~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/kernel/sys.c

Version: ~ [ linux-5.4-rc7 ] ~ [ linux-5.3.11 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.84 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.154 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.201 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.201 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.77 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 /*
  2  *  linux/kernel/sys.c
  3  *
  4  *  Copyright (C) 1991, 1992  Linus Torvalds
  5  */
  6 
  7 #include <linux/export.h>
  8 #include <linux/mm.h>
  9 #include <linux/utsname.h>
 10 #include <linux/mman.h>
 11 #include <linux/reboot.h>
 12 #include <linux/prctl.h>
 13 #include <linux/highuid.h>
 14 #include <linux/fs.h>
 15 #include <linux/kmod.h>
 16 #include <linux/perf_event.h>
 17 #include <linux/resource.h>
 18 #include <linux/kernel.h>
 19 #include <linux/workqueue.h>
 20 #include <linux/capability.h>
 21 #include <linux/device.h>
 22 #include <linux/key.h>
 23 #include <linux/times.h>
 24 #include <linux/posix-timers.h>
 25 #include <linux/security.h>
 26 #include <linux/dcookies.h>
 27 #include <linux/suspend.h>
 28 #include <linux/tty.h>
 29 #include <linux/signal.h>
 30 #include <linux/cn_proc.h>
 31 #include <linux/getcpu.h>
 32 #include <linux/task_io_accounting_ops.h>
 33 #include <linux/seccomp.h>
 34 #include <linux/cpu.h>
 35 #include <linux/personality.h>
 36 #include <linux/ptrace.h>
 37 #include <linux/fs_struct.h>
 38 #include <linux/file.h>
 39 #include <linux/mount.h>
 40 #include <linux/gfp.h>
 41 #include <linux/syscore_ops.h>
 42 #include <linux/version.h>
 43 #include <linux/ctype.h>
 44 
 45 #include <linux/compat.h>
 46 #include <linux/syscalls.h>
 47 #include <linux/kprobes.h>
 48 #include <linux/user_namespace.h>
 49 #include <linux/binfmts.h>
 50 
 51 #include <linux/sched.h>
 52 #include <linux/rcupdate.h>
 53 #include <linux/uidgid.h>
 54 #include <linux/cred.h>
 55 
 56 #include <linux/kmsg_dump.h>
 57 /* Move somewhere else to avoid recompiling? */
 58 #include <generated/utsrelease.h>
 59 
 60 #include <asm/uaccess.h>
 61 #include <asm/io.h>
 62 #include <asm/unistd.h>
 63 
 64 #ifndef SET_UNALIGN_CTL
 65 # define SET_UNALIGN_CTL(a, b)  (-EINVAL)
 66 #endif
 67 #ifndef GET_UNALIGN_CTL
 68 # define GET_UNALIGN_CTL(a, b)  (-EINVAL)
 69 #endif
 70 #ifndef SET_FPEMU_CTL
 71 # define SET_FPEMU_CTL(a, b)    (-EINVAL)
 72 #endif
 73 #ifndef GET_FPEMU_CTL
 74 # define GET_FPEMU_CTL(a, b)    (-EINVAL)
 75 #endif
 76 #ifndef SET_FPEXC_CTL
 77 # define SET_FPEXC_CTL(a, b)    (-EINVAL)
 78 #endif
 79 #ifndef GET_FPEXC_CTL
 80 # define GET_FPEXC_CTL(a, b)    (-EINVAL)
 81 #endif
 82 #ifndef GET_ENDIAN
 83 # define GET_ENDIAN(a, b)       (-EINVAL)
 84 #endif
 85 #ifndef SET_ENDIAN
 86 # define SET_ENDIAN(a, b)       (-EINVAL)
 87 #endif
 88 #ifndef GET_TSC_CTL
 89 # define GET_TSC_CTL(a)         (-EINVAL)
 90 #endif
 91 #ifndef SET_TSC_CTL
 92 # define SET_TSC_CTL(a)         (-EINVAL)
 93 #endif
 94 #ifndef MPX_ENABLE_MANAGEMENT
 95 # define MPX_ENABLE_MANAGEMENT()        (-EINVAL)
 96 #endif
 97 #ifndef MPX_DISABLE_MANAGEMENT
 98 # define MPX_DISABLE_MANAGEMENT()       (-EINVAL)
 99 #endif
100 #ifndef GET_FP_MODE
101 # define GET_FP_MODE(a)         (-EINVAL)
102 #endif
103 #ifndef SET_FP_MODE
104 # define SET_FP_MODE(a,b)       (-EINVAL)
105 #endif
106 
107 /*
108  * this is where the system-wide overflow UID and GID are defined, for
109  * architectures that now have 32-bit UID/GID but didn't in the past
110  */
111 
112 int overflowuid = DEFAULT_OVERFLOWUID;
113 int overflowgid = DEFAULT_OVERFLOWGID;
114 
115 EXPORT_SYMBOL(overflowuid);
116 EXPORT_SYMBOL(overflowgid);
117 
118 /*
119  * the same as above, but for filesystems which can only store a 16-bit
120  * UID and GID. as such, this is needed on all architectures
121  */
122 
123 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
124 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
125 
126 EXPORT_SYMBOL(fs_overflowuid);
127 EXPORT_SYMBOL(fs_overflowgid);
128 
129 /*
130  * Returns true if current's euid is same as p's uid or euid,
131  * or has CAP_SYS_NICE to p's user_ns.
132  *
133  * Called with rcu_read_lock, creds are safe
134  */
135 static bool set_one_prio_perm(struct task_struct *p)
136 {
137         const struct cred *cred = current_cred(), *pcred = __task_cred(p);
138 
139         if (uid_eq(pcred->uid,  cred->euid) ||
140             uid_eq(pcred->euid, cred->euid))
141                 return true;
142         if (ns_capable(pcred->user_ns, CAP_SYS_NICE))
143                 return true;
144         return false;
145 }
146 
147 /*
148  * set the priority of a task
149  * - the caller must hold the RCU read lock
150  */
151 static int set_one_prio(struct task_struct *p, int niceval, int error)
152 {
153         int no_nice;
154 
155         if (!set_one_prio_perm(p)) {
156                 error = -EPERM;
157                 goto out;
158         }
159         if (niceval < task_nice(p) && !can_nice(p, niceval)) {
160                 error = -EACCES;
161                 goto out;
162         }
163         no_nice = security_task_setnice(p, niceval);
164         if (no_nice) {
165                 error = no_nice;
166                 goto out;
167         }
168         if (error == -ESRCH)
169                 error = 0;
170         set_user_nice(p, niceval);
171 out:
172         return error;
173 }
174 
175 SYSCALL_DEFINE3(setpriority, int, which, int, who, int, niceval)
176 {
177         struct task_struct *g, *p;
178         struct user_struct *user;
179         const struct cred *cred = current_cred();
180         int error = -EINVAL;
181         struct pid *pgrp;
182         kuid_t uid;
183 
184         if (which > PRIO_USER || which < PRIO_PROCESS)
185                 goto out;
186         if (!ccs_capable(CCS_SYS_NICE)) {
187                 error = -EPERM;
188                 goto out;
189         }
190 
191         /* normalize: avoid signed division (rounding problems) */
192         error = -ESRCH;
193         if (niceval < MIN_NICE)
194                 niceval = MIN_NICE;
195         if (niceval > MAX_NICE)
196                 niceval = MAX_NICE;
197 
198         rcu_read_lock();
199         read_lock(&tasklist_lock);
200         switch (which) {
201         case PRIO_PROCESS:
202                 if (who)
203                         p = find_task_by_vpid(who);
204                 else
205                         p = current;
206                 if (p)
207                         error = set_one_prio(p, niceval, error);
208                 break;
209         case PRIO_PGRP:
210                 if (who)
211                         pgrp = find_vpid(who);
212                 else
213                         pgrp = task_pgrp(current);
214                 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
215                         error = set_one_prio(p, niceval, error);
216                 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
217                 break;
218         case PRIO_USER:
219                 uid = make_kuid(cred->user_ns, who);
220                 user = cred->user;
221                 if (!who)
222                         uid = cred->uid;
223                 else if (!uid_eq(uid, cred->uid)) {
224                         user = find_user(uid);
225                         if (!user)
226                                 goto out_unlock;        /* No processes for this user */
227                 }
228                 do_each_thread(g, p) {
229                         if (uid_eq(task_uid(p), uid) && task_pid_vnr(p))
230                                 error = set_one_prio(p, niceval, error);
231                 } while_each_thread(g, p);
232                 if (!uid_eq(uid, cred->uid))
233                         free_uid(user);         /* For find_user() */
234                 break;
235         }
236 out_unlock:
237         read_unlock(&tasklist_lock);
238         rcu_read_unlock();
239 out:
240         return error;
241 }
242 
243 /*
244  * Ugh. To avoid negative return values, "getpriority()" will
245  * not return the normal nice-value, but a negated value that
246  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
247  * to stay compatible.
248  */
249 SYSCALL_DEFINE2(getpriority, int, which, int, who)
250 {
251         struct task_struct *g, *p;
252         struct user_struct *user;
253         const struct cred *cred = current_cred();
254         long niceval, retval = -ESRCH;
255         struct pid *pgrp;
256         kuid_t uid;
257 
258         if (which > PRIO_USER || which < PRIO_PROCESS)
259                 return -EINVAL;
260 
261         rcu_read_lock();
262         read_lock(&tasklist_lock);
263         switch (which) {
264         case PRIO_PROCESS:
265                 if (who)
266                         p = find_task_by_vpid(who);
267                 else
268                         p = current;
269                 if (p) {
270                         niceval = nice_to_rlimit(task_nice(p));
271                         if (niceval > retval)
272                                 retval = niceval;
273                 }
274                 break;
275         case PRIO_PGRP:
276                 if (who)
277                         pgrp = find_vpid(who);
278                 else
279                         pgrp = task_pgrp(current);
280                 do_each_pid_thread(pgrp, PIDTYPE_PGID, p) {
281                         niceval = nice_to_rlimit(task_nice(p));
282                         if (niceval > retval)
283                                 retval = niceval;
284                 } while_each_pid_thread(pgrp, PIDTYPE_PGID, p);
285                 break;
286         case PRIO_USER:
287                 uid = make_kuid(cred->user_ns, who);
288                 user = cred->user;
289                 if (!who)
290                         uid = cred->uid;
291                 else if (!uid_eq(uid, cred->uid)) {
292                         user = find_user(uid);
293                         if (!user)
294                                 goto out_unlock;        /* No processes for this user */
295                 }
296                 do_each_thread(g, p) {
297                         if (uid_eq(task_uid(p), uid) && task_pid_vnr(p)) {
298                                 niceval = nice_to_rlimit(task_nice(p));
299                                 if (niceval > retval)
300                                         retval = niceval;
301                         }
302                 } while_each_thread(g, p);
303                 if (!uid_eq(uid, cred->uid))
304                         free_uid(user);         /* for find_user() */
305                 break;
306         }
307 out_unlock:
308         read_unlock(&tasklist_lock);
309         rcu_read_unlock();
310 
311         return retval;
312 }
313 
314 /*
315  * Unprivileged users may change the real gid to the effective gid
316  * or vice versa.  (BSD-style)
317  *
318  * If you set the real gid at all, or set the effective gid to a value not
319  * equal to the real gid, then the saved gid is set to the new effective gid.
