1/*
2 *  linux/kernel/compat.c
3 *
4 *  Kernel compatibililty routines for e.g. 32 bit syscall support
5 *  on 64 bit kernels.
6 *
7 *  Copyright (C) 2002-2003 Stephen Rothwell, IBM Corporation
8 *
9 *  This program is free software; you can redistribute it and/or modify
10 *  it under the terms of the GNU General Public License version 2 as
11 *  published by the Free Software Foundation.
12 */
13
14#include <linux/linkage.h>
15#include <linux/compat.h>
16#include <linux/errno.h>
17#include <linux/time.h>
18#include <linux/signal.h>
19#include <linux/sched.h>	/* for MAX_SCHEDULE_TIMEOUT */
20#include <linux/syscalls.h>
21#include <linux/unistd.h>
22#include <linux/security.h>
23#include <linux/timex.h>
24#include <linux/export.h>
25#include <linux/migrate.h>
26#include <linux/posix-timers.h>
27#include <linux/times.h>
28#include <linux/ptrace.h>
29#include <linux/gfp.h>
30
31#include <asm/uaccess.h>
32
33static int compat_get_timex(struct timex *txc, struct compat_timex __user *utp)
34{
35	memset(txc, 0, sizeof(struct timex));
36
37	if (!access_ok(VERIFY_READ, utp, sizeof(struct compat_timex)) ||
38			__get_user(txc->modes, &utp->modes) ||
39			__get_user(txc->offset, &utp->offset) ||
40			__get_user(txc->freq, &utp->freq) ||
41			__get_user(txc->maxerror, &utp->maxerror) ||
42			__get_user(txc->esterror, &utp->esterror) ||
43			__get_user(txc->status, &utp->status) ||
44			__get_user(txc->constant, &utp->constant) ||
45			__get_user(txc->precision, &utp->precision) ||
46			__get_user(txc->tolerance, &utp->tolerance) ||
47			__get_user(txc->time.tv_sec, &utp->time.tv_sec) ||
48			__get_user(txc->time.tv_usec, &utp->time.tv_usec) ||
49			__get_user(txc->tick, &utp->tick) ||
50			__get_user(txc->ppsfreq, &utp->ppsfreq) ||
51			__get_user(txc->jitter, &utp->jitter) ||
52			__get_user(txc->shift, &utp->shift) ||
53			__get_user(txc->stabil, &utp->stabil) ||
54			__get_user(txc->jitcnt, &utp->jitcnt) ||
55			__get_user(txc->calcnt, &utp->calcnt) ||
56			__get_user(txc->errcnt, &utp->errcnt) ||
57			__get_user(txc->stbcnt, &utp->stbcnt))
58		return -EFAULT;
59
60	return 0;
61}
62
63static int compat_put_timex(struct compat_timex __user *utp, struct timex *txc)
64{
65	if (!access_ok(VERIFY_WRITE, utp, sizeof(struct compat_timex)) ||
66			__put_user(txc->modes, &utp->modes) ||
67			__put_user(txc->offset, &utp->offset) ||
68			__put_user(txc->freq, &utp->freq) ||
69			__put_user(txc->maxerror, &utp->maxerror) ||
70			__put_user(txc->esterror, &utp->esterror) ||
71			__put_user(txc->status, &utp->status) ||
72			__put_user(txc->constant, &utp->constant) ||
73			__put_user(txc->precision, &utp->precision) ||
74			__put_user(txc->tolerance, &utp->tolerance) ||
75			__put_user(txc->time.tv_sec, &utp->time.tv_sec) ||
76			__put_user(txc->time.tv_usec, &utp->time.tv_usec) ||
77			__put_user(txc->tick, &utp->tick) ||
78			__put_user(txc->ppsfreq, &utp->ppsfreq) ||
79			__put_user(txc->jitter, &utp->jitter) ||
80			__put_user(txc->shift, &utp->shift) ||
81			__put_user(txc->stabil, &utp->stabil) ||
82			__put_user(txc->jitcnt, &utp->jitcnt) ||
83			__put_user(txc->calcnt, &utp->calcnt) ||
84			__put_user(txc->errcnt, &utp->errcnt) ||
85			__put_user(txc->stbcnt, &utp->stbcnt) ||
86			__put_user(txc->tai, &utp->tai))
87		return -EFAULT;
88	return 0;
89}
90
91COMPAT_SYSCALL_DEFINE2(gettimeofday, struct compat_timeval __user *, tv,
92		       struct timezone __user *, tz)
93{
94	if (tv) {
95		struct timeval ktv;
96		do_gettimeofday(&ktv);
97		if (compat_put_timeval(&ktv, tv))
98			return -EFAULT;
99	}
100	if (tz) {
101		if (copy_to_user(tz, &sys_tz, sizeof(sys_tz)))
102			return -EFAULT;
103	}
104
105	return 0;
106}
107
108COMPAT_SYSCALL_DEFINE2(settimeofday, struct compat_timeval __user *, tv,
109		       struct timezone __user *, tz)
110{
111	struct timeval user_tv;
112	struct timespec	new_ts;
113	struct timezone new_tz;
114
115	if (tv) {
116		if (compat_get_timeval(&user_tv, tv))
117			return -EFAULT;
118		new_ts.tv_sec = user_tv.tv_sec;
119		new_ts.tv_nsec = user_tv.tv_usec * NSEC_PER_USEC;
120	}
121	if (tz) {
122		if (copy_from_user(&new_tz, tz, sizeof(*tz)))
