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/proc_fs.h>
26 #include <linux/tty.h>
27 #include <linux/binfmts.h>
28 #include <linux/coredump.h>
29 #include <linux/security.h>
30 #include <linux/syscalls.h>
31 #include <linux/ptrace.h>
32 #include <linux/signal.h>
33 #include <linux/signalfd.h>
34 #include <linux/ratelimit.h>
35 #include <linux/tracehook.h>
36 #include <linux/capability.h>
37 #include <linux/freezer.h>
38 #include <linux/pid_namespace.h>
39 #include <linux/nsproxy.h>
40 #include <linux/user_namespace.h>
41 #include <linux/uprobes.h>
42 #include <linux/compat.h>
43 #include <linux/cn_proc.h>
44 #include <linux/compiler.h>
45 #include <linux/posix-timers.h>
46 #include <linux/cgroup.h>
47 #include <linux/audit.h>
48
49 #define CREATE_TRACE_POINTS
50 #include <trace/events/signal.h>
51
52 #include <asm/param.h>
53 #include <linux/uaccess.h>
54 #include <asm/unistd.h>
55 #include <asm/siginfo.h>
56 #include <asm/cacheflush.h>
57 #include <asm/syscall.h> /* for syscall_get_* */
58
59 /*
60 * SLAB caches for signal bits.
61 */
62
63 static struct kmem_cache *sigqueue_cachep;
64
65 int print_fatal_signals __read_mostly;
66
sig_handler(struct task_struct * t,int sig)67 static void __user *sig_handler(struct task_struct *t, int sig)
68 {
69 return t->sighand->action[sig - 1].sa.sa_handler;
70 }
71
sig_handler_ignored(void __user * handler,int sig)72 static inline bool sig_handler_ignored(void __user *handler, int sig)
73 {
74 /* Is it explicitly or implicitly ignored? */
75 return handler == SIG_IGN ||
76 (handler == SIG_DFL && sig_kernel_ignore(sig));
77 }
78
sig_task_ignored(struct task_struct * t,int sig,bool force)79 static bool sig_task_ignored(struct task_struct *t, int sig, bool force)
80 {
81 void __user *handler;
82
83 handler = sig_handler(t, sig);
84
85 /* SIGKILL and SIGSTOP may not be sent to the global init */
86 if (unlikely(is_global_init(t) && sig_kernel_only(sig)))
87 return true;
88
89 if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
90 handler == SIG_DFL && !(force && sig_kernel_only(sig)))
91 return true;
92
93 /* Only allow kernel generated signals to this kthread */
94 if (unlikely((t->flags & PF_KTHREAD) &&
95 (handler == SIG_KTHREAD_KERNEL) && !force))
96 return true;
97
98 return sig_handler_ignored(handler, sig);
99 }
100
sig_ignored(struct task_struct * t,int sig,bool force)101 static bool sig_ignored(struct task_struct *t, int sig, bool force)
102 {
103 /*
104 * Blocked signals are never ignored, since the
105 * signal handler may change by the time it is
106 * unblocked.
107 */
108 if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
109 return false;
110
111 /*
112 * Tracers may want to know about even ignored signal unless it
113 * is SIGKILL which can't be reported anyway but can be ignored
114 * by SIGNAL_UNKILLABLE task.
115 */
116 if (t->ptrace && sig != SIGKILL)
117 return false;
118
119 return sig_task_ignored(t, sig, force);
120 }
121
122 /*
123 * Re-calculate pending state from the set of locally pending
124 * signals, globally pending signals, and blocked signals.
125 */
has_pending_signals(sigset_t * signal,sigset_t * blocked)126 static inline bool has_pending_signals(sigset_t *signal, sigset_t *blocked)
127 {
128 unsigned long ready;
129 long i;
130
131 switch (_NSIG_WORDS) {
132 default:
133 for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
134 ready |= signal->sig[i] &~ blocked->sig[i];
135 break;
136
137 case 4: ready = signal->sig[3] &~ blocked->sig[3];
138 ready |= signal->sig[2] &~ blocked->sig[2];
139 ready |= signal->sig[1] &~ blocked->sig[1];
140 ready |= signal->sig[0] &~ blocked->sig[0];
141 break;
142
143 case 2: ready = signal->sig[1] &~ blocked->sig[1];
144 ready |= signal->sig[0] &~ blocked->sig[0];
145 break;
146
147 case 1: ready = signal->sig[0] &~ blocked->sig[0];
148 }
149 return ready != 0;
150 }
151
152 #define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
153
recalc_sigpending_tsk(struct task_struct * t)154 static bool recalc_sigpending_tsk(struct task_struct *t)
155 {
156 if ((t->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) ||
157 PENDING(&t->pending, &t->blocked) ||
158 PENDING(&t->signal->shared_pending, &t->blocked) ||
159 cgroup_task_frozen(t)) {
160 set_tsk_thread_flag(t, TIF_SIGPENDING);
161 return true;
162 }
163
164 /*
165 * We must never clear the flag in another thread, or in current
166 * when it's possible the current syscall is returning -ERESTART*.
167 * So we don't clear it here, and only callers who know they should do.
168 */
169 return false;
170 }
171
172 /*
173 * After recalculating TIF_SIGPENDING, we need to make sure the task wakes up.
174 * This is superfluous when called on current, the wakeup is a harmless no-op.
175 */
recalc_sigpending_and_wake(struct task_struct * t)176 void recalc_sigpending_and_wake(struct task_struct *t)
177 {
178 if (recalc_sigpending_tsk(t))
179 signal_wake_up(t, 0);
180 }
181
recalc_sigpending(void)182 void recalc_sigpending(void)
183 {
184 if (!recalc_sigpending_tsk(current) && !freezing(current))
185 clear_thread_flag(TIF_SIGPENDING);
186
187 }
188 EXPORT_SYMBOL(recalc_sigpending);
189
calculate_sigpending(void)190 void calculate_sigpending(void)
191 {
192 /* Have any signals or users of TIF_SIGPENDING been delayed
193 * until after fork?
194 */
195 spin_lock_irq(¤t->sighand->siglock);
196 set_tsk_thread_flag(current, TIF_SIGPENDING);
197 recalc_sigpending();
198 spin_unlock_irq(¤t->sighand->siglock);
199 }
200
201 /* Given the mask, find the first available signal that should be serviced. */
202
203 #define SYNCHRONOUS_MASK \
204 (sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
205 sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
206
next_signal(struct sigpending * pending,sigset_t * mask)207 int next_signal(struct sigpending *pending, sigset_t *mask)
208 {
209 unsigned long i, *s, *m, x;
210 int sig = 0;
211
212 s = pending->signal.sig;
213 m = mask->sig;
214
215 /*
216 * Handle the first word specially: it contains the
217 * synchronous signals that need to be dequeued first.
218 */
219 x = *s &~ *m;
220 if (x) {
221 if (x & SYNCHRONOUS_MASK)
222 x &= SYNCHRONOUS_MASK;
223 sig = ffz(~x) + 1;
224 return sig;
225 }
226
227 switch (_NSIG_WORDS) {
228 default:
229 for (i = 1; i < _NSIG_WORDS; ++i) {
230 x = *++s &~ *++m;
231 if (!x)
232 continue;
233 sig = ffz(~x) + i*_NSIG_BPW + 1;
234 break;
235 }
236 break;
237
238 case 2:
239 x = s[1] &~ m[1];
240 if (!x)
241 break;
242 sig = ffz(~x) + _NSIG_BPW + 1;
243 break;
244
245 case 1:
246 /* Nothing to do */
247 break;
248 }
249
250 return sig;
251 }
252
print_dropped_signal(int sig)253 static inline void print_dropped_signal(int sig)
254 {
255 static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
256
257 if (!print_fatal_signals)
258 return;
259
260 if (!__ratelimit(&ratelimit_state))
261 return;
262
263 pr_info("%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
264 current->comm, current->pid, sig);
265 }
266
267 /**
268 * task_set_jobctl_pending - set jobctl pending bits
269 * @task: target task
270 * @mask: pending bits to set
271 *
272 * Clear @mask from @task->jobctl. @mask must be subset of
273 * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
274 * %JOBCTL_TRAPPING. If stop signo is being set, the existing signo is
275 * cleared. If @task is already being killed or exiting, this function
276 * becomes noop.
277 *
278 * CONTEXT:
279 * Must be called with @task->sighand->siglock held.
280 *
281 * RETURNS:
282 * %true if @mask is set, %false if made noop because @task was dying.
283 */
task_set_jobctl_pending(struct task_struct * task,unsigned long mask)284 bool task_set_jobctl_pending(struct task_struct *task, unsigned long mask)
285 {
286 BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
287 JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
288 BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
289
290 if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
291 return false;
292
293 if (mask & JOBCTL_STOP_SIGMASK)
294 task->jobctl &= ~JOBCTL_STOP_SIGMASK;
295
296 task->jobctl |= mask;
297 return true;
298 }
299
300 /**
301 * task_clear_jobctl_trapping - clear jobctl trapping bit
302 * @task: target task
303 *
304 * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
305 * Clear it and wake up the ptracer. Note that we don't need any further
306 * locking. @task->siglock guarantees that @task->parent points to the
307 * ptracer.
308 *
309 * CONTEXT:
310 * Must be called with @task->sighand->siglock held.
311 */
task_clear_jobctl_trapping(struct task_struct * task)312 void task_clear_jobctl_trapping(struct task_struct *task)
313 {
314 if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
315 task->jobctl &= ~JOBCTL_TRAPPING;
316 smp_mb(); /* advised by wake_up_bit() */
317 wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
318 }
319 }
320
321 /**
322 * task_clear_jobctl_pending - clear jobctl pending bits
323 * @task: target task
324 * @mask: pending bits to clear
325 *
326 * Clear @mask from @task->jobctl. @mask must be subset of
327 * %JOBCTL_PENDING_MASK. If %JOBCTL_STOP_PENDING is being cleared, other
328 * STOP bits are cleared together.
329 *
330 * If clearing of @mask leaves no stop or trap pending, this function calls
331 * task_clear_jobctl_trapping().
332 *
333 * CONTEXT:
334 * Must be called with @task->sighand->siglock held.
335 */
task_clear_jobctl_pending(struct task_struct * task,unsigned long mask)336 void task_clear_jobctl_pending(struct task_struct *task, unsigned long mask)
337 {
338 BUG_ON(mask & ~JOBCTL_PENDING_MASK);
339
340 if (mask & JOBCTL_STOP_PENDING)
341 mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
342
343 task->jobctl &= ~mask;
344
345 if (!(task->jobctl & JOBCTL_PENDING_MASK))
346 task_clear_jobctl_trapping(task);
347 }
348
349 /**
350 * task_participate_group_stop - participate in a group stop
351 * @task: task participating in a group stop
352 *
353 * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
354 * Group stop states are cleared and the group stop count is consumed if
355 * %JOBCTL_STOP_CONSUME was set. If the consumption completes the group
356 * stop, the appropriate `SIGNAL_*` flags are set.
357 *
358 * CONTEXT:
359 * Must be called with @task->sighand->siglock held.
360 *
361 * RETURNS:
362 * %true if group stop completion should be notified to the parent, %false
363 * otherwise.
364 */
task_participate_group_stop(struct task_struct * task)365 static bool task_participate_group_stop(struct task_struct *task)
366 {
367 struct signal_struct *sig = task->signal;
368 bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
369
370 WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
371
372 task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
373
374 if (!consume)
375 return false;
376
377 if (!WARN_ON_ONCE(sig->group_stop_count == 0))
378 sig->group_stop_count--;
379
380 /*
381 * Tell the caller to notify completion iff we are entering into a
382 * fresh group stop. Read comment in do_signal_stop() for details.
383 */
384 if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
385 signal_set_stop_flags(sig, SIGNAL_STOP_STOPPED);
386 return true;
387 }
388 return false;
389 }
390
task_join_group_stop(struct task_struct * task)391 void task_join_group_stop(struct task_struct *task)
392 {
393 unsigned long mask = current->jobctl & JOBCTL_STOP_SIGMASK;
394 struct signal_struct *sig = current->signal;
395
396 if (sig->group_stop_count) {
397 sig->group_stop_count++;
398 mask |= JOBCTL_STOP_CONSUME;
399 } else if (!(sig->flags & SIGNAL_STOP_STOPPED))
400 return;
401
402 /* Have the new thread join an on-going signal group stop */
403 task_set_jobctl_pending(task, mask | JOBCTL_STOP_PENDING);
404 }
405
406 /*
407 * allocate a new signal queue record
408 * - this may be called without locks if and only if t == current, otherwise an
409 * appropriate lock must be held to stop the target task from exiting
410 */
411 static struct sigqueue *
__sigqueue_alloc(int sig,struct task_struct * t,gfp_t gfp_flags,int override_rlimit,const unsigned int sigqueue_flags)412 __sigqueue_alloc(int sig, struct task_struct *t, gfp_t gfp_flags,
413 int override_rlimit, const unsigned int sigqueue_flags)
414 {
415 struct sigqueue *q = NULL;
416 struct ucounts *ucounts = NULL;
417 long sigpending;
418
419 /*
420 * Protect access to @t credentials. This can go away when all
421 * callers hold rcu read lock.
422 *
423 * NOTE! A pending signal will hold on to the user refcount,
424 * and we get/put the refcount only when the sigpending count
425 * changes from/to zero.
426 */
427 rcu_read_lock();
428 ucounts = task_ucounts(t);
429 sigpending = inc_rlimit_get_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
430 rcu_read_unlock();
431 if (!sigpending)
432 return NULL;
433
434 if (override_rlimit || likely(sigpending <= task_rlimit(t, RLIMIT_SIGPENDING))) {
435 q = kmem_cache_alloc(sigqueue_cachep, gfp_flags);
436 } else {
437 print_dropped_signal(sig);
438 }
439
440 if (unlikely(q == NULL)) {
441 dec_rlimit_put_ucounts(ucounts, UCOUNT_RLIMIT_SIGPENDING);
442 } else {
443 INIT_LIST_HEAD(&q->list);
444 q->flags = sigqueue_flags;
445 q->ucounts = ucounts;
446 }
447 return q;
448 }
449
__sigqueue_free(struct sigqueue * q)450 static void __sigqueue_free(struct sigqueue *q)
451 {
452 if (q->flags & SIGQUEUE_PREALLOC)
453 return;
454 if (q->ucounts) {
455 dec_rlimit_put_ucounts(q->ucounts, UCOUNT_RLIMIT_SIGPENDING);
456 q->ucounts = NULL;
457 }
458 kmem_cache_free(sigqueue_cachep, q);
459 }
460
flush_sigqueue(struct sigpending * queue)461 void flush_sigqueue(struct sigpending *queue)
462 {
463 struct sigqueue *q;
464
465 sigemptyset(&queue->signal);
466 while (!list_empty(&queue->list)) {
467 q = list_entry(queue->list.next, struct sigqueue , list);
468 list_del_init(&q->list);
469 __sigqueue_free(q);
470 }
471 }
472
473 /*
474 * Flush all pending signals for this kthread.
475 */
flush_signals(struct task_struct * t)476 void flush_signals(struct task_struct *t)
477 {
478 unsigned long flags;
479
480 spin_lock_irqsave(&t->sighand->siglock, flags);
481 clear_tsk_thread_flag(t, TIF_SIGPENDING);
482 flush_sigqueue(&t->pending);
483 flush_sigqueue(&t->signal->shared_pending);
484 spin_unlock_irqrestore(&t->sighand->siglock, flags);
485 }
486 EXPORT_SYMBOL(flush_signals);
487
488 #ifdef CONFIG_POSIX_TIMERS
__flush_itimer_signals(struct sigpending * pending)489 static void __flush_itimer_signals(struct sigpending *pending)
490 {
491 sigset_t signal, retain;
492 struct sigqueue *q, *n;
493
494 signal = pending->signal;
495 sigemptyset(&retain);
496
497 list_for_each_entry_safe(q, n, &pending->list, list) {
498 int sig = q->info.si_signo;
499
500 if (likely(q->info.si_code != SI_TIMER)) {
501 sigaddset(&retain, sig);
502 } else {
503 sigdelset(&signal, sig);
504 list_del_init(&q->list);
505 __sigqueue_free(q);
506 }
507 }
508
509 sigorsets(&pending->signal, &signal, &retain);
510 }
511
flush_itimer_signals(void)512 void flush_itimer_signals(void)
513 {
514 struct task_struct *tsk = current;
515 unsigned long flags;
516
517 spin_lock_irqsave(&tsk->sighand->siglock, flags);
518 __flush_itimer_signals(&tsk->pending);
519 __flush_itimer_signals(&tsk->signal->shared_pending);
520 spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
521 }
522 #endif
523
ignore_signals(struct task_struct * t)524 void ignore_signals(struct task_struct *t)
525 {
526 int i;
527
528 for (i = 0; i < _NSIG; ++i)
529 t->sighand->action[i].sa.sa_handler = SIG_IGN;
530
531 flush_signals(t);
532 }
533
534 /*
535 * Flush all handlers for a task.
536 */
537
538 void
flush_signal_handlers(struct task_struct * t,int force_default)539 flush_signal_handlers(struct task_struct *t, int force_default)
540 {
541 int i;
542 struct k_sigaction *ka = &t->sighand->action[0];
543 for (i = _NSIG ; i != 0 ; i--) {
544 if (force_default || ka->sa.sa_handler != SIG_IGN)
545 ka->sa.sa_handler = SIG_DFL;
546 ka->sa.sa_flags = 0;
547 #ifdef __ARCH_HAS_SA_RESTORER
548 ka->sa.sa_restorer = NULL;
549 #endif
550 sigemptyset(&ka->sa.sa_mask);
551 ka++;
552 }
553 }
554
unhandled_signal(struct task_struct * tsk,int sig)555 bool unhandled_signal(struct task_struct *tsk, int sig)
556 {
557 void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
558 if (is_global_init(tsk))
559 return true;
560
561 if (handler != SIG_IGN && handler != SIG_DFL)
562 return false;
563
564 /* if ptraced, let the tracer determine */
565 return !tsk->ptrace;
566 }
567
collect_signal(int sig,struct sigpending * list,kernel_siginfo_t * info,bool * resched_timer)568 static void collect_signal(int sig, struct sigpending *list, kernel_siginfo_t *info,
569 bool *resched_timer)
570 {
571 struct sigqueue *q, *first = NULL;
572
573 /*
574 * Collect the siginfo appropriate to this signal. Check if
575 * there is another siginfo for the same signal.
576 */
577 list_for_each_entry(q, &list->list, list) {
578 if (q->info.si_signo == sig) {
579 if (first)
580 goto still_pending;
581 first = q;
582 }
583 }
584
585 sigdelset(&list->signal, sig);
586
587 if (first) {
588 still_pending:
589 list_del_init(&first->list);
590 copy_siginfo(info, &first->info);
591
592 *resched_timer =
593 (first->flags & SIGQUEUE_PREALLOC) &&
594 (info->si_code == SI_TIMER) &&
595 (info->si_sys_private);
596
597 __sigqueue_free(first);
598 } else {
599 /*
600 * Ok, it wasn't in the queue. This must be
601 * a fast-pathed signal or we must have been
602 * out of queue space. So zero out the info.
603 */
604 clear_siginfo(info);
605 info->si_signo = sig;
606 info->si_errno = 0;
607 info->si_code = SI_USER;
608 info->si_pid = 0;
609 info->si_uid = 0;
610 }
611 }
612
__dequeue_signal(struct sigpending * pending,sigset_t * mask,kernel_siginfo_t * info,bool * resched_timer)613 static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
614 kernel_siginfo_t *info, bool *resched_timer)
615 {
616 int sig = next_signal(pending, mask);
617
618 if (sig)
619 collect_signal(sig, pending, info, resched_timer);
620 return sig;
621 }
622
623 /*
624 * Dequeue a signal and return the element to the caller, which is
625 * expected to free it.
626 *
627 * All callers have to hold the siglock.
628 */
dequeue_signal(struct task_struct * tsk,sigset_t * mask,kernel_siginfo_t * info)629 int dequeue_signal(struct task_struct *tsk, sigset_t *mask, kernel_siginfo_t *info)
630 {
631 bool resched_timer = false;
632 int signr;
633
634 /* We only dequeue private signals from ourselves, we don't let
635 * signalfd steal them
636 */
637 signr = __dequeue_signal(&tsk->pending, mask, info, &resched_timer);
638 if (!signr) {
639 signr = __dequeue_signal(&tsk->signal->shared_pending,
640 mask, info, &resched_timer);
641 #ifdef CONFIG_POSIX_TIMERS
642 /*
643 * itimer signal ?
644 *
645 * itimers are process shared and we restart periodic
646 * itimers in the signal delivery path to prevent DoS
647 * attacks in the high resolution timer case. This is
648 * compliant with the old way of self-restarting
649 * itimers, as the SIGALRM is a legacy signal and only
650 * queued once. Changing the restart behaviour to
651 * restart the timer in the signal dequeue path is
652 * reducing the timer noise on heavy loaded !highres
653 * systems too.