320  *
321  * This makes it possible for a setgid program to completely drop its
322  * privileges, which is often a useful assertion to make when you are doing
323  * a security audit over a program.
324  *
325  * The general idea is that a program which uses just setregid() will be
326  * 100% compatible with BSD.  A program which uses just setgid() will be
327  * 100% compatible with POSIX with saved IDs.
328  *
329  * SMP: There are not races, the GIDs are checked only by filesystem
330  *      operations (as far as semantic preservation is concerned).
331  */
332 #ifdef CONFIG_MULTIUSER
333 SYSCALL_DEFINE2(setregid, gid_t, rgid, gid_t, egid)
334 {
335         struct user_namespace *ns = current_user_ns();
336         const struct cred *old;
337         struct cred *new;
338         int retval;
339         kgid_t krgid, kegid;
340 
341         krgid = make_kgid(ns, rgid);
342         kegid = make_kgid(ns, egid);
343 
344         if ((rgid != (gid_t) -1) && !gid_valid(krgid))
345                 return -EINVAL;
346         if ((egid != (gid_t) -1) && !gid_valid(kegid))
347                 return -EINVAL;
348 
349         new = prepare_creds();
350         if (!new)
351                 return -ENOMEM;
352         old = current_cred();
353 
354         retval = -EPERM;
355         if (rgid != (gid_t) -1) {
356                 if (gid_eq(old->gid, krgid) ||
357                     gid_eq(old->egid, krgid) ||
358                     ns_capable(old->user_ns, CAP_SETGID))
359                         new->gid = krgid;
360                 else
361                         goto error;
362         }
363         if (egid != (gid_t) -1) {
364                 if (gid_eq(old->gid, kegid) ||
365                     gid_eq(old->egid, kegid) ||
366                     gid_eq(old->sgid, kegid) ||
367                     ns_capable(old->user_ns, CAP_SETGID))
368                         new->egid = kegid;
369                 else
370                         goto error;
371         }
372 
373         if (rgid != (gid_t) -1 ||
374             (egid != (gid_t) -1 && !gid_eq(kegid, old->gid)))
375                 new->sgid = new->egid;
376         new->fsgid = new->egid;
377 
378         return commit_creds(new);
379 
380 error:
381         abort_creds(new);
382         return retval;
383 }
384 
385 /*
386  * setgid() is implemented like SysV w/ SAVED_IDS
387  *
388  * SMP: Same implicit races as above.
389  */
390 SYSCALL_DEFINE1(setgid, gid_t, gid)
391 {
392         struct user_namespace *ns = current_user_ns();
393         const struct cred *old;
394         struct cred *new;
395         int retval;
396         kgid_t kgid;
397 
398         kgid = make_kgid(ns, gid);
399         if (!gid_valid(kgid))
400                 return -EINVAL;
401 
402         new = prepare_creds();
403         if (!new)
404                 return -ENOMEM;
405         old = current_cred();
406 
407         retval = -EPERM;
408         if (ns_capable(old->user_ns, CAP_SETGID))
409                 new->gid = new->egid = new->sgid = new->fsgid = kgid;
410         else if (gid_eq(kgid, old->gid) || gid_eq(kgid, old->sgid))
411                 new->egid = new->fsgid = kgid;
412         else
413                 goto error;
414 
415         return commit_creds(new);
416 
417 error:
418         abort_creds(new);
419         return retval;
420 }
421 
422 /*
423  * change the user struct in a credentials set to match the new UID
424  */
425 static int set_user(struct cred *new)
426 {
427         struct user_struct *new_user;
428 
429         new_user = alloc_uid(new->uid);
430         if (!new_user)
431                 return -EAGAIN;
432 
433         /*
434          * We don't fail in case of NPROC limit excess here because too many
435          * poorly written programs don't check set*uid() return code, assuming
436          * it never fails if called by root.  We may still enforce NPROC limit
437          * for programs doing set*uid()+execve() by harmlessly deferring the
438          * failure to the execve() stage.
439          */
440         if (atomic_read(&new_user->processes) >= rlimit(RLIMIT_NPROC) &&
441                         new_user != INIT_USER)
442                 current->flags |= PF_NPROC_EXCEEDED;
443         else
444                 current->flags &= ~PF_NPROC_EXCEEDED;
445 
446         free_uid(new->user);
447         new->user = new_user;
448         return 0;
449 }
450 
451 /*
452  * Unprivileged users may change the real uid to the effective uid
453  * or vice versa.  (BSD-style)
454  *
455  * If you set the real uid at all, or set the effective uid to a value not
456  * equal to the real uid, then the saved uid is set to the new effective uid.
457  *
458  * This makes it possible for a setuid program to completely drop its
459  * privileges, which is often a useful assertion to make when you are doing
460  * a security audit over a program.
461  *
462  * The general idea is that a program which uses just setreuid() will be
463  * 100% compatible with BSD.  A program which uses just setuid() will be
464  * 100% compatible with POSIX with saved IDs.
465  */
466 SYSCALL_DEFINE2(setreuid, uid_t, ruid, uid_t, euid)
467 {
468         struct user_namespace *ns = current_user_ns();
469         const struct cred *old;
470         struct cred *new;
471         int retval;
472         kuid_t kruid, keuid;
473 
474         kruid = make_kuid(ns, ruid);
475         keuid = make_kuid(ns, euid);
476 
477         if ((ruid != (uid_t) -1) && !uid_valid(kruid))
478                 return -EINVAL;
479         if ((euid != (uid_t) -1) && !uid_valid(keuid))
480                 return -EINVAL;
481 
482         new = prepare_creds();
483         if (!new)
484                 return -ENOMEM;
485         old = current_cred();
486 
487         retval = -EPERM;
488         if (ruid != (uid_t) -1) {
489                 new->uid = kruid;
490                 if (!uid_eq(old->uid, kruid) &&
491                     !uid_eq(old->euid, kruid) &&
492                     !ns_capable(old->user_ns, CAP_SETUID))
493                         goto error;
494         }
495 
496         if (euid != (uid_t) -1) {
497                 new->euid = keuid;
498                 if (!uid_eq(old->uid, keuid) &&
499                     !uid_eq(old->euid, keuid) &&
500                     !uid_eq(old->suid, keuid) &&
501                     !ns_capable(old->user_ns, CAP_SETUID))
502                         goto error;
503         }
504 
505         if (!uid_eq(new->uid, old->uid)) {
506                 retval = set_user(new);
507                 if (retval < 0)
508                         goto error;
509         }
510         if (ruid != (uid_t) -1 ||
511             (euid != (uid_t) -1 && !uid_eq(keuid, old->uid)))
512                 new->suid = new->euid;
513         new->fsuid = new->euid;
514 
515         retval = security_task_fix_setuid(new, old, LSM_SETID_RE);
516         if (retval < 0)
517                 goto error;
518 
519         return commit_creds(new);
520 
521 error:
522         abort_creds(new);
523         return retval;
524 }
525 
526 /*
527  * setuid() is implemented like SysV with SAVED_IDS
528  *
529  * Note that SAVED_ID's is deficient in that a setuid root program
530  * like sendmail, for example, cannot set its uid to be a normal
531  * user and then switch back, because if you're root, setuid() sets
532  * the saved uid too.  If you don't like this, blame the bright people
533  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
534  * will allow a root program to temporarily drop privileges and be able to
535  * regain them by swapping the real and effective uid.
536  */
537 SYSCALL_DEFINE1(setuid, uid_t, uid)
538 {
539         struct user_namespace *ns = current_user_ns();
540         const struct cred *old;
541         struct cred *new;
542         int retval;
543         kuid_t kuid;
544 
545         kuid = make_kuid(ns, uid);
546         if (!uid_valid(kuid))
547                 return -EINVAL;
548 
549         new = prepare_creds();
550         if (!new)
551                 return -ENOMEM;
552         old = current_cred();
553 
554         retval = -EPERM;
555         if (ns_capable(old->user_ns, CAP_SETUID)) {
556                 new->suid = new->uid = kuid;
557                 if (!uid_eq(kuid, old->uid)) {
558                         retval = set_user(new);
559                         if (retval < 0)
560                                 goto error;
561                 }
562         } else if (!uid_eq(kuid, old->uid) && !uid_eq(kuid, new->suid)) {
563                 goto error;
564         }
565 
566         new->fsuid = new->euid = kuid;
567 
568         retval = security_task_fix_setuid(new, old, LSM_SETID_ID);
569         if (retval < 0)
570                 goto error;
571 
572         return commit_creds(new);
573 
574 error:
575         abort_creds(new);
576         return retval;
577 }
578 
579 
580 /*
581  * This function implements a generic ability to update ruid, euid,
582  * and suid.  This allows you to implement the 4.4 compatible seteuid().
