SGETMASK_SYSCALL
4643
4644 /*
4645 * For backwards compatibility. Functionality superseded by sigprocmask.
4646 */
SYSCALL_DEFINE0(sgetmask)4647 SYSCALL_DEFINE0(sgetmask)
4648 {
4649 /* SMP safe */
4650 return current->blocked.sig[0];
4651 }
4652
SYSCALL_DEFINE1(ssetmask,int,newmask)4653 SYSCALL_DEFINE1(ssetmask, int, newmask)
4654 {
4655 int old = current->blocked.sig[0];
4656 sigset_t newset;
4657
4658 siginitset(&newset, newmask);
4659 set_current_blocked(&newset);
4660
4661 return old;
4662 }
4663 #endif /* CONFIG_SGETMASK_SYSCALL */
4664
4665 #ifdef __ARCH_WANT_SYS_SIGNAL
4666 /*
4667 * For backwards compatibility. Functionality superseded by sigaction.
4668 */
SYSCALL_DEFINE2(signal,int,sig,__sighandler_t,handler)4669 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
4670 {
4671 struct k_sigaction new_sa, old_sa;
4672 int ret;
4673
4674 new_sa.sa.sa_handler = handler;
4675 new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
4676 sigemptyset(&new_sa.sa.sa_mask);
4677
4678 ret = do_sigaction(sig, &new_sa, &old_sa);
4679
4680 return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
4681 }
4682 #endif /* __ARCH_WANT_SYS_SIGNAL */
4683
4684 #ifdef __ARCH_WANT_SYS_PAUSE
4685
SYSCALL_DEFINE0(pause)4686 SYSCALL_DEFINE0(pause)
4687 {
4688 while (!signal_pending(current)) {
4689 __set_current_state(TASK_INTERRUPTIBLE);
4690 schedule();
4691 }
4692 return -ERESTARTNOHAND;
4693 }
4694
4695 #endif
4696
sigsuspend(sigset_t * set)4697 static int sigsuspend(sigset_t *set)
4698 {
4699 current->saved_sigmask = current->blocked;
4700 set_current_blocked(set);
4701
4702 while (!signal_pending(current)) {
4703 __set_current_state(TASK_INTERRUPTIBLE);
4704 schedule();
4705 }
4706 set_restore_sigmask();
4707 return -ERESTARTNOHAND;
4708 }
4709
4710 /**
4711 * sys_rt_sigsuspend - replace the signal mask for a value with the
4712 * @unewset value until a signal is received
4713 * @unewset: new signal mask value
4714 * @sigsetsize: size of sigset_t type
4715 */
SYSCALL_DEFINE2(rt_sigsuspend,sigset_t __user *,unewset,size_t,sigsetsize)4716 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
4717 {
4718 sigset_t newset;
4719
4720 /* XXX: Don't preclude handling different sized sigset_t's. */
4721 if (sigsetsize != sizeof(sigset_t))
4722 return -EINVAL;
4723
4724 if (copy_from_user(&newset, unewset, sizeof(newset)))
4725 return -EFAULT;
4726 return sigsuspend(&newset);
4727 }
4728
4729 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(rt_sigsuspend,compat_sigset_t __user *,unewset,compat_size_t,sigsetsize)4730 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
4731 {
4732 sigset_t newset;
4733
4734 /* XXX: Don't preclude handling different sized sigset_t's. */
4735 if (sigsetsize != sizeof(sigset_t))
4736 return -EINVAL;
4737
4738 if (get_compat_sigset(&newset, unewset))
4739 return -EFAULT;
4740 return sigsuspend(&newset);
4741 }
4742 #endif
4743
4744 #ifdef CONFIG_OLD_SIGSUSPEND
SYSCALL_DEFINE1(sigsuspend,old_sigset_t,mask)4745 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
4746 {
4747 sigset_t blocked;
4748 siginitset(&blocked, mask);
4749 return sigsuspend(&blocked);
4750 }
4751 #endif
4752 #ifdef CONFIG_OLD_SIGSUSPEND3
SYSCALL_DEFINE3(sigsuspend,int,unused1,int,unused2,old_sigset_t,mask)4753 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
4754 {
4755 sigset_t blocked;
4756 siginitset(&blocked, mask);
4757 return sigsuspend(&blocked);
4758 }
4759 #endif
4760
arch_vma_name(struct vm_area_struct * vma)4761 __weak const char *arch_vma_name(struct vm_area_struct *vma)
4762 {
4763 return NULL;
4764 }
4765
siginfo_buildtime_checks(void)4766 static inline void siginfo_buildtime_checks(void)
4767 {
4768 BUILD_BUG_ON(sizeof(struct siginfo) != SI_MAX_SIZE);
4769
4770 /* Verify the offsets in the two siginfos match */
4771 #define CHECK_OFFSET(field) \
4772 BUILD_BUG_ON(offsetof(siginfo_t, field) != offsetof(kernel_siginfo_t, field))
4773
4774 /* kill */