654 */
655 if (unlikely(signr == SIGALRM)) {
656 struct hrtimer *tmr = &tsk->signal->real_timer;
657
658 if (!hrtimer_is_queued(tmr) &&
659 tsk->signal->it_real_incr != 0) {
660 hrtimer_forward(tmr, tmr->base->get_time(),
661 tsk->signal->it_real_incr);
662 hrtimer_restart(tmr);
663 }
664 }
665 #endif
666 }
667
668 recalc_sigpending();
669 if (!signr)
670 return 0;
671
672 if (unlikely(sig_kernel_stop(signr))) {
673 /*
674 * Set a marker that we have dequeued a stop signal. Our
675 * caller might release the siglock and then the pending
676 * stop signal it is about to process is no longer in the
677 * pending bitmasks, but must still be cleared by a SIGCONT
678 * (and overruled by a SIGKILL). So those cases clear this
679 * shared flag after we've set it. Note that this flag may
680 * remain set after the signal we return is ignored or
681 * handled. That doesn't matter because its only purpose
682 * is to alert stop-signal processing code when another
683 * processor has come along and cleared the flag.
684 */
685 current->jobctl |= JOBCTL_STOP_DEQUEUED;
686 }
687 #ifdef CONFIG_POSIX_TIMERS
688 if (resched_timer) {
689 /*
690 * Release the siglock to ensure proper locking order
691 * of timer locks outside of siglocks. Note, we leave
692 * irqs disabled here, since the posix-timers code is
693 * about to disable them again anyway.
694 */
695 spin_unlock(&tsk->sighand->siglock);
696 posixtimer_rearm(info);
697 spin_lock(&tsk->sighand->siglock);
698
699 /* Don't expose the si_sys_private value to userspace */
700 info->si_sys_private = 0;
701 }
702 #endif
703 return signr;
704 }
705 EXPORT_SYMBOL_GPL(dequeue_signal);
706
dequeue_synchronous_signal(kernel_siginfo_t * info)707 static int dequeue_synchronous_signal(kernel_siginfo_t *info)
708 {
709 struct task_struct *tsk = current;
710 struct sigpending *pending = &tsk->pending;
711 struct sigqueue *q, *sync = NULL;
712
713 /*
714 * Might a synchronous signal be in the queue?
715 */
716 if (!((pending->signal.sig[0] & ~tsk->blocked.sig[0]) & SYNCHRONOUS_MASK))
717 return 0;
718
719 /*
720 * Return the first synchronous signal in the queue.
721 */
722 list_for_each_entry(q, &pending->list, list) {
723 /* Synchronous signals have a positive si_code */
724 if ((q->info.si_code > SI_USER) &&
725 (sigmask(q->info.si_signo) & SYNCHRONOUS_MASK)) {
726 sync = q;
727 goto next;
728 }
729 }
730 return 0;
731 next:
732 /*
733 * Check if there is another siginfo for the same signal.
734 */
735 list_for_each_entry_continue(q, &pending->list, list) {
736 if (q->info.si_signo == sync->info.si_signo)
737 goto still_pending;
738 }
739
740 sigdelset(&pending->signal, sync->info.si_signo);
741 recalc_sigpending();
742 still_pending:
743 list_del_init(&sync->list);
744 copy_siginfo(info, &sync->info);
745 __sigqueue_free(sync);
746 return info->si_signo;
747 }
748
749 /*
750 * Tell a process that it has a new active signal..
751 *
752 * NOTE! we rely on the previous spin_lock to
753 * lock interrupts for us! We can only be called with
754 * "siglock" held, and the local interrupt must
755 * have been disabled when that got acquired!
756 *
757 * No need to set need_resched since signal event passing
758 * goes through ->blocked
759 */
signal_wake_up_state(struct task_struct * t,unsigned int state)760 void signal_wake_up_state(struct task_struct *t, unsigned int state)
761 {
762 set_tsk_thread_flag(t, TIF_SIGPENDING);
763 /*
764 * TASK_WAKEKILL also means wake it up in the stopped/traced/killable
765 * case. We don't check t->state here because there is a race with it
766 * executing another processor and just now entering stopped state.
767 * By using wake_up_state, we ensure the process will wake up and
768 * handle its death signal.
769 */
770 if (!wake_up_state(t, state | TASK_INTERRUPTIBLE))
771 kick_process(t);
772 }
773
774 /*
775 * Remove signals in mask from the pending set and queue.
776 * Returns 1 if any signals were found.
777 *
778 * All callers must be holding the siglock.
779 */
flush_sigqueue_mask(sigset_t * mask,struct sigpending * s)780 static void flush_sigqueue_mask(sigset_t *mask, struct sigpending *s)
781 {
782 struct sigqueue *q, *n;
783 sigset_t m;
784
785 sigandsets(&m, mask, &s->signal);
786 if (sigisemptyset(&m))
787 return;
788
789 sigandnsets(&s->signal, &s->signal, mask);
790 list_for_each_entry_safe(q, n, &s->list, list) {
791 if (sigismember(mask, q->info.si_signo)) {
792 list_del_init(&q->list);
793 __sigqueue_free(q);
794 }
795 }
796 }
797
is_si_special(const struct kernel_siginfo * info)798 static inline int is_si_special(const struct kernel_siginfo *info)
799 {
800 return info <= SEND_SIG_PRIV;
801 }
802
si_fromuser(const struct kernel_siginfo * info)803 static inline bool si_fromuser(const struct kernel_siginfo *info)
804 {
805 return info == SEND_SIG_NOINFO ||
806 (!is_si_special(info) && SI_FROMUSER(info));
807 }
808
809 /*
810 * called with RCU read lock from check_kill_permission()
811 */
kill_ok_by_cred(struct task_struct * t)812 static bool kill_ok_by_cred(struct task_struct *t)
813 {
814 const struct cred *cred = current_cred();
815 const struct cred *tcred = __task_cred(t);
816
817 return uid_eq(cred->euid, tcred->suid) ||
818 uid_eq(cred->euid, tcred->uid) ||
819 uid_eq(cred->uid, tcred->suid) ||
820 uid_eq(cred->uid, tcred->uid) ||
821 ns_capable(tcred->user_ns, CAP_KILL);
822 }
823
824 /*
825 * Bad permissions for sending the signal
826 * - the caller must hold the RCU read lock
827 */
check_kill_permission(int sig,struct kernel_siginfo * info,struct task_struct * t)828 static int check_kill_permission(int sig, struct kernel_siginfo *info,
829 struct task_struct *t)
830 {
831 struct pid *sid;
832 int error;
833
834 if (!valid_signal(sig))
835 return -EINVAL;
836
837 if (!si_fromuser(info))
838 return 0;
839
840 error = audit_signal_info(sig, t); /* Let audit system see the signal */
841 if (error)
842 return error;
843
844 if (!same_thread_group(current, t) &&
845 !kill_ok_by_cred(t)) {
846 switch (sig) {
847 case SIGCONT:
848 sid = task_session(t);
849 /*
850 * We don't return the error if sid == NULL. The
851 * task was unhashed, the caller must notice this.
852 */
853 if (!sid || sid == task_session(current))
854 break;
855 fallthrough;
856 default:
857 return -EPERM;
858 }
859 }
860
861 return security_task_kill(t, info, sig, NULL);
862 }
863
864 /**
865 * ptrace_trap_notify - schedule trap to notify ptracer
866 * @t: tracee wanting to notify tracer
867 *
868 * This function schedules sticky ptrace trap which is cleared on the next
869 * TRAP_STOP to notify ptracer of an event. @t must have been seized by
870 * ptracer.
871 *
872 * If @t is running, STOP trap will be taken. If trapped for STOP and
873 * ptracer is listening for events, tracee is woken up so that it can
874 * re-trap for the new event. If trapped otherwise, STOP trap will be
875 * eventually taken without returning to userland after the existing traps
876 * are finished by PTRACE_CONT.
877 *
878 * CONTEXT:
879 * Must be called with @task->sighand->siglock held.
880 */
ptrace_trap_notify(struct task_struct * t)881 static void ptrace_trap_notify(struct task_struct *t)
882 {
883 WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
884 assert_spin_locked(&t->sighand->siglock);
885
886 task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
887 ptrace_signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
888 }
889
890 /*
891 * Handle magic process-wide effects of stop/continue signals. Unlike
892 * the signal actions, these happen immediately at signal-generation
893 * time regardless of blocking, ignoring, or handling. This does the
894 * actual continuing for SIGCONT, but not the actual stopping for stop
895 * signals. The process stop is done as a signal action for SIG_DFL.
896 *
897 * Returns true if the signal should be actually delivered, otherwise
898 * it should be dropped.
899 */
prepare_signal(int sig,struct task_struct * p,bool force)900 static bool prepare_signal(int sig, struct task_struct *p, bool force)
901 {
902 struct signal_struct *signal = p->signal;
903 struct task_struct *t;
904 sigset_t flush;
905
906 if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
907 if (!(signal->flags & SIGNAL_GROUP_EXIT))
908 return sig == SIGKILL;
909 /*
910 * The process is in the middle of dying, nothing to do.
911 */
912 } else if (sig_kernel_stop(sig)) {
913 /*
914 * This is a stop signal. Remove SIGCONT from all queues.
915 */
916 siginitset(&flush, sigmask(SIGCONT));
917 flush_sigqueue_mask(&flush, &signal->shared_pending);
918 for_each_thread(p, t)
919 flush_sigqueue_mask(&flush, &t->pending);
920 } else if (sig == SIGCONT) {
921 unsigned int why;
922 /*
923 * Remove all stop signals from all queues, wake all threads.
924 */
925 siginitset(&flush, SIG_KERNEL_STOP_MASK);
926 flush_sigqueue_mask(&flush, &signal->shared_pending);
927 for_each_thread(p, t) {
928 flush_sigqueue_mask(&flush, &t->pending);
929 task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
930 if (likely(!(t->ptrace & PT_SEIZED)))
931 wake_up_state(t, __TASK_STOPPED);
932 else
933 ptrace_trap_notify(t);
934 }
935
936 /*
937 * Notify the parent with CLD_CONTINUED if we were stopped.
938 *
939 * If we were in the middle of a group stop, we pretend it
940 * was already finished, and then continued. Since SIGCHLD
941 * doesn't queue we report only CLD_STOPPED, as if the next
942 * CLD_CONTINUED was dropped.
943 */
944 why = 0;
945 if (signal->flags & SIGNAL_STOP_STOPPED)
946 why |= SIGNAL_CLD_CONTINUED;
947 else if (signal->group_stop_count)
948 why |= SIGNAL_CLD_STOPPED;
949
950 if (why) {
951 /*
952 * The first thread which returns from do_signal_stop()
953 * will take ->siglock, notice SIGNAL_CLD_MASK, and
954 * notify its parent. See get_signal().
955 */
956 signal_set_stop_flags(signal, why | SIGNAL_STOP_CONTINUED);
957 signal->group_stop_count = 0;
958 signal->group_exit_code = 0;
959 }
960 }
961
962 return !sig_ignored(p, sig, force);
963 }
964
965 /*
966 * Test if P wants to take SIG. After we've checked all threads with this,
967 * it's equivalent to finding no threads not blocking SIG. Any threads not
968 * blocking SIG were ruled out because they are not running and already
969 * have pending signals. Such threads will dequeue from the shared queue
970 * as soon as they're available, so putting the signal on the shared queue
971 * will be equivalent to sending it to one such thread.
972 */
wants_signal(int sig,struct task_struct * p)973 static inline bool wants_signal(int sig, struct task_struct *p)
974 {
975 if (sigismember(&p->blocked, sig))
976 return false;
977
978 if (p->flags & PF_EXITING)
979 return false;
980
981 if (sig == SIGKILL)
982 return true;
983
984 if (task_is_stopped_or_traced(p))
985 return false;
986
987 return task_curr(p) || !task_sigpending(p);
988 }
989
complete_signal(int sig,struct task_struct * p,enum pid_type type)990 static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
991 {
992 struct signal_struct *signal = p->signal;
993 struct task_struct *t;
994
995 /*
996 * Now find a thread we can wake up to take the signal off the queue.
997 *
998 * If the main thread wants the signal, it gets first crack.
999 * Probably the least surprising to the average bear.
1000 */
1001 if (wants_signal(sig, p))
1002 t = p;
1003 else if ((type == PIDTYPE_PID) || thread_group_empty(p))
1004 /*
1005 * There is just one thread and it does not need to be woken.
1006 * It will dequeue unblocked signals before it runs again.
1007 */
1008 return;
1009 else {
1010 /*
1011 * Otherwise try to find a suitable thread.
1012 */
1013 t = signal->curr_target;
1014 while (!wants_signal(sig, t)) {
1015 t = next_thread(t);
1016 if (t == signal->curr_target)
1017 /*
1018 * No thread needs to be woken.
1019 * Any eligible threads will see
1020 * the signal in the queue soon.
1021 */
1022 return;
1023 }
1024 signal->curr_target = t;
1025 }
1026
1027 /*
1028 * Found a killable thread. If the signal will be fatal,
1029 * then start taking the whole group down immediately.
1030 */
1031 if (sig_fatal(p, sig) &&
1032 !(signal->flags & SIGNAL_GROUP_EXIT) &&
1033 !sigismember(&t->real_blocked, sig) &&
1034 (sig == SIGKILL || !p->ptrace)) {
1035 /*
1036 * This signal will be fatal to the whole group.
1037 */
1038 if (!sig_kernel_coredump(sig)) {
1039 /*
1040 * Start a group exit and wake everybody up.
1041 * This way we don't have other threads
1042 * running and doing things after a slower
1043 * thread has the fatal signal pending.
1044 */
1045 signal->flags = SIGNAL_GROUP_EXIT;
1046 signal->group_exit_code = sig;
1047 signal->group_stop_count = 0;
1048 t = p;
1049 do {
1050 task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1051 sigaddset(&t->pending.signal, SIGKILL);
1052 signal_wake_up(t, 1);
1053 } while_each_thread(p, t);
1054 return;
1055 }
1056 }
1057
1058 /*
1059 * The signal is already in the shared-pending queue.
1060 * Tell the chosen thread to wake up and dequeue it.
1061 */
1062 signal_wake_up(t, sig == SIGKILL);
1063 return;
1064 }
1065
legacy_queue(struct sigpending * signals,int sig)1066 static inline bool legacy_queue(struct sigpending *signals, int sig)
1067 {
1068 return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
1069 }
1070
__send_signal(int sig,struct kernel_siginfo * info,struct task_struct * t,enum pid_type type,bool force)1071 static int __send_signal(int sig, struct kernel_siginfo *info, struct task_struct *t,
1072 enum pid_type type, bool force)
1073 {
1074 struct sigpending *pending;
1075 struct sigqueue *q;
1076 int override_rlimit;
1077 int ret = 0, result;
1078
1079 assert_spin_locked(&t->sighand->siglock);
1080
1081 result = TRACE_SIGNAL_IGNORED;
1082 if (!prepare_signal(sig, t, force))
1083 goto ret;
1084
1085 pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1086 /*
1087 * Short-circuit ignored signals and support queuing
1088 * exactly one non-rt signal, so that we can get more
1089 * detailed information about the cause of the signal.
1090 */
1091 result = TRACE_SIGNAL_ALREADY_PENDING;
1092 if (legacy_queue(pending, sig))
1093 goto ret;
1094
1095 result = TRACE_SIGNAL_DELIVERED;
1096 /*
1097 * Skip useless siginfo allocation for SIGKILL and kernel threads.
1098 */
1099 if ((sig == SIGKILL) || (t->flags & PF_KTHREAD))
1100 goto out_set;
1101
1102 /*
1103 * Real-time signals must be queued if sent by sigqueue, or
1104 * some other real-time mechanism. It is implementation
1105 * defined whether kill() does so. We attempt to do so, on
1106 * the principle of least surprise, but since kill is not
1107 * allowed to fail with EAGAIN when low on memory we just
1108 * make sure at least one signal gets delivered and don't
1109 * pass on the info struct.
1110 */
1111 if (sig < SIGRTMIN)
1112 override_rlimit = (is_si_special(info) || info->si_code >= 0);
1113 else
1114 override_rlimit = 0;
1115
1116 q = __sigqueue_alloc(sig, t, GFP_ATOMIC, override_rlimit, 0);
1117
1118 if (q) {
1119 list_add_tail(&q->list, &pending->list);
1120 switch ((unsigned long) info) {
1121 case (unsigned long) SEND_SIG_NOINFO:
1122 clear_siginfo(&q->info);
1123 q->info.si_signo = sig;
1124 q->info.si_errno = 0;
1125 q->info.si_code = SI_USER;
1126 q->info.si_pid = task_tgid_nr_ns(current,
1127 task_active_pid_ns(t));
1128 rcu_read_lock();
1129 q->info.si_uid =
1130 from_kuid_munged(task_cred_xxx(t, user_ns),
1131 current_uid());
1132 rcu_read_unlock();
1133 break;
1134 case (unsigned long) SEND_SIG_PRIV:
1135 clear_siginfo(&q->info);
1136 q->info.si_signo = sig;
1137 q->info.si_errno = 0;
1138 q->info.si_code = SI_KERNEL;
1139 q->info.si_pid = 0;
1140 q->info.si_uid = 0;
1141 break;
1142 default:
1143 copy_siginfo(&q->info, info);
1144 break;
1145 }
1146 } else if (!is_si_special(info) &&
1147 sig >= SIGRTMIN && info->si_code != SI_USER) {
1148 /*
1149 * Queue overflow, abort. We may abort if the
1150 * signal was rt and sent by user using something
1151 * other than kill().
1152 */
1153 result = TRACE_SIGNAL_OVERFLOW_FAIL;
1154 ret = -EAGAIN;
1155 goto ret;
1156 } else {
1157 /*
1158 * This is a silent loss of information. We still
1159 * send the signal, but the *info bits are lost.
1160 */
1161 result = TRACE_SIGNAL_LOSE_INFO;
1162 }
1163
1164 out_set:
1165 signalfd_notify(t, sig);
1166 sigaddset(&pending->signal, sig);
1167
1168 /* Let multiprocess signals appear after on-going forks */
1169 if (type > PIDTYPE_TGID) {
1170 struct multiprocess_signals *delayed;
1171 hlist_for_each_entry(delayed, &t->signal->multiprocess, node) {
1172 sigset_t *signal = &delayed->signal;
1173 /* Can't queue both a stop and a continue signal */
1174 if (sig == SIGCONT)
1175 sigdelsetmask(signal, SIG_KERNEL_STOP_MASK);
1176 else if (sig_kernel_stop(sig))
1177 sigdelset(signal, SIGCONT);
1178 sigaddset(signal, sig);
1179 }
1180 }
1181
1182 complete_signal(sig, t, type);
1183 ret:
1184 trace_signal_generate(sig, info, t, type != PIDTYPE_PID, result);
1185 return ret;
1186 }
1187
has_si_pid_and_uid(struct kernel_siginfo * info)1188 static inline bool has_si_pid_and_uid(struct kernel_siginfo *info)
1189 {
1190 bool ret = false;
1191 switch (siginfo_layout(info->si_signo, info->si_code)) {
1192 case SIL_KILL:
1193 case SIL_CHLD:
1194 case SIL_RT:
1195 ret = true;
1196 break;
1197 case SIL_TIMER:
1198 case SIL_POLL:
1199 case SIL_FAULT:
1200 case SIL_FAULT_TRAPNO:
1201 case SIL_FAULT_MCEERR:
1202 case SIL_FAULT_BNDERR:
1203 case SIL_FAULT_PKUERR:
1204 case SIL_FAULT_PERF_EVENT:
1205 case SIL_SYS:
1206 ret = false;
1207 break;
1208 }
1209 return ret;
1210 }
1211
send_signal(int sig,struct kernel_siginfo * info,struct task_struct * t,enum pid_type type)1212 static int send_signal(int sig, struct kernel_siginfo *info, struct task_struct *t,
1213 enum pid_type type)
1214 {
1215 /* Should SIGKILL or SIGSTOP be received by a pid namespace init? */
1216 bool force = false;
1217
1218 if (info == SEND_SIG_NOINFO) {
1219 /* Force if sent from an ancestor pid namespace */
1220 force = !task_pid_nr_ns(current, task_active_pid_ns(t));
1221 } else if (info == SEND_SIG_PRIV) {
1222 /* Don't ignore kernel generated signals */
1223 force = true;
1224 } else if (has_si_pid_and_uid(info)) {
1225 /* SIGKILL and SIGSTOP is special or has ids */
1226 struct user_namespace *t_user_ns;
1227
1228 rcu_read_lock();
1229 t_user_ns = task_cred_xxx(t, user_ns);
1230 if (current_user_ns() != t_user_ns) {
1231 kuid_t uid = make_kuid(current_user_ns(), info->si_uid);
1232 info->si_uid = from_kuid_munged(t_user_ns, uid);
1233 }
1234 rcu_read_unlock();
1235
1236 /* A kernel generated signal? */
1237 force = (info->si_code == SI_KERNEL);
1238
1239 /* From an ancestor pid namespace? */
1240 if (!task_pid_nr_ns(current, task_active_pid_ns(t))) {
1241 info->si_pid = 0;
1242 force = true;
1243 }
1244 }
1245 return __send_signal(sig, info, t, type, force);
1246 }
1247
print_fatal_signal(int signr)1248 static void print_fatal_signal(int signr)
1249 {
1250 struct pt_regs *regs = signal_pt_regs();
1251 pr_info("potentially unexpected fatal signal %d.\n", signr);
1252
1253 #if defined(__i386__) && !defined(__arch_um__)
1254 pr_info("code at %08lx: ", regs->ip);
1255 {
1256 int i;
1257 for (i = 0; i < 16; i++) {
1258 unsigned char insn;
1259
1260 if (get_user(insn, (unsigned char *)(regs->ip + i)))
1261 break;
1262 pr_cont("%02x ", insn);
1263 }
1264 }
1265 pr_cont("\n");
1266 #endif
1267 preempt_disable();
1268 show_regs(regs);
1269 preempt_enable();
1270 }
1271
setup_print_fatal_signals(char * str)1272 static int __init setup_print_fatal_signals(char *str)
1273 {
1274 get_option (&str, &print_fatal_signals);
1275
1276 return 1;
1277 }
1278
1279 __setup("print-fatal-signals=", setup_print_fatal_signals);
1280
1281 int
__group_send_sig_info(int sig,struct kernel_siginfo * info,struct task_struct * p)1282 __group_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1283 {
1284 return send_signal(sig, info, p, PIDTYPE_TGID);
1285 }
1286
do_send_sig_info(int sig,struct kernel_siginfo * info,struct task_struct * p,enum pid_type type)1287 int do_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p,
1288 enum pid_type type)
1289 {
1290 unsigned long flags;
1291 int ret = -ESRCH;
1292
1293 if (lock_task_sighand(p, &flags)) {
1294 ret = send_signal(sig, info, p, type);
1295 unlock_task_sighand(p, &flags);
1296 }
1297
1298 return ret;
1299 }
1300
1301 enum sig_handler {
1302 HANDLER_CURRENT, /* If reachable use the current handler */
1303 HANDLER_SIG_DFL, /* Always use SIG_DFL handler semantics */
1304 HANDLER_EXIT, /* Only visible as the process exit code */
1305 };
1306
1307 /*
1308 * Force a signal that the process can't ignore: if necessary
1309 * we unblock the signal and change any SIG_IGN to SIG_DFL.