583  */
584 SYSCALL_DEFINE3(setresuid, uid_t, ruid, uid_t, euid, uid_t, suid)
585 {
586         struct user_namespace *ns = current_user_ns();
587         const struct cred *old;
588         struct cred *new;
589         int retval;
590         kuid_t kruid, keuid, ksuid;
591 
592         kruid = make_kuid(ns, ruid);
593         keuid = make_kuid(ns, euid);
594         ksuid = make_kuid(ns, suid);
595 
596         if ((ruid != (uid_t) -1) && !uid_valid(kruid))
597                 return -EINVAL;
598 
599         if ((euid != (uid_t) -1) && !uid_valid(keuid))
600                 return -EINVAL;
601 
602         if ((suid != (uid_t) -1) && !uid_valid(ksuid))
603                 return -EINVAL;
604 
605         new = prepare_creds();
606         if (!new)
607                 return -ENOMEM;
608 
609         old = current_cred();
610 
611         retval = -EPERM;
612         if (!ns_capable(old->user_ns, CAP_SETUID)) {
613                 if (ruid != (uid_t) -1        && !uid_eq(kruid, old->uid) &&
614                     !uid_eq(kruid, old->euid) && !uid_eq(kruid, old->suid))
615                         goto error;
616                 if (euid != (uid_t) -1        && !uid_eq(keuid, old->uid) &&
617                     !uid_eq(keuid, old->euid) && !uid_eq(keuid, old->suid))
618                         goto error;
619                 if (suid != (uid_t) -1        && !uid_eq(ksuid, old->uid) &&
620                     !uid_eq(ksuid, old->euid) && !uid_eq(ksuid, old->suid))
621                         goto error;
622         }
623 
624         if (ruid != (uid_t) -1) {
625                 new->uid = kruid;
626                 if (!uid_eq(kruid, old->uid)) {
627                         retval = set_user(new);
628                         if (retval < 0)
629                                 goto error;
630                 }
631         }
632         if (euid != (uid_t) -1)
633                 new->euid = keuid;
634         if (suid != (uid_t) -1)
635                 new->suid = ksuid;
636         new->fsuid = new->euid;
637 
638         retval = security_task_fix_setuid(new, old, LSM_SETID_RES);
639         if (retval < 0)
640                 goto error;
641 
642         return commit_creds(new);
643 
644 error:
645         abort_creds(new);
646         return retval;
647 }
648 
649 SYSCALL_DEFINE3(getresuid, uid_t __user *, ruidp, uid_t __user *, euidp, uid_t __user *, suidp)
650 {
651         const struct cred *cred = current_cred();
652         int retval;
653         uid_t ruid, euid, suid;
654 
655         ruid = from_kuid_munged(cred->user_ns, cred->uid);
656         euid = from_kuid_munged(cred->user_ns, cred->euid);
657         suid = from_kuid_munged(cred->user_ns, cred->suid);
658 
659         retval = put_user(ruid, ruidp);
660         if (!retval) {
661                 retval = put_user(euid, euidp);
662                 if (!retval)
663                         return put_user(suid, suidp);
664         }
665         return retval;
666 }
667 
668 /*
669  * Same as above, but for rgid, egid, sgid.
670  */
671 SYSCALL_DEFINE3(setresgid, gid_t, rgid, gid_t, egid, gid_t, sgid)
672 {
673         struct user_namespace *ns = current_user_ns();
674         const struct cred *old;
675         struct cred *new;
676         int retval;
677         kgid_t krgid, kegid, ksgid;
678 
679         krgid = make_kgid(ns, rgid);
680         kegid = make_kgid(ns, egid);
681         ksgid = make_kgid(ns, sgid);
682 
683         if ((rgid != (gid_t) -1) && !gid_valid(krgid))
684                 return -EINVAL;
685         if ((egid != (gid_t) -1) && !gid_valid(kegid))
686                 return -EINVAL;
687         if ((sgid != (gid_t) -1) && !gid_valid(ksgid))
688                 return -EINVAL;
689 
690         new = prepare_creds();
691         if (!new)
692                 return -ENOMEM;
693         old = current_cred();
694 
695         retval = -EPERM;
696         if (!ns_capable(old->user_ns, CAP_SETGID)) {
697                 if (rgid != (gid_t) -1        && !gid_eq(krgid, old->gid) &&
698                     !gid_eq(krgid, old->egid) && !gid_eq(krgid, old->sgid))
699                         goto error;
700                 if (egid != (gid_t) -1        && !gid_eq(kegid, old->gid) &&
701                     !gid_eq(kegid, old->egid) && !gid_eq(kegid, old->sgid))
702                         goto error;
703                 if (sgid != (gid_t) -1        && !gid_eq(ksgid, old->gid) &&
704                     !gid_eq(ksgid, old->egid) && !gid_eq(ksgid, old->sgid))
705                         goto error;
706         }
707 
708         if (rgid != (gid_t) -1)
709                 new->gid = krgid;
710         if (egid != (gid_t) -1)
711                 new->egid = kegid;
712         if (sgid != (gid_t) -1)
713                 new->sgid = ksgid;
714         new->fsgid = new->egid;
715 
716         return commit_creds(new);
717 
718 error:
719         abort_creds(new);
720         return retval;
721 }
722 
723 SYSCALL_DEFINE3(getresgid, gid_t __user *, rgidp, gid_t __user *, egidp, gid_t __user *, sgidp)
724 {
725         const struct cred *cred = current_cred();
726         int retval;
727         gid_t rgid, egid, sgid;
728 
729         rgid = from_kgid_munged(cred->user_ns, cred->gid);
730         egid = from_kgid_munged(cred->user_ns, cred->egid);
731         sgid = from_kgid_munged(cred->user_ns, cred->sgid);
732 
733         retval = put_user(rgid, rgidp);
734         if (!retval) {
735                 retval = put_user(egid, egidp);
736                 if (!retval)
737                         retval = put_user(sgid, sgidp);
738         }
739 
740         return retval;
741 }
742 
743 
744 /*
745  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
746  * is used for "access()" and for the NFS daemon (letting nfsd stay at
747  * whatever uid it wants to). It normally shadows "euid", except when
748  * explicitly set by setfsuid() or for access..
749  */
750 SYSCALL_DEFINE1(setfsuid, uid_t, uid)
751 {
752         const struct cred *old;
753         struct cred *new;
754         uid_t old_fsuid;
755         kuid_t kuid;
756 
757         old = current_cred();
758         old_fsuid = from_kuid_munged(old->user_ns, old->fsuid);
759 
760         kuid = make_kuid(old->user_ns, uid);
761         if (!uid_valid(kuid))
762                 return old_fsuid;
763 
764         new = prepare_creds();
765         if (!new)
766                 return old_fsuid;
767 
768         if (uid_eq(kuid, old->uid)  || uid_eq(kuid, old->euid)  ||
769             uid_eq(kuid, old->suid) || uid_eq(kuid, old->fsuid) ||
770             ns_capable(old->user_ns, CAP_SETUID)) {
771                 if (!uid_eq(kuid, old->fsuid)) {
772                         new->fsuid = kuid;
773                         if (security_task_fix_setuid(new, old, LSM_SETID_FS) == 0)
774                                 goto change_okay;
775                 }
776         }
777 
778         abort_creds(new);
779         return old_fsuid;
780 
781 change_okay:
782         commit_creds(new);
783         return old_fsuid;
784 }
785 
786 /*
787  * Samma på svenska..
788  */
789 SYSCALL_DEFINE1(setfsgid, gid_t, gid)
790 {
791         const struct cred *old;
792         struct cred *new;
793         gid_t old_fsgid;
794         kgid_t kgid;
795 
796         old = current_cred();
797         old_fsgid = from_kgid_munged(old->user_ns, old->fsgid);
798 
799         kgid = make_kgid(old->user_ns, gid);
800         if (!gid_valid(kgid))
801                 return old_fsgid;
802 
803         new = prepare_creds();
804         if (!new)
805                 return old_fsgid;
806 
807         if (gid_eq(kgid, old->gid)  || gid_eq(kgid, old->egid)  ||
808             gid_eq(kgid, old->sgid) || gid_eq(kgid, old->fsgid) ||
809             ns_capable(old->user_ns, CAP_SETGID)) {
810                 if (!gid_eq(kgid, old->fsgid)) {
811                         new->fsgid = kgid;
812                         goto change_okay;
813                 }
814         }
815 
816         abort_creds(new);
817         return old_fsgid;
818 
819 change_okay:
820         commit_creds(new);
821         return old_fsgid;
822 }
823 #endif /* CONFIG_MULTIUSER */
824 
825 /**
826  * sys_getpid - return the thread group id of the current process
827  *
828  * Note, despite the name, this returns the tgid not the pid.  The tgid and
829  * the pid are identical unless CLONE_THREAD was specified on clone() in
830  * which case the tgid is the same in all threads of the same group.
831  *
832  * This is SMP safe as current->tgid does not change.
833  */
834 SYSCALL_DEFINE0(getpid)
835 {
836         return task_tgid_vnr(current);
837 }
838 
839 /* Thread ID - the internal kernel "pid" */
840 SYSCALL_DEFINE0(gettid)
841 {
842         return task_pid_vnr(current);
843 }
844 
845 /*
846  * Accessing ->real_parent is not SMP-safe, it could
847  * change from under us. However, we can use a stale
848  * value of ->real_parent under rcu_read_lock(), see
849  * release_task()->call_rcu(delayed_put_task_struct).
850  */
851 SYSCALL_DEFINE0(getppid)
852 {
853         int pid;
854 
855         rcu_read_lock();
856         pid = task_tgid_vnr(rcu_dereference(current->real_parent));
857         rcu_read_unlock();
858 
859         return pid;
860 }
861 
862 SYSCALL_DEFINE0(getuid)
863 {
864         /* Only we change this so SMP safe */
865         return from_kuid_munged(current_user_ns(), current_uid());
866 }
867 
868 SYSCALL_DEFINE0(geteuid)
869 {
870         /* Only we change this so SMP safe */
871         return from_kuid_munged(current_user_ns(), current_euid());
872 }
873 
874 SYSCALL_DEFINE0(getgid)
875 {
876         /* Only we change this so SMP safe */
877         return from_kgid_munged(current_user_ns(), current_gid());
878 }
879 
880 SYSCALL_DEFINE0(getegid)
881 {
882         /* Only we change this so SMP safe */
883         return from_kgid_munged(current_user_ns(), current_egid());
884 }
885 
886 void do_sys_times(struct tms *tms)
887 {
888         cputime_t tgutime, tgstime, cutime, cstime;
889 
890         thread_group_cputime_adjusted(current, &tgutime, &tgstime);
891         cutime = current->signal->cutime;
892         cstime = current->signal->cstime;
893         tms->tms_utime = cputime_to_clock_t(tgutime);
894         tms->tms_stime = cputime_to_clock_t(tgstime);
895         tms->tms_cutime = cputime_to_clock_t(cutime);
896         tms->tms_cstime = cputime_to_clock_t(cstime);
897 }
898 
899 SYSCALL_DEFINE1(times, struct tms __user *, tbuf)
900 {
901         if (tbuf) {
902                 struct tms tmp;
903 
904                 do_sys_times(&tmp);
905                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
906                         return -EFAULT;
907         }
908         force_successful_syscall_return();
909         return (long) jiffies_64_to_clock_t(get_jiffies_64());
910 }
911 
912 /*
913  * This needs some heavy checking ...