4775 CHECK_OFFSET(si_pid);
4776 CHECK_OFFSET(si_uid);
4777
4778 /* timer */
4779 CHECK_OFFSET(si_tid);
4780 CHECK_OFFSET(si_overrun);
4781 CHECK_OFFSET(si_value);
4782
4783 /* rt */
4784 CHECK_OFFSET(si_pid);
4785 CHECK_OFFSET(si_uid);
4786 CHECK_OFFSET(si_value);
4787
4788 /* sigchld */
4789 CHECK_OFFSET(si_pid);
4790 CHECK_OFFSET(si_uid);
4791 CHECK_OFFSET(si_status);
4792 CHECK_OFFSET(si_utime);
4793 CHECK_OFFSET(si_stime);
4794
4795 /* sigfault */
4796 CHECK_OFFSET(si_addr);
4797 CHECK_OFFSET(si_trapno);
4798 CHECK_OFFSET(si_addr_lsb);
4799 CHECK_OFFSET(si_lower);
4800 CHECK_OFFSET(si_upper);
4801 CHECK_OFFSET(si_pkey);
4802 CHECK_OFFSET(si_perf_data);
4803 CHECK_OFFSET(si_perf_type);
4804 CHECK_OFFSET(si_perf_flags);
4805
4806 /* sigpoll */
4807 CHECK_OFFSET(si_band);
4808 CHECK_OFFSET(si_fd);
4809
4810 /* sigsys */
4811 CHECK_OFFSET(si_call_addr);
4812 CHECK_OFFSET(si_syscall);
4813 CHECK_OFFSET(si_arch);
4814 #undef CHECK_OFFSET
4815
4816 /* usb asyncio */
4817 BUILD_BUG_ON(offsetof(struct siginfo, si_pid) !=
4818 offsetof(struct siginfo, si_addr));
4819 if (sizeof(int) == sizeof(void __user *)) {
4820 BUILD_BUG_ON(sizeof_field(struct siginfo, si_pid) !=
4821 sizeof(void __user *));
4822 } else {
4823 BUILD_BUG_ON((sizeof_field(struct siginfo, si_pid) +
4824 sizeof_field(struct siginfo, si_uid)) !=
4825 sizeof(void __user *));
4826 BUILD_BUG_ON(offsetofend(struct siginfo, si_pid) !=
4827 offsetof(struct siginfo, si_uid));
4828 }
4829 #ifdef CONFIG_COMPAT
4830 BUILD_BUG_ON(offsetof(struct compat_siginfo, si_pid) !=
4831 offsetof(struct compat_siginfo, si_addr));
4832 BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4833 sizeof(compat_uptr_t));
4834 BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4835 sizeof_field(struct siginfo, si_pid));
4836 #endif
4837 }
4838
4839 #if defined(CONFIG_SYSCTL)
4840 static struct ctl_table signal_debug_table[] = {
4841 #ifdef CONFIG_SYSCTL_EXCEPTION_TRACE
4842 {
4843 .procname = "exception-trace",
4844 .data = &show_unhandled_signals,
4845 .maxlen = sizeof(int),
4846 .mode = 0644,
4847 .proc_handler = proc_dointvec
4848 },
4849 #endif
4850 };
4851
init_signal_sysctls(void)4852 static int __init init_signal_sysctls(void)
4853 {
4854 register_sysctl_init("debug", signal_debug_table);
4855 return 0;
4856 }
4857 early_initcall(init_signal_sysctls);
4858 #endif /* CONFIG_SYSCTL */
4859
signals_init(void)4860 void __init signals_init(void)
4861 {
4862 siginfo_buildtime_checks();
4863
4864 sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC | SLAB_ACCOUNT);
4865 }
4866
4867 #ifdef CONFIG_KGDB_KDB
4868 #include <linux/kdb.h>
4869 /*
4870 * kdb_send_sig - Allows kdb to send signals without exposing
4871 * signal internals. This function checks if the required locks are
4872 * available before calling the main signal code, to avoid kdb
4873 * deadlocks.
4874 */
kdb_send_sig(struct task_struct * t,int sig)4875 void kdb_send_sig(struct task_struct *t, int sig)
4876 {
4877 static struct task_struct *kdb_prev_t;
4878 int new_t, ret;
4879 if (!spin_trylock(&t->sighand->siglock)) {
4880 kdb_printf("Can't do kill command now.\n"
4881 "The sigmask lock is held somewhere else in "
4882 "kernel, try again later\n");
4883 return;
4884 }
4885 new_t = kdb_prev_t != t;
4886 kdb_prev_t = t;
4887 if (!task_is_running(t) && new_t) {
4888 spin_unlock(&t->sighand->siglock);
4889 kdb_printf("Process is not RUNNING, sending a signal from "
4890 "kdb risks deadlock\n"
4891 "on the run queue locks. "
4892 "The signal has _not_ been sent.\n"
4893 "Reissue the kill command if you want to risk "
4894 "the deadlock.\n");
4895 return;
4896 }
4897 ret = send_signal_locked(sig, SEND_SIG_PRIV, t, PIDTYPE_PID);
4898 spin_unlock(&t->sighand->siglock);
4899 if (ret)
4900 kdb_printf("Fail to deliver Signal %d to process %d.\n",
4901 sig, t->pid);
4902 else
4903 kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
4904 }
4905 #endif /* CONFIG_KGDB_KDB */
4906