1310 *
1311 * Note: If we unblock the signal, we always reset it to SIG_DFL,
1312 * since we do not want to have a signal handler that was blocked
1313 * be invoked when user space had explicitly blocked it.
1314 *
1315 * We don't want to have recursive SIGSEGV's etc, for example,
1316 * that is why we also clear SIGNAL_UNKILLABLE.
1317 */
1318 static int
force_sig_info_to_task(struct kernel_siginfo * info,struct task_struct * t,enum sig_handler handler)1319 force_sig_info_to_task(struct kernel_siginfo *info, struct task_struct *t,
1320 enum sig_handler handler)
1321 {
1322 unsigned long int flags;
1323 int ret, blocked, ignored;
1324 struct k_sigaction *action;
1325 int sig = info->si_signo;
1326
1327 spin_lock_irqsave(&t->sighand->siglock, flags);
1328 action = &t->sighand->action[sig-1];
1329 ignored = action->sa.sa_handler == SIG_IGN;
1330 blocked = sigismember(&t->blocked, sig);
1331 if (blocked || ignored || (handler != HANDLER_CURRENT)) {
1332 action->sa.sa_handler = SIG_DFL;
1333 if (handler == HANDLER_EXIT)
1334 action->sa.sa_flags |= SA_IMMUTABLE;
1335 if (blocked) {
1336 sigdelset(&t->blocked, sig);
1337 recalc_sigpending_and_wake(t);
1338 }
1339 }
1340 /*
1341 * Don't clear SIGNAL_UNKILLABLE for traced tasks, users won't expect
1342 * debugging to leave init killable.
1343 */
1344 if (action->sa.sa_handler == SIG_DFL && !t->ptrace)
1345 t->signal->flags &= ~SIGNAL_UNKILLABLE;
1346 ret = send_signal(sig, info, t, PIDTYPE_PID);
1347 spin_unlock_irqrestore(&t->sighand->siglock, flags);
1348
1349 return ret;
1350 }
1351
force_sig_info(struct kernel_siginfo * info)1352 int force_sig_info(struct kernel_siginfo *info)
1353 {
1354 return force_sig_info_to_task(info, current, HANDLER_CURRENT);
1355 }
1356
1357 /*
1358 * Nuke all other threads in the group.
1359 */
zap_other_threads(struct task_struct * p)1360 int zap_other_threads(struct task_struct *p)
1361 {
1362 struct task_struct *t = p;
1363 int count = 0;
1364
1365 p->signal->group_stop_count = 0;
1366
1367 while_each_thread(p, t) {
1368 task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1369 count++;
1370
1371 /* Don't bother with already dead threads */
1372 if (t->exit_state)
1373 continue;
1374 sigaddset(&t->pending.signal, SIGKILL);
1375 signal_wake_up(t, 1);
1376 }
1377
1378 return count;
1379 }
1380
__lock_task_sighand(struct task_struct * tsk,unsigned long * flags)1381 struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1382 unsigned long *flags)
1383 {
1384 struct sighand_struct *sighand;
1385
1386 rcu_read_lock();
1387 for (;;) {
1388 sighand = rcu_dereference(tsk->sighand);
1389 if (unlikely(sighand == NULL))
1390 break;
1391
1392 /*
1393 * This sighand can be already freed and even reused, but
1394 * we rely on SLAB_TYPESAFE_BY_RCU and sighand_ctor() which
1395 * initializes ->siglock: this slab can't go away, it has
1396 * the same object type, ->siglock can't be reinitialized.
1397 *
1398 * We need to ensure that tsk->sighand is still the same
1399 * after we take the lock, we can race with de_thread() or
1400 * __exit_signal(). In the latter case the next iteration
1401 * must see ->sighand == NULL.
1402 */
1403 spin_lock_irqsave(&sighand->siglock, *flags);
1404 if (likely(sighand == rcu_access_pointer(tsk->sighand)))
1405 break;
1406 spin_unlock_irqrestore(&sighand->siglock, *flags);
1407 }
1408 rcu_read_unlock();
1409
1410 return sighand;
1411 }
1412
1413 #ifdef CONFIG_LOCKDEP
lockdep_assert_task_sighand_held(struct task_struct * task)1414 void lockdep_assert_task_sighand_held(struct task_struct *task)
1415 {
1416 struct sighand_struct *sighand;
1417
1418 rcu_read_lock();
1419 sighand = rcu_dereference(task->sighand);
1420 if (sighand)
1421 lockdep_assert_held(&sighand->siglock);
1422 else
1423 WARN_ON_ONCE(1);
1424 rcu_read_unlock();
1425 }
1426 #endif
1427
1428 /*
1429 * send signal info to all the members of a group
1430 */
group_send_sig_info(int sig,struct kernel_siginfo * info,struct task_struct * p,enum pid_type type)1431 int group_send_sig_info(int sig, struct kernel_siginfo *info,
1432 struct task_struct *p, enum pid_type type)
1433 {
1434 int ret;
1435
1436 rcu_read_lock();
1437 ret = check_kill_permission(sig, info, p);
1438 rcu_read_unlock();
1439
1440 if (!ret && sig)
1441 ret = do_send_sig_info(sig, info, p, type);
1442
1443 return ret;
1444 }
1445
1446 /*
1447 * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1448 * control characters do (^C, ^Z etc)
1449 * - the caller must hold at least a readlock on tasklist_lock
1450 */
__kill_pgrp_info(int sig,struct kernel_siginfo * info,struct pid * pgrp)1451 int __kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1452 {
1453 struct task_struct *p = NULL;
1454 int retval, success;
1455
1456 success = 0;
1457 retval = -ESRCH;
1458 do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1459 int err = group_send_sig_info(sig, info, p, PIDTYPE_PGID);
1460 success |= !err;
1461 retval = err;
1462 } while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1463 return success ? 0 : retval;
1464 }
1465
kill_pid_info(int sig,struct kernel_siginfo * info,struct pid * pid)1466 int kill_pid_info(int sig, struct kernel_siginfo *info, struct pid *pid)
1467 {
1468 int error = -ESRCH;
1469 struct task_struct *p;
1470
1471 for (;;) {
1472 rcu_read_lock();
1473 p = pid_task(pid, PIDTYPE_PID);
1474 if (p)
1475 error = group_send_sig_info(sig, info, p, PIDTYPE_TGID);
1476 rcu_read_unlock();
1477 if (likely(!p || error != -ESRCH))
1478 return error;
1479
1480 /*
1481 * The task was unhashed in between, try again. If it
1482 * is dead, pid_task() will return NULL, if we race with
1483 * de_thread() it will find the new leader.
1484 */
1485 }
1486 }
1487
kill_proc_info(int sig,struct kernel_siginfo * info,pid_t pid)1488 static int kill_proc_info(int sig, struct kernel_siginfo *info, pid_t pid)
1489 {
1490 int error;
1491 rcu_read_lock();
1492 error = kill_pid_info(sig, info, find_vpid(pid));
1493 rcu_read_unlock();
1494 return error;
1495 }
1496
kill_as_cred_perm(const struct cred * cred,struct task_struct * target)1497 static inline bool kill_as_cred_perm(const struct cred *cred,
1498 struct task_struct *target)
1499 {
1500 const struct cred *pcred = __task_cred(target);
1501
1502 return uid_eq(cred->euid, pcred->suid) ||
1503 uid_eq(cred->euid, pcred->uid) ||
1504 uid_eq(cred->uid, pcred->suid) ||
1505 uid_eq(cred->uid, pcred->uid);
1506 }
1507
1508 /*
1509 * The usb asyncio usage of siginfo is wrong. The glibc support
1510 * for asyncio which uses SI_ASYNCIO assumes the layout is SIL_RT.
1511 * AKA after the generic fields:
1512 * kernel_pid_t si_pid;
1513 * kernel_uid32_t si_uid;
1514 * sigval_t si_value;
1515 *
1516 * Unfortunately when usb generates SI_ASYNCIO it assumes the layout
1517 * after the generic fields is:
1518 * void __user *si_addr;
1519 *
1520 * This is a practical problem when there is a 64bit big endian kernel
1521 * and a 32bit userspace. As the 32bit address will encoded in the low
1522 * 32bits of the pointer. Those low 32bits will be stored at higher
1523 * address than appear in a 32 bit pointer. So userspace will not
1524 * see the address it was expecting for it's completions.
1525 *
1526 * There is nothing in the encoding that can allow
1527 * copy_siginfo_to_user32 to detect this confusion of formats, so
1528 * handle this by requiring the caller of kill_pid_usb_asyncio to
1529 * notice when this situration takes place and to store the 32bit
1530 * pointer in sival_int, instead of sival_addr of the sigval_t addr
1531 * parameter.
1532 */
kill_pid_usb_asyncio(int sig,int errno,sigval_t addr,struct pid * pid,const struct cred * cred)1533 int kill_pid_usb_asyncio(int sig, int errno, sigval_t addr,
1534 struct pid *pid, const struct cred *cred)
1535 {
1536 struct kernel_siginfo info;
1537 struct task_struct *p;
1538 unsigned long flags;
1539 int ret = -EINVAL;
1540
1541 if (!valid_signal(sig))
1542 return ret;
1543
1544 clear_siginfo(&info);
1545 info.si_signo = sig;
1546 info.si_errno = errno;
1547 info.si_code = SI_ASYNCIO;
1548 *((sigval_t *)&info.si_pid) = addr;
1549
1550 rcu_read_lock();
1551 p = pid_task(pid, PIDTYPE_PID);
1552 if (!p) {
1553 ret = -ESRCH;
1554 goto out_unlock;
1555 }
1556 if (!kill_as_cred_perm(cred, p)) {
1557 ret = -EPERM;
1558 goto out_unlock;
1559 }
1560 ret = security_task_kill(p, &info, sig, cred);
1561 if (ret)
1562 goto out_unlock;
1563
1564 if (sig) {
1565 if (lock_task_sighand(p, &flags)) {
1566 ret = __send_signal(sig, &info, p, PIDTYPE_TGID, false);
1567 unlock_task_sighand(p, &flags);
1568 } else
1569 ret = -ESRCH;
1570 }
1571 out_unlock:
1572 rcu_read_unlock();
1573 return ret;
1574 }
1575 EXPORT_SYMBOL_GPL(kill_pid_usb_asyncio);
1576
1577 /*
1578 * kill_something_info() interprets pid in interesting ways just like kill(2).
1579 *
1580 * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1581 * is probably wrong. Should make it like BSD or SYSV.
1582 */
1583
kill_something_info(int sig,struct kernel_siginfo * info,pid_t pid)1584 static int kill_something_info(int sig, struct kernel_siginfo *info, pid_t pid)
1585 {
1586 int ret;
1587
1588 if (pid > 0)
1589 return kill_proc_info(sig, info, pid);
1590
1591 /* -INT_MIN is undefined. Exclude this case to avoid a UBSAN warning */
1592 if (pid == INT_MIN)
1593 return -ESRCH;
1594
1595 read_lock(&tasklist_lock);
1596 if (pid != -1) {
1597 ret = __kill_pgrp_info(sig, info,
1598 pid ? find_vpid(-pid) : task_pgrp(current));
1599 } else {
1600 int retval = 0, count = 0;
1601 struct task_struct * p;
1602
1603 for_each_process(p) {
1604 if (task_pid_vnr(p) > 1 &&
1605 !same_thread_group(p, current)) {
1606 int err = group_send_sig_info(sig, info, p,
1607 PIDTYPE_MAX);
1608 ++count;
1609 if (err != -EPERM)
1610 retval = err;
1611 }
1612 }
1613 ret = count ? retval : -ESRCH;
1614 }
1615 read_unlock(&tasklist_lock);
1616
1617 return ret;
1618 }
1619
1620 /*
1621 * These are for backward compatibility with the rest of the kernel source.
1622 */
1623
send_sig_info(int sig,struct kernel_siginfo * info,struct task_struct * p)1624 int send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1625 {
1626 /*
1627 * Make sure legacy kernel users don't send in bad values
1628 * (normal paths check this in check_kill_permission).
1629 */
1630 if (!valid_signal(sig))
1631 return -EINVAL;
1632
1633 return do_send_sig_info(sig, info, p, PIDTYPE_PID);
1634 }
1635 EXPORT_SYMBOL(send_sig_info);
1636
1637 #define __si_special(priv) \
1638 ((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1639
1640 int
send_sig(int sig,struct task_struct * p,int priv)1641 send_sig(int sig, struct task_struct *p, int priv)
1642 {
1643 return send_sig_info(sig, __si_special(priv), p);
1644 }
1645 EXPORT_SYMBOL(send_sig);
1646
force_sig(int sig)1647 void force_sig(int sig)
1648 {
1649 struct kernel_siginfo info;
1650
1651 clear_siginfo(&info);
1652 info.si_signo = sig;
1653 info.si_errno = 0;
1654 info.si_code = SI_KERNEL;
1655 info.si_pid = 0;
1656 info.si_uid = 0;
1657 force_sig_info(&info);
1658 }
1659 EXPORT_SYMBOL(force_sig);
1660
force_fatal_sig(int sig)1661 void force_fatal_sig(int sig)
1662 {
1663 struct kernel_siginfo info;
1664
1665 clear_siginfo(&info);
1666 info.si_signo = sig;
1667 info.si_errno = 0;
1668 info.si_code = SI_KERNEL;
1669 info.si_pid = 0;
1670 info.si_uid = 0;
1671 force_sig_info_to_task(&info, current, HANDLER_SIG_DFL);
1672 }
1673
force_exit_sig(int sig)1674 void force_exit_sig(int sig)
1675 {
1676 struct kernel_siginfo info;
1677
1678 clear_siginfo(&info);
1679 info.si_signo = sig;
1680 info.si_errno = 0;
1681 info.si_code = SI_KERNEL;
1682 info.si_pid = 0;
1683 info.si_uid = 0;
1684 force_sig_info_to_task(&info, current, HANDLER_EXIT);
1685 }
1686
1687 /*
1688 * When things go south during signal handling, we
1689 * will force a SIGSEGV. And if the signal that caused
1690 * the problem was already a SIGSEGV, we'll want to
1691 * make sure we don't even try to deliver the signal..
1692 */
force_sigsegv(int sig)1693 void force_sigsegv(int sig)
1694 {
1695 if (sig == SIGSEGV)
1696 force_fatal_sig(SIGSEGV);
1697 else
1698 force_sig(SIGSEGV);
1699 }
1700
force_sig_fault_to_task(int sig,int code,void __user * addr ___ARCH_SI_IA64 (int imm,unsigned int flags,unsigned long isr),struct task_struct * t)1701 int force_sig_fault_to_task(int sig, int code, void __user *addr
1702 ___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1703 , struct task_struct *t)
1704 {
1705 struct kernel_siginfo info;
1706
1707 clear_siginfo(&info);
1708 info.si_signo = sig;
1709 info.si_errno = 0;
1710 info.si_code = code;
1711 info.si_addr = addr;
1712 #ifdef __ia64__
1713 info.si_imm = imm;
1714 info.si_flags = flags;
1715 info.si_isr = isr;
1716 #endif
1717 return force_sig_info_to_task(&info, t, HANDLER_CURRENT);
1718 }
1719
force_sig_fault(int sig,int code,void __user * addr ___ARCH_SI_IA64 (int imm,unsigned int flags,unsigned long isr))1720 int force_sig_fault(int sig, int code, void __user *addr
1721 ___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr))
1722 {
1723 return force_sig_fault_to_task(sig, code, addr
1724 ___ARCH_SI_IA64(imm, flags, isr), current);
1725 }
1726
send_sig_fault(int sig,int code,void __user * addr ___ARCH_SI_IA64 (int imm,unsigned int flags,unsigned long isr),struct task_struct * t)1727 int send_sig_fault(int sig, int code, void __user *addr
1728 ___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1729 , struct task_struct *t)
1730 {
1731 struct kernel_siginfo info;
1732
1733 clear_siginfo(&info);
1734 info.si_signo = sig;
1735 info.si_errno = 0;
1736 info.si_code = code;
1737 info.si_addr = addr;
1738 #ifdef __ia64__
1739 info.si_imm = imm;
1740 info.si_flags = flags;
1741 info.si_isr = isr;
1742 #endif
1743 return send_sig_info(info.si_signo, &info, t);
1744 }
1745
force_sig_mceerr(int code,void __user * addr,short lsb)1746 int force_sig_mceerr(int code, void __user *addr, short lsb)
1747 {
1748 struct kernel_siginfo info;
1749
1750 WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1751 clear_siginfo(&info);
1752 info.si_signo = SIGBUS;
1753 info.si_errno = 0;
1754 info.si_code = code;
1755 info.si_addr = addr;
1756 info.si_addr_lsb = lsb;
1757 return force_sig_info(&info);
1758 }
1759
send_sig_mceerr(int code,void __user * addr,short lsb,struct task_struct * t)1760 int send_sig_mceerr(int code, void __user *addr, short lsb, struct task_struct *t)
1761 {
1762 struct kernel_siginfo info;
1763
1764 WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1765 clear_siginfo(&info);
1766 info.si_signo = SIGBUS;
1767 info.si_errno = 0;
1768 info.si_code = code;
1769 info.si_addr = addr;
1770 info.si_addr_lsb = lsb;
1771 return send_sig_info(info.si_signo, &info, t);
1772 }
1773 EXPORT_SYMBOL(send_sig_mceerr);
1774
force_sig_bnderr(void __user * addr,void __user * lower,void __user * upper)1775 int force_sig_bnderr(void __user *addr, void __user *lower, void __user *upper)
1776 {
1777 struct kernel_siginfo info;
1778
1779 clear_siginfo(&info);
1780 info.si_signo = SIGSEGV;
1781 info.si_errno = 0;
1782 info.si_code = SEGV_BNDERR;
1783 info.si_addr = addr;
1784 info.si_lower = lower;
1785 info.si_upper = upper;
1786 return force_sig_info(&info);
1787 }
1788
1789 #ifdef SEGV_PKUERR
force_sig_pkuerr(void __user * addr,u32 pkey)1790 int force_sig_pkuerr(void __user *addr, u32 pkey)
1791 {
1792 struct kernel_siginfo info;
1793
1794 clear_siginfo(&info);
1795 info.si_signo = SIGSEGV;
1796 info.si_errno = 0;
1797 info.si_code = SEGV_PKUERR;
1798 info.si_addr = addr;
1799 info.si_pkey = pkey;
1800 return force_sig_info(&info);
1801 }
1802 #endif
1803
force_sig_perf(void __user * addr,u32 type,u64 sig_data)1804 int force_sig_perf(void __user *addr, u32 type, u64 sig_data)
1805 {
1806 struct kernel_siginfo info;
1807
1808 clear_siginfo(&info);
1809 info.si_signo = SIGTRAP;
1810 info.si_errno = 0;
1811 info.si_code = TRAP_PERF;
1812 info.si_addr = addr;
1813 info.si_perf_data = sig_data;
1814 info.si_perf_type = type;
1815
1816 return force_sig_info(&info);
1817 }
1818
1819 /**
1820 * force_sig_seccomp - signals the task to allow in-process syscall emulation
1821 * @syscall: syscall number to send to userland
1822 * @reason: filter-supplied reason code to send to userland (via si_errno)
1823 *
1824 * Forces a SIGSYS with a code of SYS_SECCOMP and related sigsys info.
1825 */
force_sig_seccomp(int syscall,int reason,bool force_coredump)1826 int force_sig_seccomp(int syscall, int reason, bool force_coredump)
1827 {
1828 struct kernel_siginfo info;
1829
1830 clear_siginfo(&info);
1831 info.si_signo = SIGSYS;
1832 info.si_code = SYS_SECCOMP;
1833 info.si_call_addr = (void __user *)KSTK_EIP(current);
1834 info.si_errno = reason;
1835 info.si_arch = syscall_get_arch(current);
1836 info.si_syscall = syscall;
1837 return force_sig_info_to_task(&info, current,
1838 force_coredump ? HANDLER_EXIT : HANDLER_CURRENT);
1839 }
1840
1841 /* For the crazy architectures that include trap information in
1842 * the errno field, instead of an actual errno value.