914  * I just haven't the stomach for it. I also don't fully
915  * understand sessions/pgrp etc. Let somebody who does explain it.
916  *
917  * OK, I think I have the protection semantics right.... this is really
918  * only important on a multi-user system anyway, to make sure one user
919  * can't send a signal to a process owned by another.  -TYT, 12/12/91
920  *
921  * !PF_FORKNOEXEC check to conform completely to POSIX.
922  */
923 SYSCALL_DEFINE2(setpgid, pid_t, pid, pid_t, pgid)
924 {
925         struct task_struct *p;
926         struct task_struct *group_leader = current->group_leader;
927         struct pid *pgrp;
928         int err;
929 
930         if (!pid)
931                 pid = task_pid_vnr(group_leader);
932         if (!pgid)
933                 pgid = pid;
934         if (pgid < 0)
935                 return -EINVAL;
936         rcu_read_lock();
937 
938         /* From this point forward we keep holding onto the tasklist lock
939          * so that our parent does not change from under us. -DaveM
940          */
941         write_lock_irq(&tasklist_lock);
942 
943         err = -ESRCH;
944         p = find_task_by_vpid(pid);
945         if (!p)
946                 goto out;
947 
948         err = -EINVAL;
949         if (!thread_group_leader(p))
950                 goto out;
951 
952         if (same_thread_group(p->real_parent, group_leader)) {
953                 err = -EPERM;
954                 if (task_session(p) != task_session(group_leader))
955                         goto out;
956                 err = -EACCES;
957                 if (!(p->flags & PF_FORKNOEXEC))
958                         goto out;
959         } else {
960                 err = -ESRCH;
961                 if (p != group_leader)
962                         goto out;
963         }
964 
965         err = -EPERM;
966         if (p->signal->leader)
967                 goto out;
968 
969         pgrp = task_pid(p);
970         if (pgid != pid) {
971                 struct task_struct *g;
972 
973                 pgrp = find_vpid(pgid);
974                 g = pid_task(pgrp, PIDTYPE_PGID);
975                 if (!g || task_session(g) != task_session(group_leader))
976                         goto out;
977         }
978 
979         err = security_task_setpgid(p, pgid);
980         if (err)
981                 goto out;
982 
983         if (task_pgrp(p) != pgrp)
984                 change_pid(p, PIDTYPE_PGID, pgrp);
985 
986         err = 0;
987 out:
988         /* All paths lead to here, thus we are safe. -DaveM */
989         write_unlock_irq(&tasklist_lock);
990         rcu_read_unlock();
991         return err;
992 }
993 
994 SYSCALL_DEFINE1(getpgid, pid_t, pid)
995 {
996         struct task_struct *p;
997         struct pid *grp;
998         int retval;
999 
1000         rcu_read_lock();
1001         if (!pid)
1002                 grp = task_pgrp(current);
1003         else {
1004                 retval = -ESRCH;
1005                 p = find_task_by_vpid(pid);
1006                 if (!p)
1007                         goto out;
1008                 grp = task_pgrp(p);
1009                 if (!grp)
1010                         goto out;
1011 
1012                 retval = security_task_getpgid(p);
1013                 if (retval)
1014                         goto out;
1015         }
1016         retval = pid_vnr(grp);
1017 out:
1018         rcu_read_unlock();
1019         return retval;
1020 }
1021 
1022 #ifdef __ARCH_WANT_SYS_GETPGRP
1023 
1024 SYSCALL_DEFINE0(getpgrp)
1025 {
1026         return sys_getpgid(0);
1027 }
1028 
1029 #endif
1030 
1031 SYSCALL_DEFINE1(getsid, pid_t, pid)
1032 {
1033         struct task_struct *p;
1034         struct pid *sid;
1035         int retval;
1036 
1037         rcu_read_lock();
1038         if (!pid)
1039                 sid = task_session(current);
1040         else {
1041                 retval = -ESRCH;
1042                 p = find_task_by_vpid(pid);
1043                 if (!p)
1044                         goto out;
1045                 sid = task_session(p);
1046                 if (!sid)
1047                         goto out;
1048 
1049                 retval = security_task_getsid(p);
1050                 if (retval)
1051                         goto out;
1052         }
1053         retval = pid_vnr(sid);
1054 out:
1055         rcu_read_unlock();
1056         return retval;
1057 }
1058 
1059 static void set_special_pids(struct pid *pid)
1060 {
1061         struct task_struct *curr = current->group_leader;
1062 
1063         if (task_session(curr) != pid)
1064                 change_pid(curr, PIDTYPE_SID, pid);
1065 
1066         if (task_pgrp(curr) != pid)
1067                 change_pid(curr, PIDTYPE_PGID, pid);
1068 }
1069 
1070 SYSCALL_DEFINE0(setsid)
1071 {
1072         struct task_struct *group_leader = current->group_leader;
1073         struct pid *sid = task_pid(group_leader);
1074         pid_t session = pid_vnr(sid);
1075         int err = -EPERM;
1076 
1077         write_lock_irq(&tasklist_lock);
1078         /* Fail if I am already a session leader */
1079         if (group_leader->signal->leader)
1080                 goto out;
1081 
1082         /* Fail if a process group id already exists that equals the
1083          * proposed session id.
1084          */
1085         if (pid_task(sid, PIDTYPE_PGID))
1086                 goto out;
1087 
1088         group_leader->signal->leader = 1;
1089         set_special_pids(sid);
1090 
1091         proc_clear_tty(group_leader);
1092 
1093         err = session;
1094 out:
1095         write_unlock_irq(&tasklist_lock);
1096         if (err > 0) {
1097                 proc_sid_connector(group_leader);
1098                 sched_autogroup_create_attach(group_leader);
1099         }
1100         return err;
1101 }
1102 
1103 DECLARE_RWSEM(uts_sem);
1104 
1105 #ifdef COMPAT_UTS_MACHINE
1106 #define override_architecture(name) \
1107         (personality(current->personality) == PER_LINUX32 && \
1108          copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1109                       sizeof(COMPAT_UTS_MACHINE)))
1110 #else
1111 #define override_architecture(name)     0
1112 #endif
1113 
1114 /*
1115  * Work around broken programs that cannot handle "Linux 3.0".
1116  * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1117  * And we map 4.x to 2.6.60+x, so 4.0 would be 2.6.60.
1118  */
1119 static int override_release(char __user *release, size_t len)
1120 {
1121         int ret = 0;
1122 
1123         if (current->personality & UNAME26) {
1124                 const char *rest = UTS_RELEASE;
1125                 char buf[65] = { 0 };
1126                 int ndots = 0;
1127                 unsigned v;
1128                 size_t copy;
1129 
1130                 while (*rest) {
1131                         if (*rest == '.' && ++ndots >= 3)
1132                                 break;
1133                         if (!isdigit(*rest) && *rest != '.')
1134                                 break;
1135                         rest++;
1136                 }
1137                 v = ((LINUX_VERSION_CODE >> 8) & 0xff) + 60;
1138                 copy = clamp_t(size_t, len, 1, sizeof(buf));
1139                 copy = scnprintf(buf, copy, "2.6.%u%s", v, rest);
1140                 ret = copy_to_user(release, buf, copy + 1);
1141         }
1142         return ret;
1143 }
1144 
1145 SYSCALL_DEFINE1(newuname, struct new_utsname __user *, name)
1146 {
1147         int errno = 0;
1148 
1149         down_read(&uts_sem);
1150         if (copy_to_user(name, utsname(), sizeof *name))
1151                 errno = -EFAULT;
1152         up_read(&uts_sem);
1153 
1154         if (!errno && override_release(name->release, sizeof(name->release)))
1155                 errno = -EFAULT;
1156         if (!errno && override_architecture(name))
1157                 errno = -EFAULT;
1158         return errno;
1159 }
1160 
1161 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1162 /*
1163  * Old cruft
1164  */
1165 SYSCALL_DEFINE1(uname, struct old_utsname __user *, name)
1166 {
1167         int error = 0;
1168 
1169         if (!name)
1170                 return -EFAULT;
1171 
1172         down_read(&uts_sem);
1173         if (copy_to_user(name, utsname(), sizeof(*name)))
1174                 error = -EFAULT;
1175         up_read(&uts_sem);
1176 
1177         if (!error && override_release(name->release, sizeof(name->release)))
1178                 error = -EFAULT;
1179         if (!error && override_architecture(name))
1180                 error = -EFAULT;
1181         return error;
1182 }
1183 
1184 SYSCALL_DEFINE1(olduname, struct oldold_utsname __user *, name)
1185 {
1186         int error;
1187 
1188         if (!name)
1189                 return -EFAULT;
1190         if (!access_ok(VERIFY_WRITE, name, sizeof(struct oldold_utsname)))
1191                 return -EFAULT;
1192 
1193         down_read(&uts_sem);
1194         error = __copy_to_user(&name->sysname, &utsname()->sysname,
1195                                __OLD_UTS_LEN);
1196         error |= __put_user(0, name->sysname + __OLD_UTS_LEN);
1197         error |= __copy_to_user(&name->nodename, &utsname()->nodename,
1198                                 __OLD_UTS_LEN);
1199         error |= __put_user(0, name->nodename + __OLD_UTS_LEN);
1200         error |= __copy_to_user(&name->release, &utsname()->release,
1201                                 __OLD_UTS_LEN);
1202         error |= __put_user(0, name->release + __OLD_UTS_LEN);
1203         error |= __copy_to_user(&name->version, &utsname()->version,
1204                                 __OLD_UTS_LEN);
1205         error |= __put_user(0, name->version + __OLD_UTS_LEN);
1206         error |= __copy_to_user(&name->machine, &utsname()->machine,
1207                                 __OLD_UTS_LEN);
1208         error |= __put_user(0, name->machine + __OLD_UTS_LEN);
1209         up_read(&uts_sem);
1210 
1211         if (!error && override_architecture(name))
1212                 error = -EFAULT;
1213         if (!error && override_release(name->release, sizeof(name->release)))
1214                 error = -EFAULT;
1215         return error ? -EFAULT : 0;
1216 }
1217 #endif
1218 
1219 SYSCALL_DEFINE2(sethostname, char __user *, name, int, len)
1220 {
1221         int errno;
1222         char tmp[__NEW_UTS_LEN];
1223 
1224         if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1225                 return -EPERM;
1226 
1227         if (len < 0 || len > __NEW_UTS_LEN)
1228                 return -EINVAL;
1229         if (!ccs_capable(CCS_SYS_SETHOSTNAME))
1230                 return -EPERM;
1231         down_write(&uts_sem);
1232         errno = -EFAULT;
1233         if (!copy_from_user(tmp, name, len)) {
1234                 struct new_utsname *u = utsname();
1235 
1236                 memcpy(u->nodename, tmp, len);
1237                 memset(u->nodename + len, 0, sizeof(u->nodename) - len);
1238                 errno = 0;
1239                 uts_proc_notify(UTS_PROC_HOSTNAME);
1240         }
1241         up_write(&uts_sem);
1242         return errno;
1243 }
1244 
1245 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1246 
1247 SYSCALL_DEFINE2(gethostname, char __user *, name, int, len)
1248 {
1249         int i, errno;
1250         struct new_utsname *u;
1251 
1252         if (len < 0)
1253                 return -EINVAL;
1254         down_read(&uts_sem);
1255         u = utsname();
1256         i = 1 + strlen(u->nodename);
1257         if (i > len)
1258                 i = len;
1259         errno = 0;
1260         if (copy_to_user(name, u->nodename, i))
1261                 errno = -EFAULT;
1262         up_read(&uts_sem);
1263         return errno;
1264 }
1265 
1266 #endif
1267 
1268 /*
1269  * Only setdomainname; getdomainname can be implemented by calling
1270  * uname()
1271  */
1272 SYSCALL_DEFINE2(setdomainname, char __user *, name, int, len)
1273 {
1274         int errno;
1275         char tmp[__NEW_UTS_LEN];
1276 
1277         if (!ns_capable(current->nsproxy->uts_ns->user_ns, CAP_SYS_ADMIN))
1278                 return -EPERM;
1279         if (len < 0 || len > __NEW_UTS_LEN)
1280                 return -EINVAL;
1281         if (!ccs_capable(CCS_SYS_SETHOSTNAME))
1282                 return -EPERM;
1283 
1284         down_write(&uts_sem);
1285         errno = -EFAULT;
1286         if (!copy_from_user(tmp, name, len)) {
1287                 struct new_utsname *u = utsname();
1288 
1289                 memcpy(u->domainname, tmp, len);
1290                 memset(u->domainname + len, 0, sizeof(u->domainname) - len);
1291                 errno = 0;
1292                 uts_proc_notify(UTS_PROC_DOMAINNAME);
1293         }
1294         up_write(&uts_sem);
1295         return errno;
1296 }
1297 
1298 SYSCALL_DEFINE2(getrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1299 {
1300         struct rlimit value;
1301         int ret;
1302 
1303         ret = do_prlimit(current, resource, NULL, &value);
1304         if (!ret)
1305                 ret = copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1306 
1307         return ret;
1308 }
1309 
1310 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1311 
1312 /*
1313  *      Back compatibility for getrlimit. Needed for some apps.