123			return -EFAULT;
124	}
125
126	return do_sys_settimeofday(tv ? &new_ts : NULL, tz ? &new_tz : NULL);
127}
128
129static int __compat_get_timeval(struct timeval *tv, const struct compat_timeval __user *ctv)
130{
131	return (!access_ok(VERIFY_READ, ctv, sizeof(*ctv)) ||
132			__get_user(tv->tv_sec, &ctv->tv_sec) ||
133			__get_user(tv->tv_usec, &ctv->tv_usec)) ? -EFAULT : 0;
134}
135
136static int __compat_put_timeval(const struct timeval *tv, struct compat_timeval __user *ctv)
137{
138	return (!access_ok(VERIFY_WRITE, ctv, sizeof(*ctv)) ||
139			__put_user(tv->tv_sec, &ctv->tv_sec) ||
140			__put_user(tv->tv_usec, &ctv->tv_usec)) ? -EFAULT : 0;
141}
142
143static int __compat_get_timespec(struct timespec *ts, const struct compat_timespec __user *cts)
144{
145	return (!access_ok(VERIFY_READ, cts, sizeof(*cts)) ||
146			__get_user(ts->tv_sec, &cts->tv_sec) ||
147			__get_user(ts->tv_nsec, &cts->tv_nsec)) ? -EFAULT : 0;
148}
149
150static int __compat_put_timespec(const struct timespec *ts, struct compat_timespec __user *cts)
151{
152	return (!access_ok(VERIFY_WRITE, cts, sizeof(*cts)) ||
153			__put_user(ts->tv_sec, &cts->tv_sec) ||
154			__put_user(ts->tv_nsec, &cts->tv_nsec)) ? -EFAULT : 0;
155}
156
157int compat_get_timeval(struct timeval *tv, const void __user *utv)
158{
159	if (COMPAT_USE_64BIT_TIME)
160		return copy_from_user(tv, utv, sizeof(*tv)) ? -EFAULT : 0;
161	else
162		return __compat_get_timeval(tv, utv);
163}
164EXPORT_SYMBOL_GPL(compat_get_timeval);
165
166int compat_put_timeval(const struct timeval *tv, void __user *utv)
167{
168	if (COMPAT_USE_64BIT_TIME)
169		return copy_to_user(utv, tv, sizeof(*tv)) ? -EFAULT : 0;
170	else
171		return __compat_put_timeval(tv, utv);
172}
173EXPORT_SYMBOL_GPL(compat_put_timeval);
174
175int compat_get_timespec(struct timespec *ts, const void __user *uts)
176{
177	if (COMPAT_USE_64BIT_TIME)
178		return copy_from_user(ts, uts, sizeof(*ts)) ? -EFAULT : 0;
179	else
180		return __compat_get_timespec(ts, uts);
181}
182EXPORT_SYMBOL_GPL(compat_get_timespec);
183
184int compat_put_timespec(const struct timespec *ts, void __user *uts)
185{
186	if (COMPAT_USE_64BIT_TIME)
187		return copy_to_user(uts, ts, sizeof(*ts)) ? -EFAULT : 0;
188	else
189		return __compat_put_timespec(ts, uts);
190}
191EXPORT_SYMBOL_GPL(compat_put_timespec);
192
193int compat_convert_timespec(struct timespec __user **kts,
194			    const void __user *cts)
195{
196	struct timespec ts;
197	struct timespec __user *uts;
198
199	if (!cts || COMPAT_USE_64BIT_TIME) {
200		*kts = (struct timespec __user *)cts;
201		return 0;
202	}
203
204	uts = compat_alloc_user_space(sizeof(ts));
205	if (!uts)
206		return -EFAULT;
207	if (compat_get_timespec(&ts, cts))
208		return -EFAULT;
209	if (copy_to_user(uts, &ts, sizeof(ts)))
210		return -EFAULT;
211
212	*kts = uts;
213	return 0;
214}
215
216static long compat_nanosleep_restart(struct restart_block *restart)
217{
218	struct compat_timespec __user *rmtp;
219	struct timespec rmt;
220	mm_segment_t oldfs;
221	long ret;
222
223	restart->nanosleep.rmtp = (struct timespec __user *) &rmt;
224	oldfs = get_fs();
225	set_fs(KERNEL_DS);
226	ret = hrtimer_nanosleep_restart(restart);
227	set_fs(oldfs);
228
229	if (ret == -ERESTART_RESTARTBLOCK) {
230		rmtp = restart->nanosleep.compat_rmtp;
231
232		if (rmtp && compat_put_timespec(&rmt, rmtp))
233			return -EFAULT;
234	}
235
236	return ret;
237}
238
239COMPAT_SYSCALL_DEFINE2(nanosleep, struct compat_timespec __user *, rqtp,
240		       struct compat_timespec __user *, rmtp)
241{
242	struct timespec tu, rmt;
243	mm_segment_t oldfs;
244	long ret;
245
246	if (compat_get_timespec(&tu, rqtp))
247		return -EFAULT;
248
249	if (!timespec_valid(&tu))
250		return -EINVAL;
251
252	oldfs = get_fs();
253	set_fs(KERNEL_DS);
254	ret = hrtimer_nanosleep(&tu,
255				rmtp ? (struct timespec __user *)&rmt : NULL,
256				HRTIMER_MODE_REL, CLOCK_MONOTONIC);
257	set_fs(oldfs);
258
259	/*
260	 * hrtimer_nanosleep() can only return 0 or
261	 * -ERESTART_RESTARTBLOCK here because:
262	 *
263	 * - we call it with HRTIMER_MODE_REL and therefor exclude the
264	 *   -ERESTARTNOHAND return path.