1843 */
force_sig_ptrace_errno_trap(int errno,void __user * addr)1844 int force_sig_ptrace_errno_trap(int errno, void __user *addr)
1845 {
1846 struct kernel_siginfo info;
1847
1848 clear_siginfo(&info);
1849 info.si_signo = SIGTRAP;
1850 info.si_errno = errno;
1851 info.si_code = TRAP_HWBKPT;
1852 info.si_addr = addr;
1853 return force_sig_info(&info);
1854 }
1855
1856 /* For the rare architectures that include trap information using
1857 * si_trapno.
1858 */
force_sig_fault_trapno(int sig,int code,void __user * addr,int trapno)1859 int force_sig_fault_trapno(int sig, int code, void __user *addr, int trapno)
1860 {
1861 struct kernel_siginfo info;
1862
1863 clear_siginfo(&info);
1864 info.si_signo = sig;
1865 info.si_errno = 0;
1866 info.si_code = code;
1867 info.si_addr = addr;
1868 info.si_trapno = trapno;
1869 return force_sig_info(&info);
1870 }
1871
1872 /* For the rare architectures that include trap information using
1873 * si_trapno.
1874 */
send_sig_fault_trapno(int sig,int code,void __user * addr,int trapno,struct task_struct * t)1875 int send_sig_fault_trapno(int sig, int code, void __user *addr, int trapno,
1876 struct task_struct *t)
1877 {
1878 struct kernel_siginfo info;
1879
1880 clear_siginfo(&info);
1881 info.si_signo = sig;
1882 info.si_errno = 0;
1883 info.si_code = code;
1884 info.si_addr = addr;
1885 info.si_trapno = trapno;
1886 return send_sig_info(info.si_signo, &info, t);
1887 }
1888
kill_pgrp(struct pid * pid,int sig,int priv)1889 int kill_pgrp(struct pid *pid, int sig, int priv)
1890 {
1891 int ret;
1892
1893 read_lock(&tasklist_lock);
1894 ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1895 read_unlock(&tasklist_lock);
1896
1897 return ret;
1898 }
1899 EXPORT_SYMBOL(kill_pgrp);
1900
kill_pid(struct pid * pid,int sig,int priv)1901 int kill_pid(struct pid *pid, int sig, int priv)
1902 {
1903 return kill_pid_info(sig, __si_special(priv), pid);
1904 }
1905 EXPORT_SYMBOL(kill_pid);
1906
1907 /*
1908 * These functions support sending signals using preallocated sigqueue
1909 * structures. This is needed "because realtime applications cannot
1910 * afford to lose notifications of asynchronous events, like timer
1911 * expirations or I/O completions". In the case of POSIX Timers
1912 * we allocate the sigqueue structure from the timer_create. If this
1913 * allocation fails we are able to report the failure to the application
1914 * with an EAGAIN error.
1915 */
sigqueue_alloc(void)1916 struct sigqueue *sigqueue_alloc(void)
1917 {
1918 return __sigqueue_alloc(-1, current, GFP_KERNEL, 0, SIGQUEUE_PREALLOC);
1919 }
1920
sigqueue_free(struct sigqueue * q)1921 void sigqueue_free(struct sigqueue *q)
1922 {
1923 unsigned long flags;
1924 spinlock_t *lock = ¤t->sighand->siglock;
1925
1926 BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1927 /*
1928 * We must hold ->siglock while testing q->list
1929 * to serialize with collect_signal() or with
1930 * __exit_signal()->flush_sigqueue().
1931 */
1932 spin_lock_irqsave(lock, flags);
1933 q->flags &= ~SIGQUEUE_PREALLOC;
1934 /*
1935 * If it is queued it will be freed when dequeued,
1936 * like the "regular" sigqueue.
1937 */
1938 if (!list_empty(&q->list))
1939 q = NULL;
1940 spin_unlock_irqrestore(lock, flags);
1941
1942 if (q)
1943 __sigqueue_free(q);
1944 }
1945
send_sigqueue(struct sigqueue * q,struct pid * pid,enum pid_type type)1946 int send_sigqueue(struct sigqueue *q, struct pid *pid, enum pid_type type)
1947 {
1948 int sig = q->info.si_signo;
1949 struct sigpending *pending;
1950 struct task_struct *t;
1951 unsigned long flags;
1952 int ret, result;
1953
1954 BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1955
1956 ret = -1;
1957 rcu_read_lock();
1958 t = pid_task(pid, type);
1959 if (!t || !likely(lock_task_sighand(t, &flags)))
1960 goto ret;
1961
1962 ret = 1; /* the signal is ignored */
1963 result = TRACE_SIGNAL_IGNORED;
1964 if (!prepare_signal(sig, t, false))
1965 goto out;
1966
1967 ret = 0;
1968 if (unlikely(!list_empty(&q->list))) {
1969 /*
1970 * If an SI_TIMER entry is already queue just increment
1971 * the overrun count.
1972 */
1973 BUG_ON(q->info.si_code != SI_TIMER);
1974 q->info.si_overrun++;
1975 result = TRACE_SIGNAL_ALREADY_PENDING;
1976 goto out;
1977 }
1978 q->info.si_overrun = 0;
1979
1980 signalfd_notify(t, sig);
1981 pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1982 list_add_tail(&q->list, &pending->list);
1983 sigaddset(&pending->signal, sig);
1984 complete_signal(sig, t, type);
1985 result = TRACE_SIGNAL_DELIVERED;
1986 out:
1987 trace_signal_generate(sig, &q->info, t, type != PIDTYPE_PID, result);
1988 unlock_task_sighand(t, &flags);
1989 ret:
1990 rcu_read_unlock();
1991 return ret;
1992 }
1993
do_notify_pidfd(struct task_struct * task)1994 static void do_notify_pidfd(struct task_struct *task)
1995 {
1996 struct pid *pid;
1997
1998 WARN_ON(task->exit_state == 0);
1999 pid = task_pid(task);
2000 wake_up_all(&pid->wait_pidfd);
2001 }
2002
2003 /*
2004 * Let a parent know about the death of a child.
2005 * For a stopped/continued status change, use do_notify_parent_cldstop instead.
2006 *
2007 * Returns true if our parent ignored us and so we've switched to
2008 * self-reaping.
2009 */
do_notify_parent(struct task_struct * tsk,int sig)2010 bool do_notify_parent(struct task_struct *tsk, int sig)
2011 {
2012 struct kernel_siginfo info;
2013 unsigned long flags;
2014 struct sighand_struct *psig;
2015 bool autoreap = false;
2016 u64 utime, stime;
2017
2018 BUG_ON(sig == -1);
2019
2020 /* do_notify_parent_cldstop should have been called instead. */
2021 BUG_ON(task_is_stopped_or_traced(tsk));
2022
2023 BUG_ON(!tsk->ptrace &&
2024 (tsk->group_leader != tsk || !thread_group_empty(tsk)));
2025
2026 /* Wake up all pidfd waiters */
2027 do_notify_pidfd(tsk);
2028
2029 if (sig != SIGCHLD) {
2030 /*
2031 * This is only possible if parent == real_parent.
2032 * Check if it has changed security domain.
2033 */
2034 if (tsk->parent_exec_id != READ_ONCE(tsk->parent->self_exec_id))
2035 sig = SIGCHLD;
2036 }
2037
2038 clear_siginfo(&info);
2039 info.si_signo = sig;
2040 info.si_errno = 0;
2041 /*
2042 * We are under tasklist_lock here so our parent is tied to
2043 * us and cannot change.
2044 *
2045 * task_active_pid_ns will always return the same pid namespace
2046 * until a task passes through release_task.
2047 *
2048 * write_lock() currently calls preempt_disable() which is the
2049 * same as rcu_read_lock(), but according to Oleg, this is not
2050 * correct to rely on this
2051 */
2052 rcu_read_lock();
2053 info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
2054 info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
2055 task_uid(tsk));
2056 rcu_read_unlock();
2057
2058 task_cputime(tsk, &utime, &stime);
2059 info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
2060 info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
2061
2062 info.si_status = tsk->exit_code & 0x7f;
2063 if (tsk->exit_code & 0x80)
2064 info.si_code = CLD_DUMPED;
2065 else if (tsk->exit_code & 0x7f)
2066 info.si_code = CLD_KILLED;
2067 else {
2068 info.si_code = CLD_EXITED;
2069 info.si_status = tsk->exit_code >> 8;
2070 }
2071
2072 psig = tsk->parent->sighand;
2073 spin_lock_irqsave(&psig->siglock, flags);
2074 if (!tsk->ptrace && sig == SIGCHLD &&
2075 (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
2076 (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
2077 /*
2078 * We are exiting and our parent doesn't care. POSIX.1
2079 * defines special semantics for setting SIGCHLD to SIG_IGN
2080 * or setting the SA_NOCLDWAIT flag: we should be reaped
2081 * automatically and not left for our parent's wait4 call.
2082 * Rather than having the parent do it as a magic kind of
2083 * signal handler, we just set this to tell do_exit that we
2084 * can be cleaned up without becoming a zombie. Note that
2085 * we still call __wake_up_parent in this case, because a
2086 * blocked sys_wait4 might now return -ECHILD.
2087 *
2088 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
2089 * is implementation-defined: we do (if you don't want
2090 * it, just use SIG_IGN instead).
2091 */
2092 autoreap = true;
2093 if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
2094 sig = 0;
2095 }
2096 /*
2097 * Send with __send_signal as si_pid and si_uid are in the
2098 * parent's namespaces.
2099 */
2100 if (valid_signal(sig) && sig)
2101 __send_signal(sig, &info, tsk->parent, PIDTYPE_TGID, false);
2102 __wake_up_parent(tsk, tsk->parent);
2103 spin_unlock_irqrestore(&psig->siglock, flags);
2104
2105 return autoreap;
2106 }
2107
2108 /**
2109 * do_notify_parent_cldstop - notify parent of stopped/continued state change
2110 * @tsk: task reporting the state change
2111 * @for_ptracer: the notification is for ptracer
2112 * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
2113 *
2114 * Notify @tsk's parent that the stopped/continued state has changed. If
2115 * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
2116 * If %true, @tsk reports to @tsk->parent which should be the ptracer.
2117 *
2118 * CONTEXT:
2119 * Must be called with tasklist_lock at least read locked.
2120 */
do_notify_parent_cldstop(struct task_struct * tsk,bool for_ptracer,int why)2121 static void do_notify_parent_cldstop(struct task_struct *tsk,
2122 bool for_ptracer, int why)
2123 {
2124 struct kernel_siginfo info;
2125 unsigned long flags;
2126 struct task_struct *parent;
2127 struct sighand_struct *sighand;
2128 u64 utime, stime;
2129
2130 if (for_ptracer) {
2131 parent = tsk->parent;
2132 } else {
2133 tsk = tsk->group_leader;
2134 parent = tsk->real_parent;
2135 }
2136
2137 clear_siginfo(&info);
2138 info.si_signo = SIGCHLD;
2139 info.si_errno = 0;
2140 /*
2141 * see comment in do_notify_parent() about the following 4 lines
2142 */
2143 rcu_read_lock();
2144 info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
2145 info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
2146 rcu_read_unlock();
2147
2148 task_cputime(tsk, &utime, &stime);
2149 info.si_utime = nsec_to_clock_t(utime);
2150 info.si_stime = nsec_to_clock_t(stime);
2151
2152 info.si_code = why;
2153 switch (why) {
2154 case CLD_CONTINUED:
2155 info.si_status = SIGCONT;
2156 break;
2157 case CLD_STOPPED:
2158 info.si_status = tsk->signal->group_exit_code & 0x7f;
2159 break;
2160 case CLD_TRAPPED:
2161 info.si_status = tsk->exit_code & 0x7f;
2162 break;
2163 default:
2164 BUG();
2165 }
2166
2167 sighand = parent->sighand;
2168 spin_lock_irqsave(&sighand->siglock, flags);
2169 if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
2170 !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
2171 __group_send_sig_info(SIGCHLD, &info, parent);
2172 /*
2173 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
2174 */
2175 __wake_up_parent(tsk, parent);
2176 spin_unlock_irqrestore(&sighand->siglock, flags);
2177 }
2178
2179 /*
2180 * This must be called with current->sighand->siglock held.
2181 *
2182 * This should be the path for all ptrace stops.
2183 * We always set current->last_siginfo while stopped here.
2184 * That makes it a way to test a stopped process for
2185 * being ptrace-stopped vs being job-control-stopped.
2186 *
2187 * If we actually decide not to stop at all because the tracer
2188 * is gone, we keep current->exit_code unless clear_code.
2189 */
ptrace_stop(int exit_code,int why,int clear_code,kernel_siginfo_t * info)2190 static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t *info)
2191 __releases(¤t->sighand->siglock)
2192 __acquires(¤t->sighand->siglock)
2193 {
2194 bool gstop_done = false;
2195
2196 if (arch_ptrace_stop_needed()) {
2197 /*
2198 * The arch code has something special to do before a
2199 * ptrace stop. This is allowed to block, e.g. for faults
2200 * on user stack pages. We can't keep the siglock while
2201 * calling arch_ptrace_stop, so we must release it now.
2202 * To preserve proper semantics, we must do this before
2203 * any signal bookkeeping like checking group_stop_count.
2204 */
2205 spin_unlock_irq(¤t->sighand->siglock);
2206 arch_ptrace_stop();
2207 spin_lock_irq(¤t->sighand->siglock);
2208 }
2209
2210 /*
2211 * schedule() will not sleep if there is a pending signal that
2212 * can awaken the task.
2213 */
2214 set_special_state(TASK_TRACED);
2215
2216 /*
2217 * We're committing to trapping. TRACED should be visible before
2218 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
2219 * Also, transition to TRACED and updates to ->jobctl should be
2220 * atomic with respect to siglock and should be done after the arch
2221 * hook as siglock is released and regrabbed across it.
2222 *
2223 * TRACER TRACEE
2224 *
2225 * ptrace_attach()
2226 * [L] wait_on_bit(JOBCTL_TRAPPING) [S] set_special_state(TRACED)
2227 * do_wait()
2228 * set_current_state() smp_wmb();
2229 * ptrace_do_wait()
2230 * wait_task_stopped()
2231 * task_stopped_code()
2232 * [L] task_is_traced() [S] task_clear_jobctl_trapping();
2233 */
2234 smp_wmb();
2235
2236 current->last_siginfo = info;
2237 current->exit_code = exit_code;
2238
2239 /*
2240 * If @why is CLD_STOPPED, we're trapping to participate in a group
2241 * stop. Do the bookkeeping. Note that if SIGCONT was delievered
2242 * across siglock relocks since INTERRUPT was scheduled, PENDING
2243 * could be clear now. We act as if SIGCONT is received after
2244 * TASK_TRACED is entered - ignore it.
2245 */
2246 if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
2247 gstop_done = task_participate_group_stop(current);
2248
2249 /* any trap clears pending STOP trap, STOP trap clears NOTIFY */
2250 task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
2251 if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
2252 task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
2253
2254 /* entering a trap, clear TRAPPING */
2255 task_clear_jobctl_trapping(current);
2256
2257 spin_unlock_irq(¤t->sighand->siglock);
2258 read_lock(&tasklist_lock);
2259 if (likely(current->ptrace)) {
2260 /*
2261 * Notify parents of the stop.
2262 *
2263 * While ptraced, there are two parents - the ptracer and
2264 * the real_parent of the group_leader. The ptracer should
2265 * know about every stop while the real parent is only
2266 * interested in the completion of group stop. The states
2267 * for the two don't interact with each other. Notify
2268 * separately unless they're gonna be duplicates.
2269 */
2270 do_notify_parent_cldstop(current, true, why);
2271 if (gstop_done && ptrace_reparented(current))
2272 do_notify_parent_cldstop(current, false, why);
2273
2274 /*
2275 * Don't want to allow preemption here, because
2276 * sys_ptrace() needs this task to be inactive.
2277 *
2278 * XXX: implement read_unlock_no_resched().
2279 */
2280 preempt_disable();
2281 read_unlock(&tasklist_lock);
2282 cgroup_enter_frozen();
2283 preempt_enable_no_resched();
2284 freezable_schedule();
2285 cgroup_leave_frozen(true);
2286 } else {
2287 /*
2288 * By the time we got the lock, our tracer went away.
2289 * Don't drop the lock yet, another tracer may come.
2290 *
2291 * If @gstop_done, the ptracer went away between group stop
2292 * completion and here. During detach, it would have set
2293 * JOBCTL_STOP_PENDING on us and we'll re-enter
2294 * TASK_STOPPED in do_signal_stop() on return, so notifying
2295 * the real parent of the group stop completion is enough.
2296 */
2297 if (gstop_done)
2298 do_notify_parent_cldstop(current, false, why);
2299
2300 /* tasklist protects us from ptrace_freeze_traced() */
2301 __set_current_state(TASK_RUNNING);
2302 if (clear_code)
2303 current->exit_code = 0;
2304 read_unlock(&tasklist_lock);
2305 }
2306
2307 /*
2308 * We are back. Now reacquire the siglock before touching
2309 * last_siginfo, so that we are sure to have synchronized with
2310 * any signal-sending on another CPU that wants to examine it.
2311 */
2312 spin_lock_irq(¤t->sighand->siglock);
2313 current->last_siginfo = NULL;
2314
2315 /* LISTENING can be set only during STOP traps, clear it */
2316 current->jobctl &= ~JOBCTL_LISTENING;
2317
2318 /*
2319 * Queued signals ignored us while we were stopped for tracing.
2320 * So check for any that we should take before resuming user mode.
2321 * This sets TIF_SIGPENDING, but never clears it.
2322 */
2323 recalc_sigpending_tsk(current);
2324 }
2325
ptrace_do_notify(int signr,int exit_code,int why)2326 static void ptrace_do_notify(int signr, int exit_code, int why)
2327 {
2328 kernel_siginfo_t info;
2329
2330 clear_siginfo(&info);
2331 info.si_signo = signr;
2332 info.si_code = exit_code;
2333 info.si_pid = task_pid_vnr(current);
2334 info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2335
2336 /* Let the debugger run. */
2337 ptrace_stop(exit_code, why, 1, &info);
2338 }
2339
ptrace_notify(int exit_code)2340 void ptrace_notify(int exit_code)
2341 {
2342 BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
2343 if (unlikely(current->task_works))
2344 task_work_run();
2345
2346 spin_lock_irq(¤t->sighand->siglock);
2347 ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
2348 spin_unlock_irq(¤t->sighand->siglock);
2349 }
2350
2351 /**
2352 * do_signal_stop - handle group stop for SIGSTOP and other stop signals
2353 * @signr: signr causing group stop if initiating
2354 *
2355 * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
2356 * and participate in it. If already set, participate in the existing
2357 * group stop. If participated in a group stop (and thus slept), %true is
2358 * returned with siglock released.
2359 *
2360 * If ptraced, this function doesn't handle stop itself. Instead,
2361 * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
2362 * untouched. The caller must ensure that INTERRUPT trap handling takes
2363 * places afterwards.
2364 *
2365 * CONTEXT:
2366 * Must be called with @current->sighand->siglock held, which is released
2367 * on %true return.
2368 *
2369 * RETURNS:
2370 * %false if group stop is already cancelled or ptrace trap is scheduled.
2371 * %true if participated in group stop.
2372 */
do_signal_stop(int signr)2373 static bool do_signal_stop(int signr)
2374 __releases(¤t->sighand->siglock)
2375 {
2376 struct signal_struct *sig = current->signal;
2377
2378 if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2379 unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2380 struct task_struct *t;
2381
2382 /* signr will be recorded in task->jobctl for retries */
2383 WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2384
2385 if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2386 unlikely(signal_group_exit(sig)))
2387 return false;
2388 /*
2389 * There is no group stop already in progress. We must
2390 * initiate one now.
2391 *
2392 * While ptraced, a task may be resumed while group stop is
2393 * still in effect and then receive a stop signal and
2394 * initiate another group stop. This deviates from the
2395 * usual behavior as two consecutive stop signals can't
2396 * cause two group stops when !ptraced. That is why we
2397 * also check !task_is_stopped(t) below.
2398 *
2399 * The condition can be distinguished by testing whether
2400 * SIGNAL_STOP_STOPPED is already set. Don't generate
2401 * group_exit_code in such case.
2402 *
2403 * This is not necessary for SIGNAL_STOP_CONTINUED because
2404 * an intervening stop signal is required to cause two
2405 * continued events regardless of ptrace.
2406 */
2407 if (!(sig->flags & SIGNAL_STOP_STOPPED))
2408 sig->group_exit_code = signr;
2409
2410 sig->group_stop_count = 0;
2411
2412 if (task_set_jobctl_pending(current, signr | gstop))
2413 sig->group_stop_count++;
2414
2415 t = current;
2416 while_each_thread(current, t) {
2417 /*
2418 * Setting state to TASK_STOPPED for a group
2419 * stop is always done with the siglock held,
2420 * so this check has no races.
2421 */
2422 if (!task_is_stopped(t) &&
2423 task_set_jobctl_pending(t, signr | gstop)) {
2424 sig->group_stop_count++;
2425 if (likely(!(t->ptrace & PT_SEIZED)))
2426 signal_wake_up(t, 0);
2427 else
2428 ptrace_trap_notify(t);
2429 }
2430 }
2431 }
2432
2433 if (likely(!current->ptrace)) {
2434 int notify = 0;
2435
2436 /*
2437 * If there are no other threads in the group, or if there
2438 * is a group stop in progress and we are the last to stop,
2439 * report to the parent.