1314  */
1315 SYSCALL_DEFINE2(old_getrlimit, unsigned int, resource,
1316                 struct rlimit __user *, rlim)
1317 {
1318         struct rlimit x;
1319         if (resource >= RLIM_NLIMITS)
1320                 return -EINVAL;
1321 
1322         task_lock(current->group_leader);
1323         x = current->signal->rlim[resource];
1324         task_unlock(current->group_leader);
1325         if (x.rlim_cur > 0x7FFFFFFF)
1326                 x.rlim_cur = 0x7FFFFFFF;
1327         if (x.rlim_max > 0x7FFFFFFF)
1328                 x.rlim_max = 0x7FFFFFFF;
1329         return copy_to_user(rlim, &x, sizeof(x)) ? -EFAULT : 0;
1330 }
1331 
1332 #endif
1333 
1334 static inline bool rlim64_is_infinity(__u64 rlim64)
1335 {
1336 #if BITS_PER_LONG < 64
1337         return rlim64 >= ULONG_MAX;
1338 #else
1339         return rlim64 == RLIM64_INFINITY;
1340 #endif
1341 }
1342 
1343 static void rlim_to_rlim64(const struct rlimit *rlim, struct rlimit64 *rlim64)
1344 {
1345         if (rlim->rlim_cur == RLIM_INFINITY)
1346                 rlim64->rlim_cur = RLIM64_INFINITY;
1347         else
1348                 rlim64->rlim_cur = rlim->rlim_cur;
1349         if (rlim->rlim_max == RLIM_INFINITY)
1350                 rlim64->rlim_max = RLIM64_INFINITY;
1351         else
1352                 rlim64->rlim_max = rlim->rlim_max;
1353 }
1354 
1355 static void rlim64_to_rlim(const struct rlimit64 *rlim64, struct rlimit *rlim)
1356 {
1357         if (rlim64_is_infinity(rlim64->rlim_cur))
1358                 rlim->rlim_cur = RLIM_INFINITY;
1359         else
1360                 rlim->rlim_cur = (unsigned long)rlim64->rlim_cur;
1361         if (rlim64_is_infinity(rlim64->rlim_max))
1362                 rlim->rlim_max = RLIM_INFINITY;
1363         else
1364                 rlim->rlim_max = (unsigned long)rlim64->rlim_max;
1365 }
1366 
1367 /* make sure you are allowed to change @tsk limits before calling this */
1368 int do_prlimit(struct task_struct *tsk, unsigned int resource,
1369                 struct rlimit *new_rlim, struct rlimit *old_rlim)
1370 {
1371         struct rlimit *rlim;
1372         int retval = 0;
1373 
1374         if (resource >= RLIM_NLIMITS)
1375                 return -EINVAL;
1376         if (new_rlim) {
1377                 if (new_rlim->rlim_cur > new_rlim->rlim_max)
1378                         return -EINVAL;
1379                 if (resource == RLIMIT_NOFILE &&
1380                                 new_rlim->rlim_max > sysctl_nr_open)
1381                         return -EPERM;
1382         }
1383 
1384         /* protect tsk->signal and tsk->sighand from disappearing */
1385         read_lock(&tasklist_lock);
1386         if (!tsk->sighand) {
1387                 retval = -ESRCH;
1388                 goto out;
1389         }
1390 
1391         rlim = tsk->signal->rlim + resource;
1392         task_lock(tsk->group_leader);
1393         if (new_rlim) {
1394                 /* Keep the capable check against init_user_ns until
1395                    cgroups can contain all limits */
1396                 if (new_rlim->rlim_max > rlim->rlim_max &&
1397                                 !capable(CAP_SYS_RESOURCE))
1398                         retval = -EPERM;
1399                 if (!retval)
1400                         retval = security_task_setrlimit(tsk->group_leader,
1401                                         resource, new_rlim);
1402                 if (resource == RLIMIT_CPU && new_rlim->rlim_cur == 0) {
1403                         /*
1404                          * The caller is asking for an immediate RLIMIT_CPU
1405                          * expiry.  But we use the zero value to mean "it was
1406                          * never set".  So let's cheat and make it one second
1407                          * instead
1408                          */
1409                         new_rlim->rlim_cur = 1;
1410                 }
1411         }
1412         if (!retval) {
1413                 if (old_rlim)
1414                         *old_rlim = *rlim;
1415                 if (new_rlim)
1416                         *rlim = *new_rlim;
1417         }
1418         task_unlock(tsk->group_leader);
1419 
1420         /*
1421          * RLIMIT_CPU handling.   Note that the kernel fails to return an error
1422          * code if it rejected the user's attempt to set RLIMIT_CPU.  This is a
1423          * very long-standing error, and fixing it now risks breakage of
1424          * applications, so we live with it
1425          */
1426          if (!retval && new_rlim && resource == RLIMIT_CPU &&
1427                          new_rlim->rlim_cur != RLIM_INFINITY)
1428                 update_rlimit_cpu(tsk, new_rlim->rlim_cur);
1429 out:
1430         read_unlock(&tasklist_lock);
1431         return retval;
1432 }
1433 
1434 /* rcu lock must be held */
1435 static int check_prlimit_permission(struct task_struct *task)
1436 {
1437         const struct cred *cred = current_cred(), *tcred;
1438 
1439         if (current == task)
1440                 return 0;
1441 
1442         tcred = __task_cred(task);
1443         if (uid_eq(cred->uid, tcred->euid) &&
1444             uid_eq(cred->uid, tcred->suid) &&
1445             uid_eq(cred->uid, tcred->uid)  &&
1446             gid_eq(cred->gid, tcred->egid) &&
1447             gid_eq(cred->gid, tcred->sgid) &&
1448             gid_eq(cred->gid, tcred->gid))
1449                 return 0;
1450         if (ns_capable(tcred->user_ns, CAP_SYS_RESOURCE))
1451                 return 0;
1452 
1453         return -EPERM;
1454 }
1455 
1456 SYSCALL_DEFINE4(prlimit64, pid_t, pid, unsigned int, resource,
1457                 const struct rlimit64 __user *, new_rlim,
1458                 struct rlimit64 __user *, old_rlim)
1459 {
1460         struct rlimit64 old64, new64;
1461         struct rlimit old, new;
1462         struct task_struct *tsk;
1463         int ret;
1464 
1465         if (new_rlim) {
1466                 if (copy_from_user(&new64, new_rlim, sizeof(new64)))
1467                         return -EFAULT;
1468                 rlim64_to_rlim(&new64, &new);
1469         }
1470 
1471         rcu_read_lock();
1472         tsk = pid ? find_task_by_vpid(pid) : current;
1473         if (!tsk) {
1474                 rcu_read_unlock();
1475                 return -ESRCH;
1476         }
1477         ret = check_prlimit_permission(tsk);
1478         if (ret) {
1479                 rcu_read_unlock();
1480                 return ret;
1481         }
1482         get_task_struct(tsk);
1483         rcu_read_unlock();
1484 
1485         ret = do_prlimit(tsk, resource, new_rlim ? &new : NULL,
1486                         old_rlim ? &old : NULL);
1487 
1488         if (!ret && old_rlim) {
1489                 rlim_to_rlim64(&old, &old64);
1490                 if (copy_to_user(old_rlim, &old64, sizeof(old64)))
1491                         ret = -EFAULT;
1492         }
1493 
1494         put_task_struct(tsk);
1495         return ret;
1496 }
1497 
1498 SYSCALL_DEFINE2(setrlimit, unsigned int, resource, struct rlimit __user *, rlim)
1499 {
1500         struct rlimit new_rlim;
1501 
1502         if (copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1503                 return -EFAULT;
1504         return do_prlimit(current, resource, &new_rlim, NULL);
1505 }
1506 
1507 /*
1508  * It would make sense to put struct rusage in the task_struct,
1509  * except that would make the task_struct be *really big*.  After
1510  * task_struct gets moved into malloc'ed memory, it would
1511  * make sense to do this.  It will make moving the rest of the information
1512  * a lot simpler!  (Which we're not doing right now because we're not
1513  * measuring them yet).