265	 *
266	 * - we supply the rmtp argument from the task stack (due to
267	 *   the necessary compat conversion. So the update cannot
268	 *   fail, which excludes the -EFAULT return path as well. If
269	 *   it fails nevertheless we have a bigger problem and wont
270	 *   reach this place anymore.
271	 *
272	 * - if the return value is 0, we do not have to update rmtp
273	 *    because there is no remaining time.
274	 *
275	 * We check for -ERESTART_RESTARTBLOCK nevertheless if the
276	 * core implementation decides to return random nonsense.
277	 */
278	if (ret == -ERESTART_RESTARTBLOCK) {
279		struct restart_block *restart = &current->restart_block;
280
281		restart->fn = compat_nanosleep_restart;
282		restart->nanosleep.compat_rmtp = rmtp;
283
284		if (rmtp && compat_put_timespec(&rmt, rmtp))
285			return -EFAULT;
286	}
287	return ret;
288}
289
290static inline long get_compat_itimerval(struct itimerval *o,
291		struct compat_itimerval __user *i)
292{
293	return (!access_ok(VERIFY_READ, i, sizeof(*i)) ||
294		(__get_user(o->it_interval.tv_sec, &i->it_interval.tv_sec) |
295		 __get_user(o->it_interval.tv_usec, &i->it_interval.tv_usec) |
296		 __get_user(o->it_value.tv_sec, &i->it_value.tv_sec) |
297		 __get_user(o->it_value.tv_usec, &i->it_value.tv_usec)));
298}
299
300static inline long put_compat_itimerval(struct compat_itimerval __user *o,
301		struct itimerval *i)
302{
303	return (!access_ok(VERIFY_WRITE, o, sizeof(*o)) ||
304		(__put_user(i->it_interval.tv_sec, &o->it_interval.tv_sec) |
305		 __put_user(i->it_interval.tv_usec, &o->it_interval.tv_usec) |
306		 __put_user(i->it_value.tv_sec, &o->it_value.tv_sec) |
307		 __put_user(i->it_value.tv_usec, &o->it_value.tv_usec)));
308}
309
310COMPAT_SYSCALL_DEFINE2(getitimer, int, which,
311		struct compat_itimerval __user *, it)
312{
313	struct itimerval kit;
314	int error;
315
316	error = do_getitimer(which, &kit);
317	if (!error && put_compat_itimerval(it, &kit))
318		error = -EFAULT;
319	return error;
320}
321
322COMPAT_SYSCALL_DEFINE3(setitimer, int, which,
323		struct compat_itimerval __user *, in,
324		struct compat_itimerval __user *, out)
325{
326	struct itimerval kin, kout;
327	int error;
328
329	if (in) {
330		if (get_compat_itimerval(&kin, in))
331			return -EFAULT;
332	} else
333		memset(&kin, 0, sizeof(kin));
334
335	error = do_setitimer(which, &kin, out ? &kout : NULL);
336	if (error || !out)
337		return error;
338	if (put_compat_itimerval(out, &kout))
339		return -EFAULT;
340	return 0;
341}
342
343static compat_clock_t clock_t_to_compat_clock_t(clock_t x)
344{
345	return compat_jiffies_to_clock_t(clock_t_to_jiffies(x));
346}
347
348COMPAT_SYSCALL_DEFINE1(times, struct compat_tms __user *, tbuf)
349{
350	if (tbuf) {
351		struct tms tms;
352		struct compat_tms tmp;
353
354		do_sys_times(&tms);
355		/* Convert our struct tms to the compat version. */
356		tmp.tms_utime = clock_t_to_compat_clock_t(tms.tms_utime);
357		tmp.tms_stime = clock_t_to_compat_clock_t(tms.tms_stime);
358		tmp.tms_cutime = clock_t_to_compat_clock_t(tms.tms_cutime);
359		tmp.tms_cstime = clock_t_to_compat_clock_t(tms.tms_cstime);
360		if (copy_to_user(tbuf, &tmp, sizeof(tmp)))
361			return -EFAULT;
362	}
363	force_successful_syscall_return();
364	return compat_jiffies_to_clock_t(jiffies);
365}
366
367#ifdef __ARCH_WANT_SYS_SIGPENDING
368
369/*
370 * Assumption: old_sigset_t and compat_old_sigset_t are both
371 * types that can be passed to put_user()/get_user().