2440 */
2441 if (task_participate_group_stop(current))
2442 notify = CLD_STOPPED;
2443
2444 set_special_state(TASK_STOPPED);
2445 spin_unlock_irq(¤t->sighand->siglock);
2446
2447 /*
2448 * Notify the parent of the group stop completion. Because
2449 * we're not holding either the siglock or tasklist_lock
2450 * here, ptracer may attach inbetween; however, this is for
2451 * group stop and should always be delivered to the real
2452 * parent of the group leader. The new ptracer will get
2453 * its notification when this task transitions into
2454 * TASK_TRACED.
2455 */
2456 if (notify) {
2457 read_lock(&tasklist_lock);
2458 do_notify_parent_cldstop(current, false, notify);
2459 read_unlock(&tasklist_lock);
2460 }
2461
2462 /* Now we don't run again until woken by SIGCONT or SIGKILL */
2463 cgroup_enter_frozen();
2464 freezable_schedule();
2465 return true;
2466 } else {
2467 /*
2468 * While ptraced, group stop is handled by STOP trap.
2469 * Schedule it and let the caller deal with it.
2470 */
2471 task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2472 return false;
2473 }
2474 }
2475
2476 /**
2477 * do_jobctl_trap - take care of ptrace jobctl traps
2478 *
2479 * When PT_SEIZED, it's used for both group stop and explicit
2480 * SEIZE/INTERRUPT traps. Both generate PTRACE_EVENT_STOP trap with
2481 * accompanying siginfo. If stopped, lower eight bits of exit_code contain
2482 * the stop signal; otherwise, %SIGTRAP.
2483 *
2484 * When !PT_SEIZED, it's used only for group stop trap with stop signal
2485 * number as exit_code and no siginfo.
2486 *
2487 * CONTEXT:
2488 * Must be called with @current->sighand->siglock held, which may be
2489 * released and re-acquired before returning with intervening sleep.
2490 */
do_jobctl_trap(void)2491 static void do_jobctl_trap(void)
2492 {
2493 struct signal_struct *signal = current->signal;
2494 int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2495
2496 if (current->ptrace & PT_SEIZED) {
2497 if (!signal->group_stop_count &&
2498 !(signal->flags & SIGNAL_STOP_STOPPED))
2499 signr = SIGTRAP;
2500 WARN_ON_ONCE(!signr);
2501 ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2502 CLD_STOPPED);
2503 } else {
2504 WARN_ON_ONCE(!signr);
2505 ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2506 current->exit_code = 0;
2507 }
2508 }
2509
2510 /**
2511 * do_freezer_trap - handle the freezer jobctl trap
2512 *
2513 * Puts the task into frozen state, if only the task is not about to quit.
2514 * In this case it drops JOBCTL_TRAP_FREEZE.
2515 *
2516 * CONTEXT:
2517 * Must be called with @current->sighand->siglock held,
2518 * which is always released before returning.
2519 */
do_freezer_trap(void)2520 static void do_freezer_trap(void)
2521 __releases(¤t->sighand->siglock)
2522 {
2523 /*
2524 * If there are other trap bits pending except JOBCTL_TRAP_FREEZE,
2525 * let's make another loop to give it a chance to be handled.
2526 * In any case, we'll return back.
2527 */
2528 if ((current->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) !=
2529 JOBCTL_TRAP_FREEZE) {
2530 spin_unlock_irq(¤t->sighand->siglock);
2531 return;
2532 }
2533
2534 /*
2535 * Now we're sure that there is no pending fatal signal and no
2536 * pending traps. Clear TIF_SIGPENDING to not get out of schedule()
2537 * immediately (if there is a non-fatal signal pending), and
2538 * put the task into sleep.
2539 */
2540 __set_current_state(TASK_INTERRUPTIBLE);
2541 clear_thread_flag(TIF_SIGPENDING);
2542 spin_unlock_irq(¤t->sighand->siglock);
2543 cgroup_enter_frozen();
2544 freezable_schedule();
2545 }
2546
ptrace_signal(int signr,kernel_siginfo_t * info)2547 static int ptrace_signal(int signr, kernel_siginfo_t *info)
2548 {
2549 /*
2550 * We do not check sig_kernel_stop(signr) but set this marker
2551 * unconditionally because we do not know whether debugger will
2552 * change signr. This flag has no meaning unless we are going
2553 * to stop after return from ptrace_stop(). In this case it will
2554 * be checked in do_signal_stop(), we should only stop if it was
2555 * not cleared by SIGCONT while we were sleeping. See also the
2556 * comment in dequeue_signal().
2557 */
2558 current->jobctl |= JOBCTL_STOP_DEQUEUED;
2559 ptrace_stop(signr, CLD_TRAPPED, 0, info);
2560
2561 /* We're back. Did the debugger cancel the sig? */
2562 signr = current->exit_code;
2563 if (signr == 0)
2564 return signr;
2565
2566 current->exit_code = 0;
2567
2568 /*
2569 * Update the siginfo structure if the signal has
2570 * changed. If the debugger wanted something
2571 * specific in the siginfo structure then it should
2572 * have updated *info via PTRACE_SETSIGINFO.
2573 */
2574 if (signr != info->si_signo) {
2575 clear_siginfo(info);
2576 info->si_signo = signr;
2577 info->si_errno = 0;
2578 info->si_code = SI_USER;
2579 rcu_read_lock();
2580 info->si_pid = task_pid_vnr(current->parent);
2581 info->si_uid = from_kuid_munged(current_user_ns(),
2582 task_uid(current->parent));
2583 rcu_read_unlock();
2584 }
2585
2586 /* If the (new) signal is now blocked, requeue it. */
2587 if (sigismember(¤t->blocked, signr)) {
2588 send_signal(signr, info, current, PIDTYPE_PID);
2589 signr = 0;
2590 }
2591
2592 return signr;
2593 }
2594
hide_si_addr_tag_bits(struct ksignal * ksig)2595 static void hide_si_addr_tag_bits(struct ksignal *ksig)
2596 {
2597 switch (siginfo_layout(ksig->sig, ksig->info.si_code)) {
2598 case SIL_FAULT:
2599 case SIL_FAULT_TRAPNO:
2600 case SIL_FAULT_MCEERR:
2601 case SIL_FAULT_BNDERR:
2602 case SIL_FAULT_PKUERR:
2603 case SIL_FAULT_PERF_EVENT:
2604 ksig->info.si_addr = arch_untagged_si_addr(
2605 ksig->info.si_addr, ksig->sig, ksig->info.si_code);
2606 break;
2607 case SIL_KILL:
2608 case SIL_TIMER:
2609 case SIL_POLL:
2610 case SIL_CHLD:
2611 case SIL_RT:
2612 case SIL_SYS:
2613 break;
2614 }
2615 }
2616
get_signal(struct ksignal * ksig)2617 bool get_signal(struct ksignal *ksig)
2618 {
2619 struct sighand_struct *sighand = current->sighand;
2620 struct signal_struct *signal = current->signal;
2621 int signr;
2622
2623 if (unlikely(current->task_works))
2624 task_work_run();
2625
2626 /*
2627 * For non-generic architectures, check for TIF_NOTIFY_SIGNAL so
2628 * that the arch handlers don't all have to do it. If we get here
2629 * without TIF_SIGPENDING, just exit after running signal work.
2630 */
2631 if (!IS_ENABLED(CONFIG_GENERIC_ENTRY)) {
2632 if (test_thread_flag(TIF_NOTIFY_SIGNAL))
2633 tracehook_notify_signal();
2634 if (!task_sigpending(current))
2635 return false;
2636 }
2637
2638 if (unlikely(uprobe_deny_signal()))
2639 return false;
2640
2641 /*
2642 * Do this once, we can't return to user-mode if freezing() == T.
2643 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2644 * thus do not need another check after return.
2645 */
2646 try_to_freeze();
2647
2648 relock:
2649 spin_lock_irq(&sighand->siglock);
2650
2651 /*
2652 * Every stopped thread goes here after wakeup. Check to see if
2653 * we should notify the parent, prepare_signal(SIGCONT) encodes
2654 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2655 */
2656 if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2657 int why;
2658
2659 if (signal->flags & SIGNAL_CLD_CONTINUED)
2660 why = CLD_CONTINUED;
2661 else
2662 why = CLD_STOPPED;
2663
2664 signal->flags &= ~SIGNAL_CLD_MASK;
2665
2666 spin_unlock_irq(&sighand->siglock);
2667
2668 /*
2669 * Notify the parent that we're continuing. This event is
2670 * always per-process and doesn't make whole lot of sense
2671 * for ptracers, who shouldn't consume the state via
2672 * wait(2) either, but, for backward compatibility, notify
2673 * the ptracer of the group leader too unless it's gonna be
2674 * a duplicate.
2675 */
2676 read_lock(&tasklist_lock);
2677 do_notify_parent_cldstop(current, false, why);
2678
2679 if (ptrace_reparented(current->group_leader))
2680 do_notify_parent_cldstop(current->group_leader,
2681 true, why);
2682 read_unlock(&tasklist_lock);
2683
2684 goto relock;
2685 }
2686
2687 /* Has this task already been marked for death? */
2688 if (signal_group_exit(signal)) {
2689 ksig->info.si_signo = signr = SIGKILL;
2690 sigdelset(¤t->pending.signal, SIGKILL);
2691 trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
2692 &sighand->action[SIGKILL - 1]);
2693 recalc_sigpending();
2694 goto fatal;
2695 }
2696
2697 for (;;) {
2698 struct k_sigaction *ka;
2699
2700 if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2701 do_signal_stop(0))
2702 goto relock;
2703
2704 if (unlikely(current->jobctl &
2705 (JOBCTL_TRAP_MASK | JOBCTL_TRAP_FREEZE))) {
2706 if (current->jobctl & JOBCTL_TRAP_MASK) {
2707 do_jobctl_trap();
2708 spin_unlock_irq(&sighand->siglock);
2709 } else if (current->jobctl & JOBCTL_TRAP_FREEZE)
2710 do_freezer_trap();
2711
2712 goto relock;
2713 }
2714
2715 /*
2716 * If the task is leaving the frozen state, let's update
2717 * cgroup counters and reset the frozen bit.
2718 */
2719 if (unlikely(cgroup_task_frozen(current))) {
2720 spin_unlock_irq(&sighand->siglock);
2721 cgroup_leave_frozen(false);
2722 goto relock;
2723 }
2724
2725 /*
2726 * Signals generated by the execution of an instruction
2727 * need to be delivered before any other pending signals
2728 * so that the instruction pointer in the signal stack
2729 * frame points to the faulting instruction.
2730 */
2731 signr = dequeue_synchronous_signal(&ksig->info);
2732 if (!signr)
2733 signr = dequeue_signal(current, ¤t->blocked, &ksig->info);
2734
2735 if (!signr)
2736 break; /* will return 0 */
2737
2738 if (unlikely(current->ptrace) && (signr != SIGKILL) &&
2739 !(sighand->action[signr -1].sa.sa_flags & SA_IMMUTABLE)) {
2740 signr = ptrace_signal(signr, &ksig->info);
2741 if (!signr)
2742 continue;
2743 }
2744
2745 ka = &sighand->action[signr-1];
2746
2747 /* Trace actually delivered signals. */
2748 trace_signal_deliver(signr, &ksig->info, ka);
2749
2750 if (ka->sa.sa_handler == SIG_IGN) /* Do nothing. */
2751 continue;
2752 if (ka->sa.sa_handler != SIG_DFL) {
2753 /* Run the handler. */
2754 ksig->ka = *ka;
2755
2756 if (ka->sa.sa_flags & SA_ONESHOT)
2757 ka->sa.sa_handler = SIG_DFL;
2758
2759 break; /* will return non-zero "signr" value */
2760 }
2761
2762 /*
2763 * Now we are doing the default action for this signal.
2764 */
2765 if (sig_kernel_ignore(signr)) /* Default is nothing. */
2766 continue;
2767
2768 /*
2769 * Global init gets no signals it doesn't want.
2770 * Container-init gets no signals it doesn't want from same
2771 * container.
2772 *
2773 * Note that if global/container-init sees a sig_kernel_only()
2774 * signal here, the signal must have been generated internally
2775 * or must have come from an ancestor namespace. In either
2776 * case, the signal cannot be dropped.
2777 */
2778 if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2779 !sig_kernel_only(signr))
2780 continue;
2781
2782 if (sig_kernel_stop(signr)) {
2783 /*
2784 * The default action is to stop all threads in
2785 * the thread group. The job control signals
2786 * do nothing in an orphaned pgrp, but SIGSTOP
2787 * always works. Note that siglock needs to be
2788 * dropped during the call to is_orphaned_pgrp()
2789 * because of lock ordering with tasklist_lock.
2790 * This allows an intervening SIGCONT to be posted.
2791 * We need to check for that and bail out if necessary.
2792 */
2793 if (signr != SIGSTOP) {
2794 spin_unlock_irq(&sighand->siglock);
2795
2796 /* signals can be posted during this window */
2797
2798 if (is_current_pgrp_orphaned())
2799 goto relock;
2800
2801 spin_lock_irq(&sighand->siglock);
2802 }
2803
2804 if (likely(do_signal_stop(ksig->info.si_signo))) {
2805 /* It released the siglock. */
2806 goto relock;
2807 }
2808
2809 /*
2810 * We didn't actually stop, due to a race
2811 * with SIGCONT or something like that.
2812 */
2813 continue;
2814 }
2815
2816 fatal:
2817 spin_unlock_irq(&sighand->siglock);
2818 if (unlikely(cgroup_task_frozen(current)))
2819 cgroup_leave_frozen(true);
2820
2821 /*
2822 * Anything else is fatal, maybe with a core dump.
2823 */
2824 current->flags |= PF_SIGNALED;
2825
2826 if (sig_kernel_coredump(signr)) {
2827 if (print_fatal_signals)
2828 print_fatal_signal(ksig->info.si_signo);
2829 proc_coredump_connector(current);
2830 /*
2831 * If it was able to dump core, this kills all
2832 * other threads in the group and synchronizes with
2833 * their demise. If we lost the race with another
2834 * thread getting here, it set group_exit_code
2835 * first and our do_group_exit call below will use
2836 * that value and ignore the one we pass it.
2837 */
2838 do_coredump(&ksig->info);
2839 }
2840
2841 /*
2842 * PF_IO_WORKER threads will catch and exit on fatal signals
2843 * themselves. They have cleanup that must be performed, so
2844 * we cannot call do_exit() on their behalf.
2845 */
2846 if (current->flags & PF_IO_WORKER)
2847 goto out;
2848
2849 /*
2850 * Death signals, no core dump.
2851 */
2852 do_group_exit(ksig->info.si_signo);
2853 /* NOTREACHED */
2854 }
2855 spin_unlock_irq(&sighand->siglock);
2856 out:
2857 ksig->sig = signr;
2858
2859 if (!(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
2860 hide_si_addr_tag_bits(ksig);
2861
2862 return ksig->sig > 0;
2863 }
2864
2865 /**
2866 * signal_delivered -
2867 * @ksig: kernel signal struct
2868 * @stepping: nonzero if debugger single-step or block-step in use
2869 *
2870 * This function should be called when a signal has successfully been
2871 * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
2872 * is always blocked, and the signal itself is blocked unless %SA_NODEFER
2873 * is set in @ksig->ka.sa.sa_flags. Tracing is notified.
2874 */
signal_delivered(struct ksignal * ksig,int stepping)2875 static void signal_delivered(struct ksignal *ksig, int stepping)
2876 {
2877 sigset_t blocked;
2878
2879 /* A signal was successfully delivered, and the
2880 saved sigmask was stored on the signal frame,
2881 and will be restored by sigreturn. So we can
2882 simply clear the restore sigmask flag. */
2883 clear_restore_sigmask();
2884
2885 sigorsets(&blocked, ¤t->blocked, &ksig->ka.sa.sa_mask);
2886 if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
2887 sigaddset(&blocked, ksig->sig);
2888 set_current_blocked(&blocked);
2889 if (current->sas_ss_flags & SS_AUTODISARM)
2890 sas_ss_reset(current);
2891 tracehook_signal_handler(stepping);
2892 }
2893
signal_setup_done(int failed,struct ksignal * ksig,int stepping)2894 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
2895 {
2896 if (failed)
2897 force_sigsegv(ksig->sig);
2898 else
2899 signal_delivered(ksig, stepping);
2900 }
2901
2902 /*
2903 * It could be that complete_signal() picked us to notify about the
2904 * group-wide signal. Other threads should be notified now to take
2905 * the shared signals in @which since we will not.
2906 */
retarget_shared_pending(struct task_struct * tsk,sigset_t * which)2907 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2908 {
2909 sigset_t retarget;
2910 struct task_struct *t;
2911
2912 sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2913 if (sigisemptyset(&retarget))
2914 return;
2915
2916 t = tsk;
2917 while_each_thread(tsk, t) {
2918 if (t->flags & PF_EXITING)
2919 continue;
2920
2921 if (!has_pending_signals(&retarget, &t->blocked))
2922 continue;
2923 /* Remove the signals this thread can handle. */
2924 sigandsets(&retarget, &retarget, &t->blocked);
2925
2926 if (!task_sigpending(t))
2927 signal_wake_up(t, 0);
2928
2929 if (sigisemptyset(&retarget))
2930 break;
2931 }
2932 }
2933
exit_signals(struct task_struct * tsk)2934 void exit_signals(struct task_struct *tsk)
2935 {
2936 int group_stop = 0;
2937 sigset_t unblocked;
2938
2939 /*
2940 * @tsk is about to have PF_EXITING set - lock out users which
2941 * expect stable threadgroup.
2942 */
2943 cgroup_threadgroup_change_begin(tsk);
2944
2945 if (thread_group_empty(tsk) || signal_group_exit(tsk->signal)) {
2946 tsk->flags |= PF_EXITING;
2947 cgroup_threadgroup_change_end(tsk);
2948 return;
2949 }
2950
2951 spin_lock_irq(&tsk->sighand->siglock);
2952 /*
2953 * From now this task is not visible for group-wide signals,
2954 * see wants_signal(), do_signal_stop().
2955 */
2956 tsk->flags |= PF_EXITING;
2957
2958 cgroup_threadgroup_change_end(tsk);
2959
2960 if (!task_sigpending(tsk))
2961 goto out;
2962
2963 unblocked = tsk->blocked;
2964 signotset(&unblocked);
2965 retarget_shared_pending(tsk, &unblocked);
2966
2967 if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2968 task_participate_group_stop(tsk))
2969 group_stop = CLD_STOPPED;
2970 out:
2971 spin_unlock_irq(&tsk->sighand->siglock);
2972
2973 /*
2974 * If group stop has completed, deliver the notification. This
2975 * should always go to the real parent of the group leader.
2976 */
2977 if (unlikely(group_stop)) {
2978 read_lock(&tasklist_lock);
2979 do_notify_parent_cldstop(tsk, false, group_stop);
2980 read_unlock(&tasklist_lock);
2981 }
2982 }
2983
2984 /*
2985 * System call entry points.
2986 */
2987
2988 /**
2989 * sys_restart_syscall - restart a system call
2990 */
SYSCALL_DEFINE0(restart_syscall)2991 SYSCALL_DEFINE0(restart_syscall)
2992 {
2993 struct restart_block *restart = ¤t->restart_block;
2994 return restart->fn(restart);
2995 }
2996
do_no_restart_syscall(struct restart_block * param)2997 long do_no_restart_syscall(struct restart_block *param)
2998 {
2999 return -EINTR;
3000 }
3001
__set_task_blocked(struct task_struct * tsk,const sigset_t * newset)3002 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
3003 {
3004 if (task_sigpending(tsk) && !thread_group_empty(tsk)) {
3005 sigset_t newblocked;
3006 /* A set of now blocked but previously unblocked signals. */
3007 sigandnsets(&newblocked, newset, ¤t->blocked);
3008 retarget_shared_pending(tsk, &newblocked);
3009 }
3010 tsk->blocked = *newset;
3011 recalc_sigpending();
3012 }
3013
3014 /**
3015 * set_current_blocked - change current->blocked mask
3016 * @newset: new mask
3017 *
3018 * It is wrong to change ->blocked directly, this helper should be used
3019 * to ensure the process can't miss a shared signal we are going to block.
3020 */
set_current_blocked(sigset_t * newset)3021 void set_current_blocked(sigset_t *newset)
3022 {
3023 sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
3024 __set_current_blocked(newset);
3025 }
3026
__set_current_blocked(const sigset_t * newset)3027 void __set_current_blocked(const sigset_t *newset)
3028 {
3029 struct task_struct *tsk = current;
3030
3031 /*
3032 * In case the signal mask hasn't changed, there is nothing we need
3033 * to do. The current->blocked shouldn't be modified by other task.
3034 */
3035 if (sigequalsets(&tsk->blocked, newset))
3036 return;
3037
3038 spin_lock_irq(&tsk->sighand->siglock);
3039 __set_task_blocked(tsk, newset);
3040 spin_unlock_irq(&tsk->sighand->siglock);
3041 }
3042
3043 /*
3044 * This is also useful for kernel threads that want to temporarily
3045 * (or permanently) block certain signals.
3046 *
3047 * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
3048 * interface happily blocks "unblockable" signals like SIGKILL
3049 * and friends.