1514  *
1515  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1516  * races with threads incrementing their own counters.  But since word
1517  * reads are atomic, we either get new values or old values and we don't
1518  * care which for the sums.  We always take the siglock to protect reading
1519  * the c* fields from p->signal from races with exit.c updating those
1520  * fields when reaping, so a sample either gets all the additions of a
1521  * given child after it's reaped, or none so this sample is before reaping.
1522  *
1523  * Locking:
1524  * We need to take the siglock for CHILDEREN, SELF and BOTH
1525  * for  the cases current multithreaded, non-current single threaded
1526  * non-current multithreaded.  Thread traversal is now safe with
1527  * the siglock held.
1528  * Strictly speaking, we donot need to take the siglock if we are current and
1529  * single threaded,  as no one else can take our signal_struct away, no one
1530  * else can  reap the  children to update signal->c* counters, and no one else
1531  * can race with the signal-> fields. If we do not take any lock, the
1532  * signal-> fields could be read out of order while another thread was just
1533  * exiting. So we should  place a read memory barrier when we avoid the lock.
1534  * On the writer side,  write memory barrier is implied in  __exit_signal
1535  * as __exit_signal releases  the siglock spinlock after updating the signal->
1536  * fields. But we don't do this yet to keep things simple.
1537  *
1538  */
1539 
1540 static void accumulate_thread_rusage(struct task_struct *t, struct rusage *r)
1541 {
1542         r->ru_nvcsw += t->nvcsw;
1543         r->ru_nivcsw += t->nivcsw;
1544         r->ru_minflt += t->min_flt;
1545         r->ru_majflt += t->maj_flt;
1546         r->ru_inblock += task_io_get_inblock(t);
1547         r->ru_oublock += task_io_get_oublock(t);
1548 }
1549 
1550 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1551 {
1552         struct task_struct *t;
1553         unsigned long flags;
1554         cputime_t tgutime, tgstime, utime, stime;
1555         unsigned long maxrss = 0;
1556 
1557         memset((char *)r, 0, sizeof (*r));
1558         utime = stime = 0;
1559 
1560         if (who == RUSAGE_THREAD) {
1561                 task_cputime_adjusted(current, &utime, &stime);
1562                 accumulate_thread_rusage(p, r);
1563                 maxrss = p->signal->maxrss;
1564                 goto out;
1565         }
1566 
1567         if (!lock_task_sighand(p, &flags))
1568                 return;
1569 
1570         switch (who) {
1571         case RUSAGE_BOTH:
1572         case RUSAGE_CHILDREN:
1573                 utime = p->signal->cutime;
1574                 stime = p->signal->cstime;
1575                 r->ru_nvcsw = p->signal->cnvcsw;
1576                 r->ru_nivcsw = p->signal->cnivcsw;
1577                 r->ru_minflt = p->signal->cmin_flt;
1578                 r->ru_majflt = p->signal->cmaj_flt;
1579                 r->ru_inblock = p->signal->cinblock;
1580                 r->ru_oublock = p->signal->coublock;
1581                 maxrss = p->signal->cmaxrss;
1582 
1583                 if (who == RUSAGE_CHILDREN)
1584                         break;
1585 
1586         case RUSAGE_SELF:
1587                 thread_group_cputime_adjusted(p, &tgutime, &tgstime);
1588                 utime += tgutime;
1589                 stime += tgstime;
1590                 r->ru_nvcsw += p->signal->nvcsw;
1591                 r->ru_nivcsw += p->signal->nivcsw;
1592                 r->ru_minflt += p->signal->min_flt;
1593                 r->ru_majflt += p->signal->maj_flt;
1594                 r->ru_inblock += p->signal->inblock;
1595                 r->ru_oublock += p->signal->oublock;
1596                 if (maxrss < p->signal->maxrss)
1597                         maxrss = p->signal->maxrss;
1598                 t = p;
1599                 do {
1600                         accumulate_thread_rusage(t, r);
1601                 } while_each_thread(p, t);
1602                 break;
1603 
1604         default:
1605                 BUG();
1606         }
1607         unlock_task_sighand(p, &flags);
1608 
1609 out:
1610         cputime_to_timeval(utime, &r->ru_utime);
1611         cputime_to_timeval(stime, &r->ru_stime);
1612 
1613         if (who != RUSAGE_CHILDREN) {
1614                 struct mm_struct *mm = get_task_mm(p);
1615 
1616                 if (mm) {
1617                         setmax_mm_hiwater_rss(&maxrss, mm);
1618                         mmput(mm);
1619                 }
1620         }
1621         r->ru_maxrss = maxrss * (PAGE_SIZE / 1024); /* convert pages to KBs */
1622 }
1623 
1624 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1625 {
1626         struct rusage r;
1627 
1628         k_getrusage(p, who, &r);
1629         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1630 }
1631 
1632 SYSCALL_DEFINE2(getrusage, int, who, struct rusage __user *, ru)
1633 {
1634         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1635             who != RUSAGE_THREAD)
1636                 return -EINVAL;
1637         return getrusage(current, who, ru);
1638 }
1639 
1640 #ifdef CONFIG_COMPAT
1641 COMPAT_SYSCALL_DEFINE2(getrusage, int, who, struct compat_rusage __user *, ru)
1642 {
1643         struct rusage r;
1644 
1645         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN &&
1646             who != RUSAGE_THREAD)
1647                 return -EINVAL;
1648 
1649         k_getrusage(current, who, &r);
1650         return put_compat_rusage(&r, ru);
1651 }
1652 #endif
1653 
1654 SYSCALL_DEFINE1(umask, int, mask)
1655 {
1656         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1657         return mask;
1658 }
1659 
1660 static int prctl_set_mm_exe_file(struct mm_struct *mm, unsigned int fd)
1661 {
1662         struct fd exe;
1663         struct file *old_exe, *exe_file;
1664         struct inode *inode;
1665         int err;
1666 
1667         exe = fdget(fd);
1668         if (!exe.file)
1669                 return -EBADF;
1670 
1671         inode = file_inode(exe.file);
1672 
1673         /*
1674          * Because the original mm->exe_file points to executable file, make
1675          * sure that this one is executable as well, to avoid breaking an
1676          * overall picture.
1677          */
1678         err = -EACCES;
1679         if (!S_ISREG(inode->i_mode) || path_noexec(&exe.file->f_path))
1680                 goto exit;
1681 
1682         err = inode_permission(inode, MAY_EXEC);
1683         if (err)
1684                 goto exit;
1685 
1686         /*
1687          * Forbid mm->exe_file change if old file still mapped.
1688          */
1689         exe_file = get_mm_exe_file(mm);
1690         err = -EBUSY;
1691         if (exe_file) {
1692                 struct vm_area_struct *vma;
1693 
1694                 down_read(&mm->mmap_sem);
1695                 for (vma = mm->mmap; vma; vma = vma->vm_next) {
1696                         if (!vma->vm_file)
1697                                 continue;
1698                         if (path_equal(&vma->vm_file->f_path,
1699                                        &exe_file->f_path))
1700                                 goto exit_err;
1701                 }
1702 
1703                 up_read(&mm->mmap_sem);
1704                 fput(exe_file);
1705         }
1706 
1707         /*
1708          * The symlink can be changed only once, just to disallow arbitrary
1709          * transitions malicious software might bring in. This means one
1710          * could make a snapshot over all processes running and monitor
1711          * /proc/pid/exe changes to notice unusual activity if needed.
1712          */
1713         err = -EPERM;
1714         if (test_and_set_bit(MMF_EXE_FILE_CHANGED, &mm->flags))
1715                 goto exit;
1716 
1717         err = 0;
1718         /* set the new file, lockless */
1719         get_file(exe.file);
1720         old_exe = xchg(&mm->exe_file, exe.file);
1721         if (old_exe)
1722                 fput(old_exe);
1723 exit:
1724         fdput(exe);
1725         return err;
1726 exit_err:
1727         up_read(&mm->mmap_sem);
1728         fput(exe_file);
1729         goto exit;
1730 }
1731 
1732 /*
1733  * WARNING: we don't require any capability here so be very careful
1734  * in what is allowed for modification from userspace.
1735  */
1736 static int validate_prctl_map(struct prctl_mm_map *prctl_map)
1737 {
1738         unsigned long mmap_max_addr = TASK_SIZE;
1739         struct mm_struct *mm = current->mm;
1740         int error = -EINVAL, i;
1741 
1742         static const unsigned char offsets[] = {
1743                 offsetof(struct prctl_mm_map, start_code),
1744                 offsetof(struct prctl_mm_map, end_code),
1745                 offsetof(struct prctl_mm_map, start_data),
1746                 offsetof(struct prctl_mm_map, end_data),
1747                 offsetof(struct prctl_mm_map, start_brk),
1748                 offsetof(struct prctl_mm_map, brk),
1749                 offsetof(struct prctl_mm_map, start_stack),
1750                 offsetof(struct prctl_mm_map, arg_start),
1751                 offsetof(struct prctl_mm_map, arg_end),
1752                 offsetof(struct prctl_mm_map, env_start),
1753                 offsetof(struct prctl_mm_map, env_end),
1754         };
1755 
1756         /*
1757          * Make sure the members are not somewhere outside
1758          * of allowed address space.
1759          */
1760         for (i = 0; i < ARRAY_SIZE(offsets); i++) {
1761                 u64 val = *(u64 *)((char *)prctl_map + offsets[i]);
1762 
1763                 if ((unsigned long)val >= mmap_max_addr ||
1764                     (unsigned long)val < mmap_min_addr)
1765                         goto out;
1766         }
1767 
1768         /*
1769          * Make sure the pairs are ordered.
1770          */
1771 #define __prctl_check_order(__m1, __op, __m2)                           \
1772         ((unsigned long)prctl_map->__m1 __op                            \
1773          (unsigned long)prctl_map->__m2) ? 0 : -EINVAL
1774         error  = __prctl_check_order(start_code, <, end_code);
1775         error |= __prctl_check_order(start_data, <, end_data);
1776         error |= __prctl_check_order(start_brk, <=, brk);
1777         error |= __prctl_check_order(arg_start, <=, arg_end);
1778         error |= __prctl_check_order(env_start, <=, env_end);
1779         if (error)
1780                 goto out;
1781 #undef __prctl_check_order
1782 
1783         error = -EINVAL;
1784 
1785         /*
1786          * @brk should be after @end_data in traditional maps.