372 */
373
374COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set)
375{
376	old_sigset_t s;
377	long ret;
378	mm_segment_t old_fs = get_fs();
379
380	set_fs(KERNEL_DS);
381	ret = sys_sigpending((old_sigset_t __user *) &s);
382	set_fs(old_fs);
383	if (ret == 0)
384		ret = put_user(s, set);
385	return ret;
386}
387
388#endif
389
390#ifdef __ARCH_WANT_SYS_SIGPROCMASK
391
392/*
393 * sys_sigprocmask SIG_SETMASK sets the first (compat) word of the
394 * blocked set of signals to the supplied signal set
395 */
396static inline void compat_sig_setmask(sigset_t *blocked, compat_sigset_word set)
397{
398	memcpy(blocked->sig, &set, sizeof(set));
399}
400
401COMPAT_SYSCALL_DEFINE3(sigprocmask, int, how,
402		       compat_old_sigset_t __user *, nset,
403		       compat_old_sigset_t __user *, oset)
404{
405	old_sigset_t old_set, new_set;
406	sigset_t new_blocked;
407
408	old_set = current->blocked.sig[0];
409
410	if (nset) {
411		if (get_user(new_set, nset))
412			return -EFAULT;
413		new_set &= ~(sigmask(SIGKILL) | sigmask(SIGSTOP));
414
415		new_blocked = current->blocked;
416
417		switch (how) {
418		case SIG_BLOCK:
419			sigaddsetmask(&new_blocked, new_set);
420			break;
421		case SIG_UNBLOCK:
422			sigdelsetmask(&new_blocked, new_set);
423			break;
424		case SIG_SETMASK:
425			compat_sig_setmask(&new_blocked, new_set);
426			break;
427		default:
428			return -EINVAL;
429		}
430
431		set_current_blocked(&new_blocked);
432	}
433
434	if (oset) {
435		if (put_user(old_set, oset))
436			return -EFAULT;
437	}
438
439	return 0;
440}
441
442#endif
443
444COMPAT_SYSCALL_DEFINE2(setrlimit, unsigned int, resource,
445		       struct compat_rlimit __user *, rlim)
446{
447	struct rlimit r;
448
449	if (!access_ok(VERIFY_READ, rlim, sizeof(*rlim)) ||
450	    __get_user(r.rlim_cur, &rlim->rlim_cur) ||
451	    __get_user(r.rlim_max, &rlim->rlim_max))
452		return -EFAULT;
453
454	if (r.rlim_cur == COMPAT_RLIM_INFINITY)
455		r.rlim_cur = RLIM_INFINITY;
456	if (r.rlim_max == COMPAT_RLIM_INFINITY)
457		r.rlim_max = RLIM_INFINITY;
458	return do_prlimit(current, resource, &r, NULL);
459}
460
461#ifdef COMPAT_RLIM_OLD_INFINITY
462
463COMPAT_SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
464		       struct compat_rlimit __user *, rlim)
465{
466	struct rlimit r;
467	int ret;
468	mm_segment_t old_fs = get_fs();
469
470	set_fs(KERNEL_DS);
471	ret = sys_old_getrlimit(resource, (struct rlimit __user *)&r);
472	set_fs(old_fs);
473
474	if (!ret) {
475		if (r.rlim_cur > COMPAT_RLIM_OLD_INFINITY)
476			r.rlim_cur = COMPAT_RLIM_INFINITY;
477		if (r.rlim_max > COMPAT_RLIM_OLD_INFINITY)
478			r.rlim_max = COMPAT_RLIM_INFINITY;
479
480		if (!access_ok(VERIFY_WRITE, rlim, sizeof(*rlim)) ||
481		    __put_user(r.rlim_cur, &rlim->rlim_cur) ||
482		    __put_user(r.rlim_max, &rlim->rlim_max))
483			return -EFAULT;
484	}
485	return ret;
486}
487
488#endif
489
490COMPAT_SYSCALL_DEFINE2(getrlimit, unsigned int, resource,
491		       struct compat_rlimit __user *, rlim)
492{
493	struct rlimit r;
494	int ret;
495
496	ret = do_prlimit(current, resource, NULL, &r);
497	if (!ret) {
498		if (r.rlim_cur > COMPAT_RLIM_INFINITY)
499			r.rlim_cur = COMPAT_RLIM_INFINITY;
500		if (r.rlim_max > COMPAT_RLIM_INFINITY)
501			r.rlim_max = COMPAT_RLIM_INFINITY;
502
503		if (!access_ok(VERIFY_WRITE, rlim, sizeof(*rlim)) ||
504		    __put_user(r.rlim_cur, &rlim->rlim_cur) ||
505		    __put_user(r.rlim_max, &rlim->rlim_max))
506			return -EFAULT;
507	}
508	return ret;
509}
510
511int put_compat_rusage(const struct rusage *r, struct compat_rusage __user *ru)
512{
513	if (!access_ok(VERIFY_WRITE, ru, sizeof(*ru)) ||
514	    __put_user(r->ru_utime.tv_sec, &ru->ru_utime.tv_sec) ||
515	    __put_user(r->ru_utime.tv_usec, &ru->ru_utime.tv_usec) ||
516	    __put_user(r->ru_stime.tv_sec, &ru->ru_stime.tv_sec) ||
517	    __put_user(r->ru_stime.tv_usec, &ru->ru_stime.tv_usec) ||
518	    __put_user(r->ru_maxrss, &ru->ru_maxrss) ||
519	    __put_user(r->ru_ixrss, &ru->ru_ixrss) ||
520	    __put_user(r->ru_idrss, &ru->ru_idrss) ||
521	    __put_user(r->ru_isrss, &ru->ru_isrss) ||
522	    __put_user(r->ru_minflt, &ru->ru_minflt) ||
523	    __put_user(r->ru_majflt, &ru->ru_majflt) ||
524	    __put_user(r->ru_nswap, &ru->ru_nswap) ||
525	    __put_user(r->ru_inblock, &ru->ru_inblock) ||
526	    __put_user(r->ru_oublock, &ru->ru_oublock) ||
527	    __put_user(r->ru_msgsnd, &ru->ru_msgsnd) ||
528	    __put_user(r->ru_msgrcv, &ru->ru_msgrcv) ||
529	    __put_user(r->ru_nsignals, &ru->ru_nsignals) ||
530	    __put_user(r->ru_nvcsw, &ru->ru_nvcsw) ||
531	    __put_user(r->ru_nivcsw, &ru->ru_nivcsw))
532		return -EFAULT;
533	return 0;
534}
535
536COMPAT_SYSCALL_DEFINE4(wait4,
537	compat_pid_t, pid,
538	compat_uint_t __user *, stat_addr,
539	int, options,
540	struct compat_rusage __user *, ru)
541{
542	if (!ru) {
543		return sys_wait4(pid, stat_addr, options, NULL);
544	} else {
545		struct rusage r;
546		int ret;
547		unsigned int status;
548		mm_segment_t old_fs = get_fs();
549
550		set_fs (KERNEL_DS);
551		ret = sys_wait4(pid,
552				(stat_addr ?