3050 */
sigprocmask(int how,sigset_t * set,sigset_t * oldset)3051 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
3052 {
3053 struct task_struct *tsk = current;
3054 sigset_t newset;
3055
3056 /* Lockless, only current can change ->blocked, never from irq */
3057 if (oldset)
3058 *oldset = tsk->blocked;
3059
3060 switch (how) {
3061 case SIG_BLOCK:
3062 sigorsets(&newset, &tsk->blocked, set);
3063 break;
3064 case SIG_UNBLOCK:
3065 sigandnsets(&newset, &tsk->blocked, set);
3066 break;
3067 case SIG_SETMASK:
3068 newset = *set;
3069 break;
3070 default:
3071 return -EINVAL;
3072 }
3073
3074 __set_current_blocked(&newset);
3075 return 0;
3076 }
3077 EXPORT_SYMBOL(sigprocmask);
3078
3079 /*
3080 * The api helps set app-provided sigmasks.
3081 *
3082 * This is useful for syscalls such as ppoll, pselect, io_pgetevents and
3083 * epoll_pwait where a new sigmask is passed from userland for the syscalls.
3084 *
3085 * Note that it does set_restore_sigmask() in advance, so it must be always
3086 * paired with restore_saved_sigmask_unless() before return from syscall.
3087 */
set_user_sigmask(const sigset_t __user * umask,size_t sigsetsize)3088 int set_user_sigmask(const sigset_t __user *umask, size_t sigsetsize)
3089 {
3090 sigset_t kmask;
3091
3092 if (!umask)
3093 return 0;
3094 if (sigsetsize != sizeof(sigset_t))
3095 return -EINVAL;
3096 if (copy_from_user(&kmask, umask, sizeof(sigset_t)))
3097 return -EFAULT;
3098
3099 set_restore_sigmask();
3100 current->saved_sigmask = current->blocked;
3101 set_current_blocked(&kmask);
3102
3103 return 0;
3104 }
3105
3106 #ifdef CONFIG_COMPAT
set_compat_user_sigmask(const compat_sigset_t __user * umask,size_t sigsetsize)3107 int set_compat_user_sigmask(const compat_sigset_t __user *umask,
3108 size_t sigsetsize)
3109 {
3110 sigset_t kmask;
3111
3112 if (!umask)
3113 return 0;
3114 if (sigsetsize != sizeof(compat_sigset_t))
3115 return -EINVAL;
3116 if (get_compat_sigset(&kmask, umask))
3117 return -EFAULT;
3118
3119 set_restore_sigmask();
3120 current->saved_sigmask = current->blocked;
3121 set_current_blocked(&kmask);
3122
3123 return 0;
3124 }
3125 #endif
3126
3127 /**
3128 * sys_rt_sigprocmask - change the list of currently blocked signals
3129 * @how: whether to add, remove, or set signals
3130 * @nset: stores pending signals
3131 * @oset: previous value of signal mask if non-null
3132 * @sigsetsize: size of sigset_t type
3133 */
SYSCALL_DEFINE4(rt_sigprocmask,int,how,sigset_t __user *,nset,sigset_t __user *,oset,size_t,sigsetsize)3134 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
3135 sigset_t __user *, oset, size_t, sigsetsize)
3136 {
3137 sigset_t old_set, new_set;
3138 int error;
3139
3140 /* XXX: Don't preclude handling different sized sigset_t's. */
3141 if (sigsetsize != sizeof(sigset_t))
3142 return -EINVAL;
3143
3144 old_set = current->blocked;
3145
3146 if (nset) {
3147 if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
3148 return -EFAULT;
3149 sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3150
3151 error = sigprocmask(how, &new_set, NULL);
3152 if (error)
3153 return error;
3154 }
3155
3156 if (oset) {
3157 if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
3158 return -EFAULT;
3159 }
3160
3161 return 0;
3162 }
3163
3164 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(rt_sigprocmask,int,how,compat_sigset_t __user *,nset,compat_sigset_t __user *,oset,compat_size_t,sigsetsize)3165 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
3166 compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
3167 {
3168 sigset_t old_set = current->blocked;
3169
3170 /* XXX: Don't preclude handling different sized sigset_t's. */
3171 if (sigsetsize != sizeof(sigset_t))
3172 return -EINVAL;
3173
3174 if (nset) {
3175 sigset_t new_set;
3176 int error;
3177 if (get_compat_sigset(&new_set, nset))
3178 return -EFAULT;
3179 sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3180
3181 error = sigprocmask(how, &new_set, NULL);
3182 if (error)
3183 return error;
3184 }
3185 return oset ? put_compat_sigset(oset, &old_set, sizeof(*oset)) : 0;
3186 }
3187 #endif
3188
do_sigpending(sigset_t * set)3189 static void do_sigpending(sigset_t *set)
3190 {
3191 spin_lock_irq(¤t->sighand->siglock);
3192 sigorsets(set, ¤t->pending.signal,
3193 ¤t->signal->shared_pending.signal);
3194 spin_unlock_irq(¤t->sighand->siglock);
3195
3196 /* Outside the lock because only this thread touches it. */
3197 sigandsets(set, ¤t->blocked, set);
3198 }
3199
3200 /**
3201 * sys_rt_sigpending - examine a pending signal that has been raised
3202 * while blocked
3203 * @uset: stores pending signals
3204 * @sigsetsize: size of sigset_t type or larger
3205 */
SYSCALL_DEFINE2(rt_sigpending,sigset_t __user *,uset,size_t,sigsetsize)3206 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
3207 {
3208 sigset_t set;
3209
3210 if (sigsetsize > sizeof(*uset))
3211 return -EINVAL;
3212
3213 do_sigpending(&set);
3214
3215 if (copy_to_user(uset, &set, sigsetsize))
3216 return -EFAULT;
3217
3218 return 0;
3219 }
3220
3221 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(rt_sigpending,compat_sigset_t __user *,uset,compat_size_t,sigsetsize)3222 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
3223 compat_size_t, sigsetsize)
3224 {
3225 sigset_t set;
3226
3227 if (sigsetsize > sizeof(*uset))
3228 return -EINVAL;
3229
3230 do_sigpending(&set);
3231
3232 return put_compat_sigset(uset, &set, sigsetsize);
3233 }
3234 #endif
3235
3236 static const struct {
3237 unsigned char limit, layout;
3238 } sig_sicodes[] = {
3239 [SIGILL] = { NSIGILL, SIL_FAULT },
3240 [SIGFPE] = { NSIGFPE, SIL_FAULT },
3241 [SIGSEGV] = { NSIGSEGV, SIL_FAULT },
3242 [SIGBUS] = { NSIGBUS, SIL_FAULT },
3243 [SIGTRAP] = { NSIGTRAP, SIL_FAULT },
3244 #if defined(SIGEMT)
3245 [SIGEMT] = { NSIGEMT, SIL_FAULT },
3246 #endif
3247 [SIGCHLD] = { NSIGCHLD, SIL_CHLD },
3248 [SIGPOLL] = { NSIGPOLL, SIL_POLL },
3249 [SIGSYS] = { NSIGSYS, SIL_SYS },
3250 };
3251
known_siginfo_layout(unsigned sig,int si_code)3252 static bool known_siginfo_layout(unsigned sig, int si_code)
3253 {
3254 if (si_code == SI_KERNEL)
3255 return true;
3256 else if ((si_code > SI_USER)) {
3257 if (sig_specific_sicodes(sig)) {
3258 if (si_code <= sig_sicodes[sig].limit)
3259 return true;
3260 }
3261 else if (si_code <= NSIGPOLL)
3262 return true;
3263 }
3264 else if (si_code >= SI_DETHREAD)
3265 return true;
3266 else if (si_code == SI_ASYNCNL)
3267 return true;
3268 return false;
3269 }
3270
siginfo_layout(unsigned sig,int si_code)3271 enum siginfo_layout siginfo_layout(unsigned sig, int si_code)
3272 {
3273 enum siginfo_layout layout = SIL_KILL;
3274 if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
3275 if ((sig < ARRAY_SIZE(sig_sicodes)) &&
3276 (si_code <= sig_sicodes[sig].limit)) {
3277 layout = sig_sicodes[sig].layout;
3278 /* Handle the exceptions */
3279 if ((sig == SIGBUS) &&
3280 (si_code >= BUS_MCEERR_AR) && (si_code <= BUS_MCEERR_AO))
3281 layout = SIL_FAULT_MCEERR;
3282 else if ((sig == SIGSEGV) && (si_code == SEGV_BNDERR))
3283 layout = SIL_FAULT_BNDERR;
3284 #ifdef SEGV_PKUERR
3285 else if ((sig == SIGSEGV) && (si_code == SEGV_PKUERR))
3286 layout = SIL_FAULT_PKUERR;
3287 #endif
3288 else if ((sig == SIGTRAP) && (si_code == TRAP_PERF))
3289 layout = SIL_FAULT_PERF_EVENT;
3290 else if (IS_ENABLED(CONFIG_SPARC) &&
3291 (sig == SIGILL) && (si_code == ILL_ILLTRP))
3292 layout = SIL_FAULT_TRAPNO;
3293 else if (IS_ENABLED(CONFIG_ALPHA) &&
3294 ((sig == SIGFPE) ||
3295 ((sig == SIGTRAP) && (si_code == TRAP_UNK))))
3296 layout = SIL_FAULT_TRAPNO;
3297 }
3298 else if (si_code <= NSIGPOLL)
3299 layout = SIL_POLL;
3300 } else {
3301 if (si_code == SI_TIMER)
3302 layout = SIL_TIMER;
3303 else if (si_code == SI_SIGIO)
3304 layout = SIL_POLL;
3305 else if (si_code < 0)
3306 layout = SIL_RT;
3307 }
3308 return layout;
3309 }
3310
si_expansion(const siginfo_t __user * info)3311 static inline char __user *si_expansion(const siginfo_t __user *info)
3312 {
3313 return ((char __user *)info) + sizeof(struct kernel_siginfo);
3314 }
3315
copy_siginfo_to_user(siginfo_t __user * to,const kernel_siginfo_t * from)3316 int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from)
3317 {
3318 char __user *expansion = si_expansion(to);
3319 if (copy_to_user(to, from , sizeof(struct kernel_siginfo)))
3320 return -EFAULT;
3321 if (clear_user(expansion, SI_EXPANSION_SIZE))
3322 return -EFAULT;
3323 return 0;
3324 }
3325
post_copy_siginfo_from_user(kernel_siginfo_t * info,const siginfo_t __user * from)3326 static int post_copy_siginfo_from_user(kernel_siginfo_t *info,
3327 const siginfo_t __user *from)
3328 {
3329 if (unlikely(!known_siginfo_layout(info->si_signo, info->si_code))) {
3330 char __user *expansion = si_expansion(from);
3331 char buf[SI_EXPANSION_SIZE];
3332 int i;
3333 /*
3334 * An unknown si_code might need more than
3335 * sizeof(struct kernel_siginfo) bytes. Verify all of the
3336 * extra bytes are 0. This guarantees copy_siginfo_to_user
3337 * will return this data to userspace exactly.
3338 */
3339 if (copy_from_user(&buf, expansion, SI_EXPANSION_SIZE))
3340 return -EFAULT;
3341 for (i = 0; i < SI_EXPANSION_SIZE; i++) {
3342 if (buf[i] != 0)
3343 return -E2BIG;
3344 }
3345 }
3346 return 0;
3347 }
3348
__copy_siginfo_from_user(int signo,kernel_siginfo_t * to,const siginfo_t __user * from)3349 static int __copy_siginfo_from_user(int signo, kernel_siginfo_t *to,
3350 const siginfo_t __user *from)
3351 {
3352 if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3353 return -EFAULT;
3354 to->si_signo = signo;
3355 return post_copy_siginfo_from_user(to, from);
3356 }
3357
copy_siginfo_from_user(kernel_siginfo_t * to,const siginfo_t __user * from)3358 int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from)
3359 {
3360 if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3361 return -EFAULT;
3362 return post_copy_siginfo_from_user(to, from);
3363 }
3364
3365 #ifdef CONFIG_COMPAT
3366 /**
3367 * copy_siginfo_to_external32 - copy a kernel siginfo into a compat user siginfo
3368 * @to: compat siginfo destination
3369 * @from: kernel siginfo source
3370 *
3371 * Note: This function does not work properly for the SIGCHLD on x32, but
3372 * fortunately it doesn't have to. The only valid callers for this function are
3373 * copy_siginfo_to_user32, which is overriden for x32 and the coredump code.
3374 * The latter does not care because SIGCHLD will never cause a coredump.
3375 */
copy_siginfo_to_external32(struct compat_siginfo * to,const struct kernel_siginfo * from)3376 void copy_siginfo_to_external32(struct compat_siginfo *to,
3377 const struct kernel_siginfo *from)
3378 {
3379 memset(to, 0, sizeof(*to));
3380
3381 to->si_signo = from->si_signo;
3382 to->si_errno = from->si_errno;
3383 to->si_code = from->si_code;
3384 switch(siginfo_layout(from->si_signo, from->si_code)) {
3385 case SIL_KILL:
3386 to->si_pid = from->si_pid;
3387 to->si_uid = from->si_uid;
3388 break;
3389 case SIL_TIMER:
3390 to->si_tid = from->si_tid;
3391 to->si_overrun = from->si_overrun;
3392 to->si_int = from->si_int;
3393 break;
3394 case SIL_POLL:
3395 to->si_band = from->si_band;
3396 to->si_fd = from->si_fd;
3397 break;
3398 case SIL_FAULT:
3399 to->si_addr = ptr_to_compat(from->si_addr);
3400 break;
3401 case SIL_FAULT_TRAPNO:
3402 to->si_addr = ptr_to_compat(from->si_addr);
3403 to->si_trapno = from->si_trapno;
3404 break;
3405 case SIL_FAULT_MCEERR:
3406 to->si_addr = ptr_to_compat(from->si_addr);
3407 to->si_addr_lsb = from->si_addr_lsb;
3408 break;
3409 case SIL_FAULT_BNDERR:
3410 to->si_addr = ptr_to_compat(from->si_addr);
3411 to->si_lower = ptr_to_compat(from->si_lower);
3412 to->si_upper = ptr_to_compat(from->si_upper);
3413 break;
3414 case SIL_FAULT_PKUERR:
3415 to->si_addr = ptr_to_compat(from->si_addr);
3416 to->si_pkey = from->si_pkey;
3417 break;
3418 case SIL_FAULT_PERF_EVENT:
3419 to->si_addr = ptr_to_compat(from->si_addr);
3420 to->si_perf_data = from->si_perf_data;
3421 to->si_perf_type = from->si_perf_type;
3422 break;
3423 case SIL_CHLD:
3424 to->si_pid = from->si_pid;
3425 to->si_uid = from->si_uid;
3426 to->si_status = from->si_status;
3427 to->si_utime = from->si_utime;
3428 to->si_stime = from->si_stime;
3429 break;
3430 case SIL_RT:
3431 to->si_pid = from->si_pid;
3432 to->si_uid = from->si_uid;
3433 to->si_int = from->si_int;
3434 break;
3435 case SIL_SYS:
3436 to->si_call_addr = ptr_to_compat(from->si_call_addr);
3437 to->si_syscall = from->si_syscall;
3438 to->si_arch = from->si_arch;
3439 break;
3440 }
3441 }
3442
__copy_siginfo_to_user32(struct compat_siginfo __user * to,const struct kernel_siginfo * from)3443 int __copy_siginfo_to_user32(struct compat_siginfo __user *to,
3444 const struct kernel_siginfo *from)
3445 {
3446 struct compat_siginfo new;
3447
3448 copy_siginfo_to_external32(&new, from);
3449 if (copy_to_user(to, &new, sizeof(struct compat_siginfo)))
3450 return -EFAULT;
3451 return 0;
3452 }
3453
post_copy_siginfo_from_user32(kernel_siginfo_t * to,const struct compat_siginfo * from)3454 static int post_copy_siginfo_from_user32(kernel_siginfo_t *to,
3455 const struct compat_siginfo *from)
3456 {
3457 clear_siginfo(to);
3458 to->si_signo = from->si_signo;
3459 to->si_errno = from->si_errno;
3460 to->si_code = from->si_code;
3461 switch(siginfo_layout(from->si_signo, from->si_code)) {
3462 case SIL_KILL:
3463 to->si_pid = from->si_pid;
3464 to->si_uid = from->si_uid;
3465 break;
3466 case SIL_TIMER:
3467 to->si_tid = from->si_tid;
3468 to->si_overrun = from->si_overrun;
3469 to->si_int = from->si_int;
3470 break;
3471 case SIL_POLL:
3472 to->si_band = from->si_band;
3473 to->si_fd = from->si_fd;
3474 break;
3475 case SIL_FAULT:
3476 to->si_addr = compat_ptr(from->si_addr);
3477 break;
3478 case SIL_FAULT_TRAPNO:
3479 to->si_addr = compat_ptr(from->si_addr);
3480 to->si_trapno = from->si_trapno;
3481 break;
3482 case SIL_FAULT_MCEERR:
3483 to->si_addr = compat_ptr(from->si_addr);
3484 to->si_addr_lsb = from->si_addr_lsb;
3485 break;
3486 case SIL_FAULT_BNDERR:
3487 to->si_addr = compat_ptr(from->si_addr);
3488 to->si_lower = compat_ptr(from->si_lower);
3489 to->si_upper = compat_ptr(from->si_upper);
3490 break;
3491 case SIL_FAULT_PKUERR:
3492 to->si_addr = compat_ptr(from->si_addr);
3493 to->si_pkey = from->si_pkey;
3494 break;
3495 case SIL_FAULT_PERF_EVENT:
3496 to->si_addr = compat_ptr(from->si_addr);
3497 to->si_perf_data = from->si_perf_data;
3498 to->si_perf_type = from->si_perf_type;
3499 break;
3500 case SIL_CHLD:
3501 to->si_pid = from->si_pid;
3502 to->si_uid = from->si_uid;
3503 to->si_status = from->si_status;
3504 #ifdef CONFIG_X86_X32_ABI
3505 if (in_x32_syscall()) {
3506 to->si_utime = from->_sifields._sigchld_x32._utime;
3507 to->si_stime = from->_sifields._sigchld_x32._stime;
3508 } else
3509 #endif
3510 {
3511 to->si_utime = from->si_utime;
3512 to->si_stime = from->si_stime;
3513 }
3514 break;
3515 case SIL_RT:
3516 to->si_pid = from->si_pid;
3517 to->si_uid = from->si_uid;
3518 to->si_int = from->si_int;
3519 break;
3520 case SIL_SYS:
3521 to->si_call_addr = compat_ptr(from->si_call_addr);
3522 to->si_syscall = from->si_syscall;
3523 to->si_arch = from->si_arch;
3524 break;
3525 }
3526 return 0;
3527 }
3528
__copy_siginfo_from_user32(int signo,struct kernel_siginfo * to,const struct compat_siginfo __user * ufrom)3529 static int __copy_siginfo_from_user32(int signo, struct kernel_siginfo *to,
3530 const struct compat_siginfo __user *ufrom)
3531 {
3532 struct compat_siginfo from;
3533
3534 if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3535 return -EFAULT;
3536
3537 from.si_signo = signo;
3538 return post_copy_siginfo_from_user32(to, &from);
3539 }
3540
copy_siginfo_from_user32(struct kernel_siginfo * to,const struct compat_siginfo __user * ufrom)3541 int copy_siginfo_from_user32(struct kernel_siginfo *to,
3542 const struct compat_siginfo __user *ufrom)
3543 {
3544 struct compat_siginfo from;
3545
3546 if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3547 return -EFAULT;
3548
3549 return post_copy_siginfo_from_user32(to, &from);
3550 }
3551 #endif /* CONFIG_COMPAT */
3552
3553 /**
3554 * do_sigtimedwait - wait for queued signals specified in @which
3555 * @which: queued signals to wait for
3556 * @info: if non-null, the signal's siginfo is returned here
3557 * @ts: upper bound on process time suspension
3558 */
do_sigtimedwait(const sigset_t * which,kernel_siginfo_t * info,const struct timespec64 * ts)3559 static int do_sigtimedwait(const sigset_t *which, kernel_siginfo_t *info,
3560 const struct timespec64 *ts)
3561 {
3562 ktime_t *to = NULL, timeout = KTIME_MAX;
3563 struct task_struct *tsk = current;
3564 sigset_t mask = *which;
3565 int sig, ret = 0;
3566
3567 if (ts) {
3568 if (!timespec64_valid(ts))
3569 return -EINVAL;
3570 timeout = timespec64_to_ktime(*ts);
3571 to = &timeout;
3572 }
3573
3574 /*
3575 * Invert the set of allowed signals to get those we want to block.
3576 */
3577 sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
3578 signotset(&mask);
3579
3580 spin_lock_irq(&tsk->sighand->siglock);
3581 sig = dequeue_signal(tsk, &mask, info);
3582 if (!sig && timeout) {
3583 /*
3584 * None ready, temporarily unblock those we're interested
3585 * while we are sleeping in so that we'll be awakened when
3586 * they arrive. Unblocking is always fine, we can avoid
3587 * set_current_blocked().