1787          */
1788         if (prctl_map->start_brk <= prctl_map->end_data ||
1789             prctl_map->brk <= prctl_map->end_data)
1790                 goto out;
1791 
1792         /*
1793          * Neither we should allow to override limits if they set.
1794          */
1795         if (check_data_rlimit(rlimit(RLIMIT_DATA), prctl_map->brk,
1796                               prctl_map->start_brk, prctl_map->end_data,
1797                               prctl_map->start_data))
1798                         goto out;
1799 
1800         /*
1801          * Someone is trying to cheat the auxv vector.
1802          */
1803         if (prctl_map->auxv_size) {
1804                 if (!prctl_map->auxv || prctl_map->auxv_size > sizeof(mm->saved_auxv))
1805                         goto out;
1806         }
1807 
1808         /*
1809          * Finally, make sure the caller has the rights to
1810          * change /proc/pid/exe link: only local root should
1811          * be allowed to.
1812          */
1813         if (prctl_map->exe_fd != (u32)-1) {
1814                 struct user_namespace *ns = current_user_ns();
1815                 const struct cred *cred = current_cred();
1816 
1817                 if (!uid_eq(cred->uid, make_kuid(ns, 0)) ||
1818                     !gid_eq(cred->gid, make_kgid(ns, 0)))
1819                         goto out;
1820         }
1821 
1822         error = 0;
1823 out:
1824         return error;
1825 }
1826 
1827 #ifdef CONFIG_CHECKPOINT_RESTORE
1828 static int prctl_set_mm_map(int opt, const void __user *addr, unsigned long data_size)
1829 {
1830         struct prctl_mm_map prctl_map = { .exe_fd = (u32)-1, };
1831         unsigned long user_auxv[AT_VECTOR_SIZE];
1832         struct mm_struct *mm = current->mm;
1833         int error;
1834 
1835         BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1836         BUILD_BUG_ON(sizeof(struct prctl_mm_map) > 256);
1837 
1838         if (opt == PR_SET_MM_MAP_SIZE)
1839                 return put_user((unsigned int)sizeof(prctl_map),
1840                                 (unsigned int __user *)addr);
1841 
1842         if (data_size != sizeof(prctl_map))
1843                 return -EINVAL;
1844 
1845         if (copy_from_user(&prctl_map, addr, sizeof(prctl_map)))
1846                 return -EFAULT;
1847 
1848         error = validate_prctl_map(&prctl_map);
1849         if (error)
1850                 return error;
1851 
1852         if (prctl_map.auxv_size) {
1853                 memset(user_auxv, 0, sizeof(user_auxv));
1854                 if (copy_from_user(user_auxv,
1855                                    (const void __user *)prctl_map.auxv,
1856                                    prctl_map.auxv_size))
1857                         return -EFAULT;
1858 
1859                 /* Last entry must be AT_NULL as specification requires */
1860                 user_auxv[AT_VECTOR_SIZE - 2] = AT_NULL;
1861                 user_auxv[AT_VECTOR_SIZE - 1] = AT_NULL;
1862         }
1863 
1864         if (prctl_map.exe_fd != (u32)-1) {
1865                 error = prctl_set_mm_exe_file(mm, prctl_map.exe_fd);
1866                 if (error)
1867                         return error;
1868         }
1869 
1870         down_write(&mm->mmap_sem);
1871 
1872         /*
1873          * We don't validate if these members are pointing to
1874          * real present VMAs because application may have correspond
1875          * VMAs already unmapped and kernel uses these members for statistics
1876          * output in procfs mostly, except
1877          *
1878          *  - @start_brk/@brk which are used in do_brk but kernel lookups
1879          *    for VMAs when updating these memvers so anything wrong written
1880          *    here cause kernel to swear at userspace program but won't lead
1881          *    to any problem in kernel itself
1882          */
1883 
1884         mm->start_code  = prctl_map.start_code;
1885         mm->end_code    = prctl_map.end_code;
1886         mm->start_data  = prctl_map.start_data;
1887         mm->end_data    = prctl_map.end_data;
1888         mm->start_brk   = prctl_map.start_brk;
1889         mm->brk         = prctl_map.brk;
1890         mm->start_stack = prctl_map.start_stack;
1891         mm->arg_start   = prctl_map.arg_start;
1892         mm->arg_end     = prctl_map.arg_end;
1893         mm->env_start   = prctl_map.env_start;
1894         mm->env_end     = prctl_map.env_end;
1895 
1896         /*
1897          * Note this update of @saved_auxv is lockless thus
1898          * if someone reads this member in procfs while we're
1899          * updating -- it may get partly updated results. It's
1900          * known and acceptable trade off: we leave it as is to
1901          * not introduce additional locks here making the kernel
1902          * more complex.
1903          */
1904         if (prctl_map.auxv_size)
1905                 memcpy(mm->saved_auxv, user_auxv, sizeof(user_auxv));
1906 
1907         up_write(&mm->mmap_sem);
1908         return 0;
1909 }
1910 #endif /* CONFIG_CHECKPOINT_RESTORE */
1911 
1912 static int prctl_set_auxv(struct mm_struct *mm, unsigned long addr,
1913                           unsigned long len)
1914 {
1915         /*
1916          * This doesn't move the auxiliary vector itself since it's pinned to
1917          * mm_struct, but it permits filling the vector with new values.  It's
1918          * up to the caller to provide sane values here, otherwise userspace
1919          * tools which use this vector might be unhappy.
1920          */
1921         unsigned long user_auxv[AT_VECTOR_SIZE];
1922 
1923         if (len > sizeof(user_auxv))
1924                 return -EINVAL;
1925 
1926         if (copy_from_user(user_auxv, (const void __user *)addr, len))
1927                 return -EFAULT;
1928 
1929         /* Make sure the last entry is always AT_NULL */
1930         user_auxv[AT_VECTOR_SIZE - 2] = 0;
1931         user_auxv[AT_VECTOR_SIZE - 1] = 0;
1932 
1933         BUILD_BUG_ON(sizeof(user_auxv) != sizeof(mm->saved_auxv));
1934 
1935         task_lock(current);
1936         memcpy(mm->saved_auxv, user_auxv, len);
1937         task_unlock(current);
1938 
1939         return 0;
1940 }
1941 
1942 static int prctl_set_mm(int opt, unsigned long addr,
1943                         unsigned long arg4, unsigned long arg5)
1944 {
1945         struct mm_struct *mm = current->mm;
1946         struct prctl_mm_map prctl_map;
1947         struct vm_area_struct *vma;
1948         int error;
1949 
1950         if (arg5 || (arg4 && (opt != PR_SET_MM_AUXV &&
1951                               opt != PR_SET_MM_MAP &&
1952                               opt != PR_SET_MM_MAP_SIZE)))
1953                 return -EINVAL;
1954 
1955 #ifdef CONFIG_CHECKPOINT_RESTORE
1956         if (opt == PR_SET_MM_MAP || opt == PR_SET_MM_MAP_SIZE)
1957                 return prctl_set_mm_map(opt, (const void __user *)addr, arg4);
1958 #endif
1959 
1960         if (!capable(CAP_SYS_RESOURCE))
1961                 return -EPERM;
1962 
1963         if (opt == PR_SET_MM_EXE_FILE)
1964                 return prctl_set_mm_exe_file(mm, (unsigned int)addr);
1965 
1966         if (opt == PR_SET_MM_AUXV)
1967                 return prctl_set_auxv(mm, addr, arg4);
1968 
1969         if (addr >= TASK_SIZE || addr < mmap_min_addr)
1970                 return -EINVAL;
1971 
1972         error = -EINVAL;
1973 
1974         down_write(&mm->mmap_sem);
1975         vma = find_vma(mm, addr);
1976 
1977         prctl_map.start_code    = mm->start_code;
1978         prctl_map.end_code      = mm->end_code;
1979         prctl_map.start_data    = mm->start_data;
1980         prctl_map.end_data      = mm->end_data;
1981         prctl_map.start_brk     = mm->start_brk;
1982         prctl_map.brk           = mm->brk;
1983         prctl_map.start_stack   = mm->start_stack;
1984         prctl_map.arg_start     = mm->arg_start;
1985         prctl_map.arg_end       = mm->arg_end;
1986         prctl_map.env_start     = mm->env_start;
1987         prctl_map.env_end       = mm->env_end;
1988         prctl_map.auxv          = NULL;
1989         prctl_map.auxv_size     = 0;
1990         prctl_map.exe_fd        = -1;
1991 
1992         switch (opt) {
1993         case PR_SET_MM_START_CODE:
1994                 prctl_map.start_code = addr;
1995                 break;
1996         case PR_SET_MM_END_CODE:
1997                 prctl_map.end_code = addr;
1998                 break;
1999         case PR_SET_MM_START_DATA:
2000                 prctl_map.start_data = addr;
2001                 break;
2002         case PR_SET_MM_END_DATA:
2003                 prctl_map.end_data = addr;
2004                 break;
2005         case PR_SET_MM_START_STACK:
2006                 prctl_map.start_stack = addr;
2007                 break;
2008         case PR_SET_MM_START_BRK:
2009                 prctl_map.start_brk = addr;
2010                 break;
2011         case PR_SET_MM_BRK:
2012                 prctl_map.brk = addr;
2013                 break;
2014         case PR_SET_MM_ARG_START:
2015                 prctl_map.arg_start = addr;
2016                 break;
2017         case PR_SET_MM_ARG_END:
2018                 prctl_map.arg_end = addr;
2019                 break;
2020         case PR_SET_MM_ENV_START:
2021                 prctl_map.env_start = addr;
2022                 break;
2023         case PR_SET_MM_ENV_END:
2024                 prctl_map.env_end = addr;
2025                 break;
2026         default:
2027                 goto out;
2028         }
2029 
2030         error = validate_prctl_map(&prctl_map);
2031         if (error)
2032                 goto out;
2033 
2034         switch (opt) {
2035         /*
2036          * If command line arguments and environment
2037          * are placed somewhere else on stack, we can
2038          * set them up here, ARG_START/END to setup
2039          * command line argumets and ENV_START/END
2040          * for environment.