553				 (unsigned int __user *) &status : NULL),
554				options, (struct rusage __user *) &r);
555		set_fs (old_fs);
556
557		if (ret > 0) {
558			if (put_compat_rusage(&r, ru))
559				return -EFAULT;
560			if (stat_addr && put_user(status, stat_addr))
561				return -EFAULT;
562		}
563		return ret;
564	}
565}
566
567COMPAT_SYSCALL_DEFINE5(waitid,
568		int, which, compat_pid_t, pid,
569		struct compat_siginfo __user *, uinfo, int, options,
570		struct compat_rusage __user *, uru)
571{
572	siginfo_t info;
573	struct rusage ru;
574	long ret;
575	mm_segment_t old_fs = get_fs();
576
577	memset(&info, 0, sizeof(info));
578
579	set_fs(KERNEL_DS);
580	ret = sys_waitid(which, pid, (siginfo_t __user *)&info, options,
581			 uru ? (struct rusage __user *)&ru : NULL);
582	set_fs(old_fs);
583
584	if ((ret < 0) || (info.si_signo == 0))
585		return ret;
586
587	if (uru) {
588		/* sys_waitid() overwrites everything in ru */
589		if (COMPAT_USE_64BIT_TIME)
590			ret = copy_to_user(uru, &ru, sizeof(ru));
591		else
592			ret = put_compat_rusage(&ru, uru);
593		if (ret)
594			return -EFAULT;
595	}
596
597	BUG_ON(info.si_code & __SI_MASK);
598	info.si_code |= __SI_CHLD;
599	return copy_siginfo_to_user32(uinfo, &info);
600}
601
602static int compat_get_user_cpu_mask(compat_ulong_t __user *user_mask_ptr,
603				    unsigned len, struct cpumask *new_mask)
604{
605	unsigned long *k;
606
607	if (len < cpumask_size())
608		memset(new_mask, 0, cpumask_size());
609	else if (len > cpumask_size())
610		len = cpumask_size();
611
612	k = cpumask_bits(new_mask);
613	return compat_get_bitmap(k, user_mask_ptr, len * 8);
614}
615
616COMPAT_SYSCALL_DEFINE3(sched_setaffinity, compat_pid_t, pid,
617		       unsigned int, len,
618		       compat_ulong_t __user *, user_mask_ptr)
619{
620	cpumask_var_t new_mask;
621	int retval;
622
623	if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
624		return -ENOMEM;
625
626	retval = compat_get_user_cpu_mask(user_mask_ptr, len, new_mask);
627	if (retval)
628		goto out;
629
630	retval = sched_setaffinity(pid, new_mask);
631out:
632	free_cpumask_var(new_mask);
633	return retval;
634}
635
636COMPAT_SYSCALL_DEFINE3(sched_getaffinity, compat_pid_t,  pid, unsigned int, len,
637		       compat_ulong_t __user *, user_mask_ptr)
638{
639	int ret;
640	cpumask_var_t mask;
641
642	if ((len * BITS_PER_BYTE) < nr_cpu_ids)
643		return -EINVAL;
644	if (len & (sizeof(compat_ulong_t)-1))
645		return -EINVAL;
646
647	if (!alloc_cpumask_var(&mask, GFP_KERNEL))
648		return -ENOMEM;
649
650	ret = sched_getaffinity(pid, mask);
651	if (ret == 0) {
652		size_t retlen = min_t(size_t, len, cpumask_size());
653
654		if (compat_put_bitmap(user_mask_ptr, cpumask_bits(mask), retlen * 8))
655			ret = -EFAULT;
656		else
657			ret = retlen;
658	}
659	free_cpumask_var(mask);
660
661	return ret;
662}
663
664int get_compat_itimerspec(struct itimerspec *dst,
665			  const struct compat_itimerspec __user *src)
666{
667	if (__compat_get_timespec(&dst->it_interval, &src->it_interval) ||
668	    __compat_get_timespec(&dst->it_value, &src->it_value))
669		return -EFAULT;
670	return 0;
671}
672
673int put_compat_itimerspec(struct compat_itimerspec __user *dst,
674			  const struct itimerspec *src)
675{
676	if (__compat_put_timespec(&src->it_interval, &dst->it_interval) ||
677	    __compat_put_timespec(&src->it_value, &dst->it_value))
678		return -EFAULT;
679	return 0;
680}
681
682COMPAT_SYSCALL_DEFINE3(timer_create, clockid_t, which_clock,
683		       struct compat_sigevent __user *, timer_event_spec,
684		       timer_t __user *, created_timer_id)
685{
686	struct sigevent __user *event = NULL;
687
688	if (timer_event_spec) {
689		struct sigevent kevent;
690
691		event = compat_alloc_user_space(sizeof(*event));
692		if (get_compat_sigevent(&kevent, timer_event_spec) ||
693		    copy_to_user(event, &kevent, sizeof(*event)))
694			return -EFAULT;
695	}
696
697	return sys_timer_create(which_clock, event, created_timer_id);
698}
699
700COMPAT_SYSCALL_DEFINE4(timer_settime, timer_t, timer_id, int, flags,
701		       struct compat_itimerspec __user *, new,
702		       struct compat_itimerspec __user *, old)
703{
704	long err;
705	mm_segment_t oldfs;
706	struct itimerspec newts, oldts;
707
708	if (!new)
709		return -EINVAL;
710	if (get_compat_itimerspec(&newts, new))