3588 */
3589 tsk->real_blocked = tsk->blocked;
3590 sigandsets(&tsk->blocked, &tsk->blocked, &mask);
3591 recalc_sigpending();
3592 spin_unlock_irq(&tsk->sighand->siglock);
3593
3594 __set_current_state(TASK_INTERRUPTIBLE);
3595 ret = freezable_schedule_hrtimeout_range(to, tsk->timer_slack_ns,
3596 HRTIMER_MODE_REL);
3597 spin_lock_irq(&tsk->sighand->siglock);
3598 __set_task_blocked(tsk, &tsk->real_blocked);
3599 sigemptyset(&tsk->real_blocked);
3600 sig = dequeue_signal(tsk, &mask, info);
3601 }
3602 spin_unlock_irq(&tsk->sighand->siglock);
3603
3604 if (sig)
3605 return sig;
3606 return ret ? -EINTR : -EAGAIN;
3607 }
3608
3609 /**
3610 * sys_rt_sigtimedwait - synchronously wait for queued signals specified
3611 * in @uthese
3612 * @uthese: queued signals to wait for
3613 * @uinfo: if non-null, the signal's siginfo is returned here
3614 * @uts: upper bound on process time suspension
3615 * @sigsetsize: size of sigset_t type
3616 */
SYSCALL_DEFINE4(rt_sigtimedwait,const sigset_t __user *,uthese,siginfo_t __user *,uinfo,const struct __kernel_timespec __user *,uts,size_t,sigsetsize)3617 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
3618 siginfo_t __user *, uinfo,
3619 const struct __kernel_timespec __user *, uts,
3620 size_t, sigsetsize)
3621 {
3622 sigset_t these;
3623 struct timespec64 ts;
3624 kernel_siginfo_t info;
3625 int ret;
3626
3627 /* XXX: Don't preclude handling different sized sigset_t's. */
3628 if (sigsetsize != sizeof(sigset_t))
3629 return -EINVAL;
3630
3631 if (copy_from_user(&these, uthese, sizeof(these)))
3632 return -EFAULT;
3633
3634 if (uts) {
3635 if (get_timespec64(&ts, uts))
3636 return -EFAULT;
3637 }
3638
3639 ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3640
3641 if (ret > 0 && uinfo) {
3642 if (copy_siginfo_to_user(uinfo, &info))
3643 ret = -EFAULT;
3644 }
3645
3646 return ret;
3647 }
3648
3649 #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)3650 SYSCALL_DEFINE4(rt_sigtimedwait_time32, const sigset_t __user *, uthese,
3651 siginfo_t __user *, uinfo,
3652 const struct old_timespec32 __user *, uts,
3653 size_t, sigsetsize)
3654 {
3655 sigset_t these;
3656 struct timespec64 ts;
3657 kernel_siginfo_t info;
3658 int ret;
3659
3660 if (sigsetsize != sizeof(sigset_t))
3661 return -EINVAL;
3662
3663 if (copy_from_user(&these, uthese, sizeof(these)))
3664 return -EFAULT;
3665
3666 if (uts) {
3667 if (get_old_timespec32(&ts, uts))
3668 return -EFAULT;
3669 }
3670
3671 ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3672
3673 if (ret > 0 && uinfo) {
3674 if (copy_siginfo_to_user(uinfo, &info))
3675 ret = -EFAULT;
3676 }
3677
3678 return ret;
3679 }
3680 #endif
3681
3682 #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)3683 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64, compat_sigset_t __user *, uthese,
3684 struct compat_siginfo __user *, uinfo,
3685 struct __kernel_timespec __user *, uts, compat_size_t, sigsetsize)
3686 {
3687 sigset_t s;
3688 struct timespec64 t;
3689 kernel_siginfo_t info;
3690 long ret;
3691
3692 if (sigsetsize != sizeof(sigset_t))
3693 return -EINVAL;
3694
3695 if (get_compat_sigset(&s, uthese))
3696 return -EFAULT;
3697
3698 if (uts) {
3699 if (get_timespec64(&t, uts))
3700 return -EFAULT;
3701 }
3702
3703 ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3704
3705 if (ret > 0 && uinfo) {
3706 if (copy_siginfo_to_user32(uinfo, &info))
3707 ret = -EFAULT;
3708 }
3709
3710 return ret;
3711 }
3712
3713 #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)3714 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32, compat_sigset_t __user *, uthese,
3715 struct compat_siginfo __user *, uinfo,
3716 struct old_timespec32 __user *, uts, compat_size_t, sigsetsize)
3717 {
3718 sigset_t s;
3719 struct timespec64 t;
3720 kernel_siginfo_t info;
3721 long ret;
3722
3723 if (sigsetsize != sizeof(sigset_t))
3724 return -EINVAL;
3725
3726 if (get_compat_sigset(&s, uthese))
3727 return -EFAULT;
3728
3729 if (uts) {
3730 if (get_old_timespec32(&t, uts))
3731 return -EFAULT;
3732 }
3733
3734 ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3735
3736 if (ret > 0 && uinfo) {
3737 if (copy_siginfo_to_user32(uinfo, &info))
3738 ret = -EFAULT;
3739 }
3740
3741 return ret;
3742 }
3743 #endif
3744 #endif
3745
prepare_kill_siginfo(int sig,struct kernel_siginfo * info)3746 static inline void prepare_kill_siginfo(int sig, struct kernel_siginfo *info)
3747 {
3748 clear_siginfo(info);
3749 info->si_signo = sig;
3750 info->si_errno = 0;
3751 info->si_code = SI_USER;
3752 info->si_pid = task_tgid_vnr(current);
3753 info->si_uid = from_kuid_munged(current_user_ns(), current_uid());
3754 }
3755
3756 /**
3757 * sys_kill - send a signal to a process
3758 * @pid: the PID of the process
3759 * @sig: signal to be sent
3760 */
SYSCALL_DEFINE2(kill,pid_t,pid,int,sig)3761 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
3762 {
3763 struct kernel_siginfo info;
3764
3765 prepare_kill_siginfo(sig, &info);
3766
3767 return kill_something_info(sig, &info, pid);
3768 }
3769
3770 /*
3771 * Verify that the signaler and signalee either are in the same pid namespace
3772 * or that the signaler's pid namespace is an ancestor of the signalee's pid
3773 * namespace.
3774 */
access_pidfd_pidns(struct pid * pid)3775 static bool access_pidfd_pidns(struct pid *pid)
3776 {
3777 struct pid_namespace *active = task_active_pid_ns(current);
3778 struct pid_namespace *p = ns_of_pid(pid);
3779
3780 for (;;) {
3781 if (!p)
3782 return false;
3783 if (p == active)
3784 break;
3785 p = p->parent;
3786 }
3787
3788 return true;
3789 }
3790
copy_siginfo_from_user_any(kernel_siginfo_t * kinfo,siginfo_t __user * info)3791 static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo,
3792 siginfo_t __user *info)
3793 {
3794 #ifdef CONFIG_COMPAT
3795 /*
3796 * Avoid hooking up compat syscalls and instead handle necessary
3797 * conversions here. Note, this is a stop-gap measure and should not be
3798 * considered a generic solution.
3799 */
3800 if (in_compat_syscall())
3801 return copy_siginfo_from_user32(
3802 kinfo, (struct compat_siginfo __user *)info);
3803 #endif
3804 return copy_siginfo_from_user(kinfo, info);
3805 }
3806
pidfd_to_pid(const struct file * file)3807 static struct pid *pidfd_to_pid(const struct file *file)
3808 {
3809 struct pid *pid;
3810
3811 pid = pidfd_pid(file);
3812 if (!IS_ERR(pid))
3813 return pid;
3814
3815 return tgid_pidfd_to_pid(file);
3816 }
3817
3818 /**
3819 * sys_pidfd_send_signal - Signal a process through a pidfd
3820 * @pidfd: file descriptor of the process
3821 * @sig: signal to send
3822 * @info: signal info
3823 * @flags: future flags
3824 *
3825 * The syscall currently only signals via PIDTYPE_PID which covers
3826 * kill(<positive-pid>, <signal>. It does not signal threads or process
3827 * groups.
3828 * In order to extend the syscall to threads and process groups the @flags
3829 * argument should be used. In essence, the @flags argument will determine
3830 * what is signaled and not the file descriptor itself. Put in other words,
3831 * grouping is a property of the flags argument not a property of the file
3832 * descriptor.
3833 *
3834 * Return: 0 on success, negative errno on failure
3835 */
SYSCALL_DEFINE4(pidfd_send_signal,int,pidfd,int,sig,siginfo_t __user *,info,unsigned int,flags)3836 SYSCALL_DEFINE4(pidfd_send_signal, int, pidfd, int, sig,
3837 siginfo_t __user *, info, unsigned int, flags)
3838 {
3839 int ret;
3840 struct fd f;
3841 struct pid *pid;
3842 kernel_siginfo_t kinfo;
3843
3844 /* Enforce flags be set to 0 until we add an extension. */
3845 if (flags)
3846 return -EINVAL;
3847
3848 f = fdget(pidfd);
3849 if (!f.file)
3850 return -EBADF;
3851
3852 /* Is this a pidfd? */
3853 pid = pidfd_to_pid(f.file);
3854 if (IS_ERR(pid)) {
3855 ret = PTR_ERR(pid);
3856 goto err;
3857 }
3858
3859 ret = -EINVAL;
3860 if (!access_pidfd_pidns(pid))
3861 goto err;
3862
3863 if (info) {
3864 ret = copy_siginfo_from_user_any(&kinfo, info);
3865 if (unlikely(ret))
3866 goto err;
3867
3868 ret = -EINVAL;
3869 if (unlikely(sig != kinfo.si_signo))
3870 goto err;
3871
3872 /* Only allow sending arbitrary signals to yourself. */
3873 ret = -EPERM;
3874 if ((task_pid(current) != pid) &&
3875 (kinfo.si_code >= 0 || kinfo.si_code == SI_TKILL))
3876 goto err;
3877 } else {
3878 prepare_kill_siginfo(sig, &kinfo);
3879 }
3880
3881 ret = kill_pid_info(sig, &kinfo, pid);
3882
3883 err:
3884 fdput(f);
3885 return ret;
3886 }
3887
3888 static int
do_send_specific(pid_t tgid,pid_t pid,int sig,struct kernel_siginfo * info)3889 do_send_specific(pid_t tgid, pid_t pid, int sig, struct kernel_siginfo *info)
3890 {
3891 struct task_struct *p;
3892 int error = -ESRCH;
3893
3894 rcu_read_lock();
3895 p = find_task_by_vpid(pid);
3896 if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
3897 error = check_kill_permission(sig, info, p);
3898 /*
3899 * The null signal is a permissions and process existence
3900 * probe. No signal is actually delivered.
3901 */
3902 if (!error && sig) {
3903 error = do_send_sig_info(sig, info, p, PIDTYPE_PID);
3904 /*
3905 * If lock_task_sighand() failed we pretend the task
3906 * dies after receiving the signal. The window is tiny,
3907 * and the signal is private anyway.
3908 */
3909 if (unlikely(error == -ESRCH))
3910 error = 0;
3911 }
3912 }
3913 rcu_read_unlock();
3914
3915 return error;
3916 }
3917
do_tkill(pid_t tgid,pid_t pid,int sig)3918 static int do_tkill(pid_t tgid, pid_t pid, int sig)
3919 {
3920 struct kernel_siginfo info;
3921
3922 clear_siginfo(&info);
3923 info.si_signo = sig;
3924 info.si_errno = 0;
3925 info.si_code = SI_TKILL;
3926 info.si_pid = task_tgid_vnr(current);
3927 info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
3928
3929 return do_send_specific(tgid, pid, sig, &info);
3930 }
3931
3932 /**
3933 * sys_tgkill - send signal to one specific thread
3934 * @tgid: the thread group ID of the thread
3935 * @pid: the PID of the thread
3936 * @sig: signal to be sent
3937 *
3938 * This syscall also checks the @tgid and returns -ESRCH even if the PID
3939 * exists but it's not belonging to the target process anymore. This
3940 * method solves the problem of threads exiting and PIDs getting reused.
3941 */
SYSCALL_DEFINE3(tgkill,pid_t,tgid,pid_t,pid,int,sig)3942 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
3943 {
3944 /* This is only valid for single tasks */
3945 if (pid <= 0 || tgid <= 0)
3946 return -EINVAL;
3947
3948 return do_tkill(tgid, pid, sig);
3949 }
3950
3951 /**
3952 * sys_tkill - send signal to one specific task
3953 * @pid: the PID of the task
3954 * @sig: signal to be sent
3955 *
3956 * Send a signal to only one task, even if it's a CLONE_THREAD task.
3957 */
SYSCALL_DEFINE2(tkill,pid_t,pid,int,sig)3958 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
3959 {
3960 /* This is only valid for single tasks */
3961 if (pid <= 0)
3962 return -EINVAL;
3963
3964 return do_tkill(0, pid, sig);
3965 }
3966
do_rt_sigqueueinfo(pid_t pid,int sig,kernel_siginfo_t * info)3967 static int do_rt_sigqueueinfo(pid_t pid, int sig, kernel_siginfo_t *info)
3968 {
3969 /* Not even root can pretend to send signals from the kernel.
3970 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3971 */
3972 if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3973 (task_pid_vnr(current) != pid))
3974 return -EPERM;
3975
3976 /* POSIX.1b doesn't mention process groups. */
3977 return kill_proc_info(sig, info, pid);
3978 }
3979
3980 /**
3981 * sys_rt_sigqueueinfo - send signal information to a signal
3982 * @pid: the PID of the thread
3983 * @sig: signal to be sent
3984 * @uinfo: signal info to be sent
3985 */
SYSCALL_DEFINE3(rt_sigqueueinfo,pid_t,pid,int,sig,siginfo_t __user *,uinfo)3986 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
3987 siginfo_t __user *, uinfo)
3988 {
3989 kernel_siginfo_t info;
3990 int ret = __copy_siginfo_from_user(sig, &info, uinfo);
3991 if (unlikely(ret))
3992 return ret;
3993 return do_rt_sigqueueinfo(pid, sig, &info);
3994 }
3995
3996 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,compat_pid_t,pid,int,sig,struct compat_siginfo __user *,uinfo)3997 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
3998 compat_pid_t, pid,
3999 int, sig,
4000 struct compat_siginfo __user *, uinfo)
4001 {
4002 kernel_siginfo_t info;
4003 int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4004 if (unlikely(ret))
4005 return ret;
4006 return do_rt_sigqueueinfo(pid, sig, &info);
4007 }
4008 #endif
4009
do_rt_tgsigqueueinfo(pid_t tgid,pid_t pid,int sig,kernel_siginfo_t * info)4010 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, kernel_siginfo_t *info)
4011 {
4012 /* This is only valid for single tasks */
4013 if (pid <= 0 || tgid <= 0)
4014 return -EINVAL;
4015
4016 /* Not even root can pretend to send signals from the kernel.
4017 * Nor can they impersonate a kill()/tgkill(), which adds source info.
4018 */
4019 if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
4020 (task_pid_vnr(current) != pid))
4021 return -EPERM;
4022
4023 return do_send_specific(tgid, pid, sig, info);
4024 }
4025
SYSCALL_DEFINE4(rt_tgsigqueueinfo,pid_t,tgid,pid_t,pid,int,sig,siginfo_t __user *,uinfo)4026 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
4027 siginfo_t __user *, uinfo)
4028 {
4029 kernel_siginfo_t info;
4030 int ret = __copy_siginfo_from_user(sig, &info, uinfo);
4031 if (unlikely(ret))
4032 return ret;
4033 return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4034 }
4035
4036 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,compat_pid_t,tgid,compat_pid_t,pid,int,sig,struct compat_siginfo __user *,uinfo)4037 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
4038 compat_pid_t, tgid,
4039 compat_pid_t, pid,
4040 int, sig,
4041 struct compat_siginfo __user *, uinfo)
4042 {
4043 kernel_siginfo_t info;
4044 int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
4045 if (unlikely(ret))
4046 return ret;
4047 return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
4048 }
4049 #endif
4050
4051 /*
4052 * For kthreads only, must not be used if cloned with CLONE_SIGHAND
4053 */
kernel_sigaction(int sig,__sighandler_t action)4054 void kernel_sigaction(int sig, __sighandler_t action)
4055 {
4056 spin_lock_irq(¤t->sighand->siglock);
4057 current->sighand->action[sig - 1].sa.sa_handler = action;
4058 if (action == SIG_IGN) {
4059 sigset_t mask;
4060
4061 sigemptyset(&mask);
4062 sigaddset(&mask, sig);
4063
4064 flush_sigqueue_mask(&mask, ¤t->signal->shared_pending);
4065 flush_sigqueue_mask(&mask, ¤t->pending);
4066 recalc_sigpending();
4067 }
4068 spin_unlock_irq(¤t->sighand->siglock);
4069 }
4070 EXPORT_SYMBOL(kernel_sigaction);
4071
sigaction_compat_abi(struct k_sigaction * act,struct k_sigaction * oact)4072 void __weak sigaction_compat_abi(struct k_sigaction *act,
4073 struct k_sigaction *oact)
4074 {
4075 }
4076
do_sigaction(int sig,struct k_sigaction * act,struct k_sigaction * oact)4077 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
4078 {
4079 struct task_struct *p = current, *t;
4080 struct k_sigaction *k;
4081 sigset_t mask;
4082
4083 if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
4084 return -EINVAL;
4085
4086 k = &p->sighand->action[sig-1];
4087
4088 spin_lock_irq(&p->sighand->siglock);
4089 if (k->sa.sa_flags & SA_IMMUTABLE) {
4090 spin_unlock_irq(&p->sighand->siglock);
4091 return -EINVAL;
4092 }
4093 if (oact)
4094 *oact = *k;
4095
4096 /*
4097 * Make sure that we never accidentally claim to support SA_UNSUPPORTED,
4098 * e.g. by having an architecture use the bit in their uapi.
4099 */
4100 BUILD_BUG_ON(UAPI_SA_FLAGS & SA_UNSUPPORTED);
4101
4102 /*
4103 * Clear unknown flag bits in order to allow userspace to detect missing
4104 * support for flag bits and to allow the kernel to use non-uapi bits
4105 * internally.
4106 */
4107 if (act)
4108 act->sa.sa_flags &= UAPI_SA_FLAGS;
4109 if (oact)
4110 oact->sa.sa_flags &= UAPI_SA_FLAGS;
4111
4112 sigaction_compat_abi(act, oact);
4113
4114 if (act) {
4115 sigdelsetmask(&act->sa.sa_mask,
4116 sigmask(SIGKILL) | sigmask(SIGSTOP));
4117 *k = *act;
4118 /*
4119 * POSIX 3.3.1.3:
4120 * "Setting a signal action to SIG_IGN for a signal that is
4121 * pending shall cause the pending signal to be discarded,
4122 * whether or not it is blocked."
4123 *
4124 * "Setting a signal action to SIG_DFL for a signal that is
4125 * pending and whose default action is to ignore the signal
4126 * (for example, SIGCHLD), shall cause the pending signal to
4127 * be discarded, whether or not it is blocked"
4128 */
4129 if (sig_handler_ignored(sig_handler(p, sig), sig)) {
4130 sigemptyset(&mask);
4131 sigaddset(&mask, sig);
4132 flush_sigqueue_mask(&mask, &p->signal->shared_pending);
4133 for_each_thread(p, t)
4134 flush_sigqueue_mask(&mask, &t->pending);
4135 }
4136 }
4137
4138 spin_unlock_irq(&p->sighand->siglock);
4139 return 0;
4140 }
4141
4142 #ifdef CONFIG_DYNAMIC_SIGFRAME
sigaltstack_lock(void)4143 static inline void sigaltstack_lock(void)
4144 __acquires(¤t->sighand->siglock)
4145 {
4146 spin_lock_irq(¤t->sighand->siglock);
4147 }
4148
sigaltstack_unlock(void)4149 static inline void sigaltstack_unlock(void)
4150 __releases(¤t->sighand->siglock)
4151 {
4152 spin_unlock_irq(¤t->sighand->siglock);
4153 }
4154 #else
sigaltstack_lock(void)4155 static inline void sigaltstack_lock(void) { }
sigaltstack_unlock(void)4156 static inline void sigaltstack_unlock(void) { }
4157 #endif
4158
4159 static int
do_sigaltstack(const stack_t * ss,stack_t * oss,unsigned long sp,size_t min_ss_size)4160 do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
4161 size_t min_ss_size)
4162 {
4163 struct task_struct *t = current;
4164 int ret = 0;
4165
4166 if (oss) {
4167 memset(oss, 0, sizeof(stack_t));
4168 oss->ss_sp = (void __user *) t->sas_ss_sp;
4169 oss->ss_size = t->sas_ss_size;
4170 oss->ss_flags = sas_ss_flags(sp) |
4171 (current->sas_ss_flags & SS_FLAG_BITS);
4172 }
4173
4174 if (ss) {
4175 void __user *ss_sp = ss->ss_sp;
4176 size_t ss_size = ss->ss_size;
4177 unsigned ss_flags = ss->ss_flags;
4178 int ss_mode;
4179
4180 if (unlikely(on_sig_stack(sp)))
4181 return -EPERM;
4182
4183 ss_mode = ss_flags & ~SS_FLAG_BITS;
4184 if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
4185 ss_mode != 0))
4186 return -EINVAL;
4187
4188 /*
4189 * Return before taking any locks if no actual
4190 * sigaltstack changes were requested.
4191 */
4192 if (t->sas_ss_sp == (unsigned long)ss_sp &&
4193 t->sas_ss_size == ss_size &&
4194 t->sas_ss_flags == ss_flags)
4195 return 0;
4196
4197 sigaltstack_lock();
4198 if (ss_mode == SS_DISABLE) {
4199 ss_size = 0;
4200 ss_sp = NULL;
4201 } else {
4202 if (unlikely(ss_size < min_ss_size))
4203 ret = -ENOMEM;
4204 if (!sigaltstack_size_valid(ss_size))
4205 ret = -ENOMEM;
4206 }
4207 if (!ret) {
4208 t->sas_ss_sp = (unsigned long) ss_sp;
4209 t->sas_ss_size = ss_size;
4210 t->sas_ss_flags = ss_flags;
4211 }
4212 sigaltstack_unlock();
4213 }
4214 return ret;
4215 }
4216
SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss,stack_t __user *,uoss)4217 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
4218 {
4219 stack_t new, old;
4220 int err;
4221 if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
4222 return -EFAULT;
4223 err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
4224 current_user_stack_pointer(),
4225 MINSIGSTKSZ);
4226 if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
4227 err = -EFAULT;
4228 return err;
4229 }
4230
restore_altstack(const stack_t __user * uss)4231 int restore_altstack(const stack_t __user *uss)
4232 {
4233 stack_t new;
4234 if (copy_from_user(&new, uss, sizeof(stack_t)))
4235 return -EFAULT;
4236 (void)do_sigaltstack(&new, NULL, current_user_stack_pointer(),
4237 MINSIGSTKSZ);
4238 /* squash all but EFAULT for now */
4239 return 0;
4240 }
4241
__save_altstack(stack_t __user * uss,unsigned long sp)4242 int __save_altstack(stack_t __user *uss, unsigned long sp)
4243 {
4244 struct task_struct *t = current;
4245 int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
4246 __put_user(t->sas_ss_flags, &uss->ss_flags) |
4247 __put_user(t->sas_ss_size, &uss->ss_size);
4248 return err;
4249 }
4250
4251 #ifdef CONFIG_COMPAT
do_compat_sigaltstack(const compat_stack_t __user * uss_ptr,compat_stack_t __user * uoss_ptr)4252 static int do_compat_sigaltstack(const compat_stack_t __user *uss_ptr,
4253 compat_stack_t __user *uoss_ptr)
4254 {
4255 stack_t uss, uoss;
4256 int ret;
4257
4258 if (uss_ptr) {
4259 compat_stack_t uss32;
4260 if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
4261 return -EFAULT;
4262 uss.ss_sp = compat_ptr(uss32.ss_sp);
4263 uss.ss_flags = uss32.ss_flags;
4264 uss.ss_size = uss32.ss_size;
4265 }
4266 ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
4267 compat_user_stack_pointer(),
4268 COMPAT_MINSIGSTKSZ);
4269 if (ret >= 0 && uoss_ptr) {
4270 compat_stack_t old;
4271 memset(&old, 0, sizeof(old));
4272 old.ss_sp = ptr_to_compat(uoss.ss_sp);
4273 old.ss_flags = uoss.ss_flags;
4274 old.ss_size = uoss.ss_size;
4275 if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
4276 ret = -EFAULT;
4277 }
4278 return ret;
4279 }
4280
COMPAT_SYSCALL_DEFINE2(sigaltstack,const compat_stack_t __user *,uss_ptr,compat_stack_t __user *,uoss_ptr)4281 COMPAT_SYSCALL_DEFINE2(sigaltstack,
4282 const compat_stack_t __user *, uss_ptr,
4283 compat_stack_t __user *, uoss_ptr)
4284 {
4285 return do_compat_sigaltstack(uss_ptr, uoss_ptr);
4286 }
4287
compat_restore_altstack(const compat_stack_t __user * uss)4288 int compat_restore_altstack(const compat_stack_t __user *uss)
4289 {
4290 int err = do_compat_sigaltstack(uss, NULL);
4291 /* squash all but -EFAULT for now */
4292 return err == -EFAULT ? err : 0;
4293 }
4294
__compat_save_altstack(compat_stack_t __user * uss,unsigned long sp)4295 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
4296 {
4297 int err;
4298 struct task_struct *t = current;
4299 err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
4300 &uss->ss_sp) |
4301 __put_user(t->sas_ss_flags, &uss->ss_flags) |
4302 __put_user(t->sas_ss_size, &uss->ss_size);
4303 return err;
4304 }
4305 #endif
4306
4307 #ifdef __ARCH_WANT_SYS_SIGPENDING
4308
4309 /**
4310 * sys_sigpending - examine pending signals
4311 * @uset: where mask of pending signal is returned
4312 */
SYSCALL_DEFINE1(sigpending,old_sigset_t __user *,uset)4313 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, uset)
4314 {
4315 sigset_t set;
4316
4317 if (sizeof(old_sigset_t) > sizeof(*uset))
4318 return -EINVAL;
4319
4320 do_sigpending(&set);
4321
4322 if (copy_to_user(uset, &set, sizeof(old_sigset_t)))
4323 return -EFAULT;
4324
4325 return 0;
4326 }
4327
4328 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE1(sigpending,compat_old_sigset_t __user *,set32)4329 COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
4330 {
4331 sigset_t set;
4332
4333 do_sigpending(&set);
4334
4335 return put_user(set.sig[0], set32);
4336 }
4337 #endif
4338
4339 #endif
4340
4341 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
4342 /**
4343 * sys_sigprocmask - examine and change blocked signals
4344 * @how: whether to add, remove, or set signals
4345 * @nset: signals to add or remove (if non-null)
4346 * @oset: previous value of signal mask if non-null
4347 *
4348 * Some platforms have their own version with special arguments;
4349 * others support only sys_rt_sigprocmask.
4350 */
4351
SYSCALL_DEFINE3(sigprocmask,int,how,old_sigset_t __user *,nset,old_sigset_t __user *,oset)4352 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
4353 old_sigset_t __user *, oset)
4354 {
4355 old_sigset_t old_set, new_set;
4356 sigset_t new_blocked;
4357
4358 old_set = current->blocked.sig[0];
4359
4360 if (nset) {
4361 if (copy_from_user(&new_set, nset, sizeof(*nset)))
4362 return -EFAULT;
4363
4364 new_blocked = current->blocked;
4365
4366 switch (how) {
4367 case SIG_BLOCK:
4368 sigaddsetmask(&new_blocked, new_set);
4369 break;
4370 case SIG_UNBLOCK:
4371 sigdelsetmask(&new_blocked, new_set);
4372 break;
4373 case SIG_SETMASK:
4374 new_blocked.sig[0] = new_set;
4375 break;
4376 default:
4377 return -EINVAL;
4378 }
4379
4380 set_current_blocked(&new_blocked);
4381 }
4382
4383 if (oset) {
4384 if (copy_to_user(oset, &old_set, sizeof(*oset)))
4385 return -EFAULT;
4386 }
4387
4388 return 0;
4389 }
4390 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
4391
4392 #ifndef CONFIG_ODD_RT_SIGACTION
4393 /**
4394 * sys_rt_sigaction - alter an action taken by a process
4395 * @sig: signal to be sent
4396 * @act: new sigaction
4397 * @oact: used to save the previous sigaction
4398 * @sigsetsize: size of sigset_t type
4399 */
SYSCALL_DEFINE4(rt_sigaction,int,sig,const struct sigaction __user *,act,struct sigaction __user *,oact,size_t,sigsetsize)4400 SYSCALL_DEFINE4(rt_sigaction, int, sig,
4401 const struct sigaction __user *, act,
4402 struct sigaction __user *, oact,
4403 size_t, sigsetsize)
4404 {
4405 struct k_sigaction new_sa, old_sa;
4406 int ret;
4407
4408 /* XXX: Don't preclude handling different sized sigset_t's. */
4409 if (sigsetsize != sizeof(sigset_t))
4410 return -EINVAL;
4411
4412 if (act && copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
4413 return -EFAULT;
4414
4415 ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
4416 if (ret)
4417 return ret;
4418
4419 if (oact && copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
4420 return -EFAULT;
4421
4422 return 0;
4423 }
4424 #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)4425 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
4426 const struct compat_sigaction __user *, act,
4427 struct compat_sigaction __user *, oact,
4428 compat_size_t, sigsetsize)
4429 {
4430 struct k_sigaction new_ka, old_ka;
4431 #ifdef __ARCH_HAS_SA_RESTORER
4432 compat_uptr_t restorer;
4433 #endif
4434 int ret;
4435
4436 /* XXX: Don't preclude handling different sized sigset_t's. */
4437 if (sigsetsize != sizeof(compat_sigset_t))
4438 return -EINVAL;
4439
4440 if (act) {
4441 compat_uptr_t handler;
4442 ret = get_user(handler, &act->sa_handler);
4443 new_ka.sa.sa_handler = compat_ptr(handler);
4444 #ifdef __ARCH_HAS_SA_RESTORER
4445 ret |= get_user(restorer, &act->sa_restorer);
4446 new_ka.sa.sa_restorer = compat_ptr(restorer);
4447 #endif
4448 ret |= get_compat_sigset(&new_ka.sa.sa_mask, &act->sa_mask);
4449 ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
4450 if (ret)
4451 return -EFAULT;
4452 }
4453
4454 ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4455 if (!ret && oact) {
4456 ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
4457 &oact->sa_handler);
4458 ret |= put_compat_sigset(&oact->sa_mask, &old_ka.sa.sa_mask,
4459 sizeof(oact->sa_mask));
4460 ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
4461 #ifdef __ARCH_HAS_SA_RESTORER
4462 ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4463 &oact->sa_restorer);
4464 #endif
4465 }
4466 return ret;
4467 }
4468 #endif
4469 #endif /* !CONFIG_ODD_RT_SIGACTION */
4470
4471 #ifdef CONFIG_OLD_SIGACTION
SYSCALL_DEFINE3(sigaction,int,sig,const struct old_sigaction __user *,act,struct old_sigaction __user *,oact)4472 SYSCALL_DEFINE3(sigaction, int, sig,
4473 const struct old_sigaction __user *, act,
4474 struct old_sigaction __user *, oact)
4475 {
4476 struct k_sigaction new_ka, old_ka;
4477 int ret;
4478
4479 if (act) {
4480 old_sigset_t mask;
4481 if (!access_ok(act, sizeof(*act)) ||
4482 __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
4483 __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
4484 __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4485 __get_user(mask, &act->sa_mask))
4486 return -EFAULT;
4487 #ifdef __ARCH_HAS_KA_RESTORER
4488 new_ka.ka_restorer = NULL;
4489 #endif
4490 siginitset(&new_ka.sa.sa_mask, mask);
4491 }
4492
4493 ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4494
4495 if (!ret && oact) {
4496 if (!access_ok(oact, sizeof(*oact)) ||
4497 __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
4498 __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
4499 __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4500 __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4501 return -EFAULT;
4502 }
4503
4504 return ret;
4505 }
4506 #endif
4507 #ifdef CONFIG_COMPAT_OLD_SIGACTION
COMPAT_SYSCALL_DEFINE3(sigaction,int,sig,const struct compat_old_sigaction __user *,act,struct compat_old_sigaction __user *,oact)4508 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
4509 const struct compat_old_sigaction __user *, act,
4510 struct compat_old_sigaction __user *, oact)
4511 {
4512 struct k_sigaction new_ka, old_ka;
4513 int ret;
4514 compat_old_sigset_t mask;
4515 compat_uptr_t handler, restorer;
4516
4517 if (act) {
4518 if (!access_ok(act, sizeof(*act)) ||
4519 __get_user(handler, &act->sa_handler) ||
4520 __get_user(restorer, &act->sa_restorer) ||
4521 __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4522 __get_user(mask, &act->sa_mask))
4523 return -EFAULT;
4524
4525 #ifdef __ARCH_HAS_KA_RESTORER
4526 new_ka.ka_restorer = NULL;
4527 #endif
4528 new_ka.sa.sa_handler = compat_ptr(handler);
4529 new_ka.sa.sa_restorer = compat_ptr(restorer);
4530 siginitset(&new_ka.sa.sa_mask, mask);
4531 }
4532
4533 ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4534
4535 if (!ret && oact) {
4536 if (!access_ok(oact, sizeof(*oact)) ||
4537 __put_user(ptr_to_compat(old_ka.sa.sa_handler),
4538 &oact->sa_handler) ||
4539 __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4540 &oact->sa_restorer) ||
4541 __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4542 __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4543 return -EFAULT;
4544 }
4545 return ret;
4546 }
4547 #endif
4548
4549 #ifdef CONFIG_SGETMASK_SYSCALL
4550
4551 /*
4552 * For backwards compatibility. Functionality superseded by sigprocmask.
4553 */
SYSCALL_DEFINE0(sgetmask)4554 SYSCALL_DEFINE0(sgetmask)
4555 {
4556 /* SMP safe */
4557 return current->blocked.sig[0];
4558 }
4559
SYSCALL_DEFINE1(ssetmask,int,newmask)4560 SYSCALL_DEFINE1(ssetmask, int, newmask)
4561 {
4562 int old = current->blocked.sig[0];
4563 sigset_t newset;
4564
4565 siginitset(&newset, newmask);
4566 set_current_blocked(&newset);
4567
4568 return old;
4569 }
4570 #endif /* CONFIG_SGETMASK_SYSCALL */
4571
4572 #ifdef __ARCH_WANT_SYS_SIGNAL
4573 /*
4574 * For backwards compatibility. Functionality superseded by sigaction.
4575 */
SYSCALL_DEFINE2(signal,int,sig,__sighandler_t,handler)4576 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
4577 {
4578 struct k_sigaction new_sa, old_sa;
4579 int ret;
4580
4581 new_sa.sa.sa_handler = handler;
4582 new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
4583 sigemptyset(&new_sa.sa.sa_mask);
4584
4585 ret = do_sigaction(sig, &new_sa, &old_sa);
4586
4587 return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
4588 }
4589 #endif /* __ARCH_WANT_SYS_SIGNAL */
4590
4591 #ifdef __ARCH_WANT_SYS_PAUSE
4592
SYSCALL_DEFINE0(pause)4593 SYSCALL_DEFINE0(pause)
4594 {
4595 while (!signal_pending(current)) {
4596 __set_current_state(TASK_INTERRUPTIBLE);
4597 schedule();
4598 }
4599 return -ERESTARTNOHAND;
4600 }
4601
4602 #endif
4603
sigsuspend(sigset_t * set)4604 static int sigsuspend(sigset_t *set)
4605 {
4606 current->saved_sigmask = current->blocked;
4607 set_current_blocked(set);
4608
4609 while (!signal_pending(current)) {
4610 __set_current_state(TASK_INTERRUPTIBLE);
4611 schedule();
4612 }
4613 set_restore_sigmask();
4614 return -ERESTARTNOHAND;
4615 }
4616
4617 /**
4618 * sys_rt_sigsuspend - replace the signal mask for a value with the
4619 * @unewset value until a signal is received
4620 * @unewset: new signal mask value
4621 * @sigsetsize: size of sigset_t type
4622 */
SYSCALL_DEFINE2(rt_sigsuspend,sigset_t __user *,unewset,size_t,sigsetsize)4623 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
4624 {
4625 sigset_t newset;
4626
4627 /* XXX: Don't preclude handling different sized sigset_t's. */
4628 if (sigsetsize != sizeof(sigset_t))
4629 return -EINVAL;
4630
4631 if (copy_from_user(&newset, unewset, sizeof(newset)))
4632 return -EFAULT;
4633 return sigsuspend(&newset);
4634 }
4635
4636 #ifdef CONFIG_COMPAT
COMPAT_SYSCALL_DEFINE2(rt_sigsuspend,compat_sigset_t __user *,unewset,compat_size_t,sigsetsize)4637 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
4638 {
4639 sigset_t newset;
4640
4641 /* XXX: Don't preclude handling different sized sigset_t's. */
4642 if (sigsetsize != sizeof(sigset_t))
4643 return -EINVAL;
4644
4645 if (get_compat_sigset(&newset, unewset))
4646 return -EFAULT;
4647 return sigsuspend(&newset);
4648 }
4649 #endif
4650
4651 #ifdef CONFIG_OLD_SIGSUSPEND
SYSCALL_DEFINE1(sigsuspend,old_sigset_t,mask)4652 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
4653 {
4654 sigset_t blocked;
4655 siginitset(&blocked, mask);
4656 return sigsuspend(&blocked);
4657 }
4658 #endif
4659 #ifdef CONFIG_OLD_SIGSUSPEND3
SYSCALL_DEFINE3(sigsuspend,int,unused1,int,unused2,old_sigset_t,mask)4660 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
4661 {
4662 sigset_t blocked;
4663 siginitset(&blocked, mask);
4664 return sigsuspend(&blocked);
4665 }
4666 #endif
4667
arch_vma_name(struct vm_area_struct * vma)4668 __weak const char *arch_vma_name(struct vm_area_struct *vma)
4669 {
4670 return NULL;
4671 }
4672
siginfo_buildtime_checks(void)4673 static inline void siginfo_buildtime_checks(void)
4674 {
4675 BUILD_BUG_ON(sizeof(struct siginfo) != SI_MAX_SIZE);
4676
4677 /* Verify the offsets in the two siginfos match */
4678 #define CHECK_OFFSET(field) \
4679 BUILD_BUG_ON(offsetof(siginfo_t, field) != offsetof(kernel_siginfo_t, field))
4680
4681 /* kill */
4682 CHECK_OFFSET(si_pid);
4683 CHECK_OFFSET(si_uid);
4684
4685 /* timer */
4686 CHECK_OFFSET(si_tid);
4687 CHECK_OFFSET(si_overrun);
4688 CHECK_OFFSET(si_value);
4689
4690 /* rt */
4691 CHECK_OFFSET(si_pid);
4692 CHECK_OFFSET(si_uid);
4693 CHECK_OFFSET(si_value);
4694
4695 /* sigchld */
4696 CHECK_OFFSET(si_pid);
4697 CHECK_OFFSET(si_uid);
4698 CHECK_OFFSET(si_status);
4699 CHECK_OFFSET(si_utime);
4700 CHECK_OFFSET(si_stime);
4701
4702 /* sigfault */
4703 CHECK_OFFSET(si_addr);
4704 CHECK_OFFSET(si_trapno);
4705 CHECK_OFFSET(si_addr_lsb);
4706 CHECK_OFFSET(si_lower);
4707 CHECK_OFFSET(si_upper);
4708 CHECK_OFFSET(si_pkey);
4709 CHECK_OFFSET(si_perf_data);
4710 CHECK_OFFSET(si_perf_type);
4711
4712 /* sigpoll */
4713 CHECK_OFFSET(si_band);
4714 CHECK_OFFSET(si_fd);
4715
4716 /* sigsys */
4717 CHECK_OFFSET(si_call_addr);
4718 CHECK_OFFSET(si_syscall);
4719 CHECK_OFFSET(si_arch);
4720 #undef CHECK_OFFSET
4721
4722 /* usb asyncio */
4723 BUILD_BUG_ON(offsetof(struct siginfo, si_pid) !=
4724 offsetof(struct siginfo, si_addr));
4725 if (sizeof(int) == sizeof(void __user *)) {
4726 BUILD_BUG_ON(sizeof_field(struct siginfo, si_pid) !=
4727 sizeof(void __user *));
4728 } else {
4729 BUILD_BUG_ON((sizeof_field(struct siginfo, si_pid) +
4730 sizeof_field(struct siginfo, si_uid)) !=
4731 sizeof(void __user *));
4732 BUILD_BUG_ON(offsetofend(struct siginfo, si_pid) !=
4733 offsetof(struct siginfo, si_uid));
4734 }
4735 #ifdef CONFIG_COMPAT
4736 BUILD_BUG_ON(offsetof(struct compat_siginfo, si_pid) !=
4737 offsetof(struct compat_siginfo, si_addr));
4738 BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4739 sizeof(compat_uptr_t));
4740 BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4741 sizeof_field(struct siginfo, si_pid));
4742 #endif
4743 }
4744
signals_init(void)4745 void __init signals_init(void)
4746 {
4747 siginfo_buildtime_checks();
4748
4749 sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC | SLAB_ACCOUNT);
4750 }
4751
4752 #ifdef CONFIG_KGDB_KDB
4753 #include <linux/kdb.h>
4754 /*
4755 * kdb_send_sig - Allows kdb to send signals without exposing
4756 * signal internals. This function checks if the required locks are
4757 * available before calling the main signal code, to avoid kdb
4758 * deadlocks.
4759 */
kdb_send_sig(struct task_struct * t,int sig)4760 void kdb_send_sig(struct task_struct *t, int sig)
4761 {
4762 static struct task_struct *kdb_prev_t;
4763 int new_t, ret;
4764 if (!spin_trylock(&t->sighand->siglock)) {
4765 kdb_printf("Can't do kill command now.\n"
4766 "The sigmask lock is held somewhere else in "
4767 "kernel, try again later\n");
4768 return;
4769 }
4770 new_t = kdb_prev_t != t;
4771 kdb_prev_t = t;
4772 if (!task_is_running(t) && new_t) {
4773 spin_unlock(&t->sighand->siglock);
4774 kdb_printf("Process is not RUNNING, sending a signal from "
4775 "kdb risks deadlock\n"
4776 "on the run queue locks. "
4777 "The signal has _not_ been sent.\n"
4778 "Reissue the kill command if you want to risk "
4779 "the deadlock.\n");
4780 return;
4781 }
4782 ret = send_signal(sig, SEND_SIG_PRIV, t, PIDTYPE_PID);
4783 spin_unlock(&t->sighand->siglock);
4784 if (ret)
4785 kdb_printf("Fail to deliver Signal %d to process %d.\n",
4786 sig, t->pid);
4787 else
4788 kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
4789 }
4790 #endif /* CONFIG_KGDB_KDB */
4791