711		return -EFAULT;
712	oldfs = get_fs();
713	set_fs(KERNEL_DS);
714	err = sys_timer_settime(timer_id, flags,
715				(struct itimerspec __user *) &newts,
716				(struct itimerspec __user *) &oldts);
717	set_fs(oldfs);
718	if (!err && old && put_compat_itimerspec(old, &oldts))
719		return -EFAULT;
720	return err;
721}
722
723COMPAT_SYSCALL_DEFINE2(timer_gettime, timer_t, timer_id,
724		       struct compat_itimerspec __user *, setting)
725{
726	long err;
727	mm_segment_t oldfs;
728	struct itimerspec ts;
729
730	oldfs = get_fs();
731	set_fs(KERNEL_DS);
732	err = sys_timer_gettime(timer_id,
733				(struct itimerspec __user *) &ts);
734	set_fs(oldfs);
735	if (!err && put_compat_itimerspec(setting, &ts))
736		return -EFAULT;
737	return err;
738}
739
740COMPAT_SYSCALL_DEFINE2(clock_settime, clockid_t, which_clock,
741		       struct compat_timespec __user *, tp)
742{
743	long err;
744	mm_segment_t oldfs;
745	struct timespec ts;
746
747	if (compat_get_timespec(&ts, tp))
748		return -EFAULT;
749	oldfs = get_fs();
750	set_fs(KERNEL_DS);
751	err = sys_clock_settime(which_clock,
752				(struct timespec __user *) &ts);
753	set_fs(oldfs);
754	return err;
755}
756
757COMPAT_SYSCALL_DEFINE2(clock_gettime, clockid_t, which_clock,
758		       struct compat_timespec __user *, tp)
759{
760	long err;
761	mm_segment_t oldfs;
762	struct timespec ts;
763
764	oldfs = get_fs();
765	set_fs(KERNEL_DS);
766	err = sys_clock_gettime(which_clock,
767				(struct timespec __user *) &ts);
768	set_fs(oldfs);
769	if (!err && compat_put_timespec(&ts, tp))
770		return -EFAULT;
771	return err;
772}
773
774COMPAT_SYSCALL_DEFINE2(clock_adjtime, clockid_t, which_clock,
775		       struct compat_timex __user *, utp)
776{
777	struct timex txc;
778	mm_segment_t oldfs;
779	int err, ret;
780
781	err = compat_get_timex(&txc, utp);
782	if (err)
783		return err;
784
785	oldfs = get_fs();
786	set_fs(KERNEL_DS);
787	ret = sys_clock_adjtime(which_clock, (struct timex __user *) &txc);
788	set_fs(oldfs);
789
790	err = compat_put_timex(utp, &txc);
791	if (err)
792		return err;
793
794	return ret;
795}
796
797COMPAT_SYSCALL_DEFINE2(clock_getres, clockid_t, which_clock,
798		       struct compat_timespec __user *, tp)
799{
800	long err;
801	mm_segment_t oldfs;
802	struct timespec ts;
803
804	oldfs = get_fs();
805	set_fs(KERNEL_DS);
806	err = sys_clock_getres(which_clock,
807			       (struct timespec __user *) &ts);
808	set_fs(oldfs);
809	if (!err && tp && compat_put_timespec(&ts, tp))
810		return -EFAULT;
811	return err;
812}
813
814static long compat_clock_nanosleep_restart(struct restart_block *restart)
815{
816	long err;
817	mm_segment_t oldfs;
818	struct timespec tu;
819	struct compat_timespec __user *rmtp = restart->nanosleep.compat_rmtp;
820
821	restart->nanosleep.rmtp = (struct timespec __user *) &tu;
822	oldfs = get_fs();
823	set_fs(KERNEL_DS);
824	err = clock_nanosleep_restart(restart);
825	set_fs(oldfs);
826
827	if ((err == -ERESTART_RESTARTBLOCK) && rmtp &&
828	    compat_put_timespec(&tu, rmtp))
829		return -EFAULT;
830
831	if (err == -ERESTART_RESTARTBLOCK) {
832		restart->fn = compat_clock_nanosleep_restart;
833		restart->nanosleep.compat_rmtp = rmtp;
834	}
835	return err;
836}
837
838COMPAT_SYSCALL_DEFINE4(clock_nanosleep, clockid_t, which_clock, int, flags,
839		       struct compat_timespec __user *, rqtp,
840		       struct compat_timespec __user *, rmtp)
841{
842	long err;
843	mm_segment_t oldfs;
844	struct timespec in, out;
845	struct restart_block *restart;
846
847	if (compat_get_timespec(&in, rqtp))
848		return -EFAULT;
849
850	oldfs = get_fs();
851	set_fs(KERNEL_DS);
852	err = sys_clock_nanosleep(which_clock, flags,
853				  (struct timespec __user *) &in,
854				  (struct timespec __user *) &out);
855	set_fs(oldfs);
856
857	if ((err == -ERESTART_RESTARTBLOCK) && rmtp &&
858	    compat_put_timespec(&out, rmtp))
859		return -EFAULT;
860
861	if (err == -ERESTART_RESTARTBLOCK) {
862		restart = &current->restart_block;
863		restart->fn = compat_clock_nanosleep_restart;
864		restart->nanosleep.compat_rmtp = rmtp;
865	}
866	return err;
867}
868
869/*
870 * We currently only need the following fields from the sigevent
871 * structure: sigev_value, sigev_signo, sig_notify and (sometimes
872 * sigev_notify_thread_id).  The others are handled in user mode.
873 * We also assume that copying sigev_value.sival_int is sufficient
874 * to keep all the bits of sigev_value.sival_ptr intact.
875 */
876int get_compat_sigevent(struct sigevent *event,
877		const struct compat_sigevent __user *u_event)
878{
879	memset(event, 0, sizeof(*event));
880	return (!access_ok(VERIFY_READ, u_event, sizeof(*u_event)) ||
881		__get_user(event->sigev_value.sival_int,
882			&u_event->sigev_value.sival_int) ||
883		__get_user(event->sigev_signo, &u_event->sigev_signo) ||
884		__get_user(event->sigev_notify, &u_event->sigev_notify) ||
885		__get_user(event->sigev_notify_thread_id,
886			&u_event->sigev_notify_thread_id))
887		? -EFAULT : 0;
888}
889
890long compat_get_bitmap(unsigned long *mask, const compat_ulong_t __user *umask,
891		       unsigned long bitmap_size)
892{
893	int i, j;
894	unsigned long m;
895	compat_ulong_t um;
896	unsigned long nr_compat_longs;
897
898	/* align bitmap up to nearest compat_long_t boundary */
899	bitmap_size = ALIGN(bitmap_size, BITS_PER_COMPAT_LONG);
900
901	if (!access_ok(VERIFY_READ, umask, bitmap_size / 8))
902		return -EFAULT;
903
904	nr_compat_longs = BITS_TO_COMPAT_LONGS(bitmap_size);
905
906	for (i = 0; i < BITS_TO_LONGS(bitmap_size); i++) {
907		m = 0;
908
909		for (j = 0; j < sizeof(m)/sizeof(um); j++) {
910			/*
911			 * We dont want to read past the end of the userspace
912			 * bitmap. We must however ensure the end of the
913			 * kernel bitmap is zeroed.
914			 */
915			if (nr_compat_longs) {
916				nr_compat_longs--;
917				if (__get_user(um, umask))
918					return -EFAULT;
919			} else {
920				um = 0;
921			}
922
923			umask++;
924			m |= (long)um << (j * BITS_PER_COMPAT_LONG);
925		}
926		*mask++ = m;
927	}
928
929	return 0;
930}
931
932long compat_put_bitmap(compat_ulong_t __user *umask, unsigned long *mask,
933		       unsigned long bitmap_size)
934{
935	int i, j;
936	unsigned long m;
937	compat_ulong_t um;
938	unsigned long nr_compat_longs;
939
940	/* align bitmap up to nearest compat_long_t boundary */
941	bitmap_size = ALIGN(bitmap_size, BITS_PER_COMPAT_LONG);
942
943	if (!access_ok(VERIFY_WRITE, umask, bitmap_size / 8))
944		return -EFAULT;
945
946	nr_compat_longs = BITS_TO_COMPAT_LONGS(bitmap_size);
947
948	for (i = 0; i < BITS_TO_LONGS(bitmap_size); i++) {
949		m = *mask++;
950
951		for (j = 0; j < sizeof(m)/sizeof(um); j++) {
952			um = m;
953
954			/*
955			 * We dont want to write past the end of the userspace
956			 * bitmap.
957			 */
958			if (nr_compat_longs) {
959				nr_compat_longs--;
960				if (__put_user(um, umask))
961					return -EFAULT;
962			}
963
964			umask++;
965			m >>= 4*sizeof(um);
966			m >>= 4*sizeof(um);
967		}
968	}
969
970	return 0;
971}
972
973void
974sigset_from_compat(sigset_t *set, const compat_sigset_t *compat)
975{
976	switch (_NSIG_WORDS) {
977	case 4: set->sig[3] = compat->sig[6] | (((long)compat->sig[7]) << 32 );
978	case 3: set->sig[2] = compat->sig[4] | (((long)compat->sig[5]) << 32 );
979	case 2: set->sig[1] = compat->sig[2] | (((long)compat->sig[3]) << 32 );
980	case 1: set->sig[0] = compat->sig[0] | (((long)compat->sig[1]) << 32 );
981	}
982}
983EXPORT_SYMBOL_GPL(sigset_from_compat);
984
985void
986sigset_to_compat(compat_sigset_t *compat, const sigset_t *set)
987{
988	switch (_NSIG_WORDS) {
989	case 4: compat->sig[7] = (set->sig[3] >> 32); compat->sig[6] = set->sig[3];
990	case 3: compat->sig[5] = (set->sig[2] >> 32); compat->sig[4] = set->sig[2];
991	case 2: compat->sig[3] = (set->sig[1] >> 32); compat->sig[2] = set->sig[1];
992	case 1: compat->sig[1] = (set->sig[0] >> 32); compat->sig[0] = set->sig[0];
993	}
994}
995
996COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait, compat_sigset_t __user *, uthese,
997		struct compat_siginfo __user *, uinfo,
998		struct compat_timespec __user *, uts, compat_size_t, sigsetsize)
999{
1000	compat_sigset_t s32;
1001	sigset_t s;
1002	struct timespec t;
1003	siginfo_t info;
1004	long ret;
1005
1006	if (sigsetsize != sizeof(sigset_t))
1007		return -EINVAL;
1008
1009	if (copy_from_user(&s32, uthese, sizeof(compat_sigset_t)))
1010		return -EFAULT;
1011	sigset_from_compat(&s, &s32);
1012
1013	if (uts) {
1014		if (compat_get_timespec(&t, uts))
1015			return -EFAULT;
1016	}
1017
1018	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
1019
1020	if (ret > 0 && uinfo) {
1021		if (copy_siginfo_to_user32(uinfo, &info))
1022			ret = -EFAULT;
1023	}
1024
1025	return ret;
1026}
1027
1028#ifdef __ARCH_WANT_COMPAT_SYS_TIME
1029
1030/* compat_time_t is a 32 bit "long" and needs to get converted. */
1031
1032COMPAT_SYSCALL_DEFINE1(time, compat_time_t __user *, tloc)
1033{
1034	compat_time_t i;
1035	struct timeval tv;
1036
1037	do_gettimeofday(&tv);
1038	i = tv.tv_sec;
1039
1040	if (tloc) {
1041		if (put_user(i,tloc))
1042			return -EFAULT;
1043	}
1044	force_successful_syscall_return();
1045	return i;
1046}
1047
1048COMPAT_SYSCALL_DEFINE1(stime, compat_time_t __user *, tptr)
1049{
1050	struct timespec tv;
1051	int err;
1052
1053	if (get_user(tv.tv_sec, tptr))
1054		return -EFAULT;
1055
1056	tv.tv_nsec = 0;
1057
1058	err = security_settime(&tv, NULL);
1059	if (err)
1060		return err;
1061
1062	do_settimeofday(&tv);
1063	return 0;
1064}
1065
1066#endif /* __ARCH_WANT_COMPAT_SYS_TIME */
1067
1068COMPAT_SYSCALL_DEFINE1(adjtimex, struct compat_timex __user *, utp)
1069{
1070	struct timex txc;
1071	int err, ret;
1072
1073	err = compat_get_timex(&txc, utp);
1074	if (err)
1075		return err;
1076
1077	ret = do_adjtimex(&txc);
1078
1079	err = compat_put_timex(utp, &txc);
1080	if (err)
1081		return err;
1082
1083	return ret;
1084}
1085
1086#ifdef CONFIG_NUMA
1087COMPAT_SYSCALL_DEFINE6(move_pages, pid_t, pid, compat_ulong_t, nr_pages,
1088		       compat_uptr_t __user *, pages32,
1089		       const int __user *, nodes,
1090		       int __user *, status,
1091		       int, flags)
1092{
1093	const void __user * __user *pages;
1094	int i;
1095
1096	pages = compat_alloc_user_space(nr_pages * sizeof(void *));
1097	for (i = 0; i < nr_pages; i++) {
1098		compat_uptr_t p;
1099
1100		if (get_user(p, pages32 + i) ||
1101			put_user(compat_ptr(p), pages + i))
1102			return -EFAULT;
1103	}
1104	return sys_move_pages(pid, nr_pages, pages, nodes, status, flags);
1105}
1106
1107COMPAT_SYSCALL_DEFINE4(migrate_pages, compat_pid_t, pid,
1108		       compat_ulong_t, maxnode,
1109		       const compat_ulong_t __user *, old_nodes,
1110		       const compat_ulong_t __user *, new_nodes)
1111{
1112	unsigned long __user *old = NULL;
1113	unsigned long __user *new = NULL;
1114	nodemask_t tmp_mask;
1115	unsigned long nr_bits;
1116	unsigned long size;
1117
1118	nr_bits = min_t(unsigned long, maxnode - 1, MAX_NUMNODES);
1119	size = ALIGN(nr_bits, BITS_PER_LONG) / 8;
1120	if (old_nodes) {
1121		if (compat_get_bitmap(nodes_addr(tmp_mask), old_nodes, nr_bits))
1122			return -EFAULT;
1123		old = compat_alloc_user_space(new_nodes ? size * 2 : size);
1124		if (new_nodes)
1125			new = old + size / sizeof(unsigned long);
1126		if (copy_to_user(old, nodes_addr(tmp_mask), size))
1127			return -EFAULT;
1128	}
1129	if (new_nodes) {
1130		if (compat_get_bitmap(nodes_addr(tmp_mask), new_nodes, nr_bits))
1131			return -EFAULT;
1132		if (new == NULL)
1133			new = compat_alloc_user_space(size);
1134		if (copy_to_user(new, nodes_addr(tmp_mask), size))
1135			return -EFAULT;
1136	}
1137	return sys_migrate_pages(pid, nr_bits + 1, old, new);
1138}
1139#endif
1140
1141COMPAT_SYSCALL_DEFINE2(sched_rr_get_interval,
1142		       compat_pid_t, pid,
1143		       struct compat_timespec __user *, interval)
1144{
1145	struct timespec t;
1146	int ret;
1147	mm_segment_t old_fs = get_fs();
1148
1149	set_fs(KERNEL_DS);
1150	ret = sys_sched_rr_get_interval(pid, (struct timespec __user *)&t);
1151	set_fs(old_fs);
1152	if (compat_put_timespec(&t, interval))
1153		return -EFAULT;
1154	return ret;
1155}
1156
1157/*
1158 * Allocate user-space memory for the duration of a single system call,
1159 * in order to marshall parameters inside a compat thunk.
1160 */
1161void __user *compat_alloc_user_space(unsigned long len)
1162{
1163	void __user *ptr;
1164
1165	/* If len would occupy more than half of the entire compat space... */
1166	if (unlikely(len > (((compat_uptr_t)~0) >> 1)))
1167		return NULL;
1168
1169	ptr = arch_compat_alloc_user_space(len);
1170
1171	if (unlikely(!access_ok(VERIFY_WRITE, ptr, len)))
1172		return NULL;
1173
1174	return ptr;
1175}
1176EXPORT_SYMBOL_GPL(compat_alloc_user_space);
1177