2041          */
2042         case PR_SET_MM_START_STACK:
2043         case PR_SET_MM_ARG_START:
2044         case PR_SET_MM_ARG_END:
2045         case PR_SET_MM_ENV_START:
2046         case PR_SET_MM_ENV_END:
2047                 if (!vma) {
2048                         error = -EFAULT;
2049                         goto out;
2050                 }
2051         }
2052 
2053         mm->start_code  = prctl_map.start_code;
2054         mm->end_code    = prctl_map.end_code;
2055         mm->start_data  = prctl_map.start_data;
2056         mm->end_data    = prctl_map.end_data;
2057         mm->start_brk   = prctl_map.start_brk;
2058         mm->brk         = prctl_map.brk;
2059         mm->start_stack = prctl_map.start_stack;
2060         mm->arg_start   = prctl_map.arg_start;
2061         mm->arg_end     = prctl_map.arg_end;
2062         mm->env_start   = prctl_map.env_start;
2063         mm->env_end     = prctl_map.env_end;
2064 
2065         error = 0;
2066 out:
2067         up_write(&mm->mmap_sem);
2068         return error;
2069 }
2070 
2071 #ifdef CONFIG_CHECKPOINT_RESTORE
2072 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2073 {
2074         return put_user(me->clear_child_tid, tid_addr);
2075 }
2076 #else
2077 static int prctl_get_tid_address(struct task_struct *me, int __user **tid_addr)
2078 {
2079         return -EINVAL;
2080 }
2081 #endif
2082 
2083 SYSCALL_DEFINE5(prctl, int, option, unsigned long, arg2, unsigned long, arg3,
2084                 unsigned long, arg4, unsigned long, arg5)
2085 {
2086         struct task_struct *me = current;
2087         unsigned char comm[sizeof(me->comm)];
2088         long error;
2089 
2090         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
2091         if (error != -ENOSYS)
2092                 return error;
2093 
2094         error = 0;
2095         switch (option) {
2096         case PR_SET_PDEATHSIG:
2097                 if (!valid_signal(arg2)) {
2098                         error = -EINVAL;
2099                         break;
2100                 }
2101                 me->pdeath_signal = arg2;
2102                 break;
2103         case PR_GET_PDEATHSIG:
2104                 error = put_user(me->pdeath_signal, (int __user *)arg2);
2105                 break;
2106         case PR_GET_DUMPABLE:
2107                 error = get_dumpable(me->mm);
2108                 break;
2109         case PR_SET_DUMPABLE:
2110                 if (arg2 != SUID_DUMP_DISABLE && arg2 != SUID_DUMP_USER) {
2111                         error = -EINVAL;
2112                         break;
2113                 }
2114                 set_dumpable(me->mm, arg2);
2115                 break;
2116 
2117         case PR_SET_UNALIGN:
2118                 error = SET_UNALIGN_CTL(me, arg2);
2119                 break;
2120         case PR_GET_UNALIGN:
2121                 error = GET_UNALIGN_CTL(me, arg2);
2122                 break;
2123         case PR_SET_FPEMU:
2124                 error = SET_FPEMU_CTL(me, arg2);
2125                 break;
2126         case PR_GET_FPEMU:
2127                 error = GET_FPEMU_CTL(me, arg2);
2128                 break;
2129         case PR_SET_FPEXC:
2130                 error = SET_FPEXC_CTL(me, arg2);
2131                 break;
2132         case PR_GET_FPEXC:
2133                 error = GET_FPEXC_CTL(me, arg2);
2134                 break;
2135         case PR_GET_TIMING:
2136                 error = PR_TIMING_STATISTICAL;
2137                 break;
2138         case PR_SET_TIMING:
2139                 if (arg2 != PR_TIMING_STATISTICAL)
2140                         error = -EINVAL;
2141                 break;
2142         case PR_SET_NAME:
2143                 comm[sizeof(me->comm) - 1] = 0;
2144                 if (strncpy_from_user(comm, (char __user *)arg2,
2145                                       sizeof(me->comm) - 1) < 0)
2146                         return -EFAULT;
2147                 set_task_comm(me, comm);
2148                 proc_comm_connector(me);
2149                 break;
2150         case PR_GET_NAME:
2151                 get_task_comm(comm, me);
2152                 if (copy_to_user((char __user *)arg2, comm, sizeof(comm)))
2153                         return -EFAULT;
2154                 break;
2155         case PR_GET_ENDIAN:
2156                 error = GET_ENDIAN(me, arg2);
2157                 break;
2158         case PR_SET_ENDIAN:
2159                 error = SET_ENDIAN(me, arg2);
2160                 break;
2161         case PR_GET_SECCOMP:
2162                 error = prctl_get_seccomp();
2163                 break;
2164         case PR_SET_SECCOMP:
2165                 error = prctl_set_seccomp(arg2, (char __user *)arg3);
2166                 break;
2167         case PR_GET_TSC:
2168                 error = GET_TSC_CTL(arg2);
2169                 break;
2170         case PR_SET_TSC:
2171                 error = SET_TSC_CTL(arg2);
2172                 break;
2173         case PR_TASK_PERF_EVENTS_DISABLE:
2174                 error = perf_event_task_disable();
2175                 break;
2176         case PR_TASK_PERF_EVENTS_ENABLE:
2177                 error = perf_event_task_enable();
2178                 break;
2179         case PR_GET_TIMERSLACK:
2180                 if (current->timer_slack_ns > ULONG_MAX)
2181                         error = ULONG_MAX;
2182                 else
2183                         error = current->timer_slack_ns;
2184                 break;
2185         case PR_SET_TIMERSLACK:
2186                 if (arg2 <= 0)
2187                         current->timer_slack_ns =
2188                                         current->default_timer_slack_ns;
2189                 else
2190                         current->timer_slack_ns = arg2;
2191                 break;
2192         case PR_MCE_KILL:
2193                 if (arg4 | arg5)
2194                         return -EINVAL;
2195                 switch (arg2) {
2196                 case PR_MCE_KILL_CLEAR:
2197                         if (arg3 != 0)
2198                                 return -EINVAL;
2199                         current->flags &= ~PF_MCE_PROCESS;
2200                         break;
2201                 case PR_MCE_KILL_SET:
2202                         current->flags |= PF_MCE_PROCESS;
2203                         if (arg3 == PR_MCE_KILL_EARLY)
2204                                 current->flags |= PF_MCE_EARLY;
2205                         else if (arg3 == PR_MCE_KILL_LATE)
2206                                 current->flags &= ~PF_MCE_EARLY;
2207                         else if (arg3 == PR_MCE_KILL_DEFAULT)
2208                                 current->flags &=
2209                                                 ~(PF_MCE_EARLY|PF_MCE_PROCESS);
2210                         else
2211                                 return -EINVAL;
2212                         break;
2213                 default:
2214                         return -EINVAL;
2215                 }
2216                 break;
2217         case PR_MCE_KILL_GET:
2218                 if (arg2 | arg3 | arg4 | arg5)
2219                         return -EINVAL;
2220                 if (current->flags & PF_MCE_PROCESS)
2221                         error = (current->flags & PF_MCE_EARLY) ?
2222                                 PR_MCE_KILL_EARLY : PR_MCE_KILL_LATE;
2223                 else
2224                         error = PR_MCE_KILL_DEFAULT;
2225                 break;
2226         case PR_SET_MM:
2227                 error = prctl_set_mm(arg2, arg3, arg4, arg5);
2228                 break;
2229         case PR_GET_TID_ADDRESS:
2230                 error = prctl_get_tid_address(me, (int __user **)arg2);
2231                 break;
2232         case PR_SET_CHILD_SUBREAPER:
2233                 me->signal->is_child_subreaper = !!arg2;
2234                 break;
2235         case PR_GET_CHILD_SUBREAPER:
2236                 error = put_user(me->signal->is_child_subreaper,
2237                                  (int __user *)arg2);
2238                 break;
2239         case PR_SET_NO_NEW_PRIVS:
2240                 if (arg2 != 1 || arg3 || arg4 || arg5)
2241                         return -EINVAL;
2242 
2243                 task_set_no_new_privs(current);
2244                 break;
2245         case PR_GET_NO_NEW_PRIVS:
2246                 if (arg2 || arg3 || arg4 || arg5)
2247                         return -EINVAL;
2248                 return task_no_new_privs(current) ? 1 : 0;
2249         case PR_GET_THP_DISABLE:
2250                 if (arg2 || arg3 || arg4 || arg5)
2251                         return -EINVAL;
2252                 error = !!(me->mm->def_flags & VM_NOHUGEPAGE);
2253                 break;
2254         case PR_SET_THP_DISABLE:
2255                 if (arg3 || arg4 || arg5)
2256                         return -EINVAL;
2257                 down_write(&me->mm->mmap_sem);
2258                 if (arg2)
2259                         me->mm->def_flags |= VM_NOHUGEPAGE;
2260                 else
2261                         me->mm->def_flags &= ~VM_NOHUGEPAGE;
2262                 up_write(&me->mm->mmap_sem);
2263                 break;
2264         case PR_MPX_ENABLE_MANAGEMENT:
2265                 if (arg2 || arg3 || arg4 || arg5)
2266                         return -EINVAL;
2267                 error = MPX_ENABLE_MANAGEMENT();
2268                 break;
2269         case PR_MPX_DISABLE_MANAGEMENT:
2270                 if (arg2 || arg3 || arg4 || arg5)
2271                         return -EINVAL;
2272                 error = MPX_DISABLE_MANAGEMENT();
2273                 break;
2274         case PR_SET_FP_MODE:
2275                 error = SET_FP_MODE(me, arg2);
2276                 break;
2277         case PR_GET_FP_MODE:
2278                 error = GET_FP_MODE(me);
2279                 break;
2280         default:
2281                 error = -EINVAL;
2282                 break;
2283         }
2284         return error;
2285 }
2286 
2287 SYSCALL_DEFINE3(getcpu, unsigned __user *, cpup, unsigned __user *, nodep,
2288                 struct getcpu_cache __user *, unused)
2289 {
2290         int err = 0;
2291         int cpu = raw_smp_processor_id();
2292 
2293         if (cpup)
2294                 err |= put_user(cpu, cpup);
2295         if (nodep)
2296                 err |= put_user(cpu_to_node(cpu), nodep);
2297         return err ? -EFAULT : 0;
2298 }
2299 
2300 /**
2301  * do_sysinfo - fill in sysinfo struct
2302  * @info: pointer to buffer to fill
2303  */
2304 static int do_sysinfo(struct sysinfo *info)
2305 {
2306         unsigned long mem_total, sav_total;
2307         unsigned int mem_unit, bitcount;
2308         struct timespec tp;
2309 
2310         memset(info, 0, sizeof(struct sysinfo));
2311 
2312         get_monotonic_boottime(&tp);
2313         info->uptime = tp.tv_sec + (tp.tv_nsec ? 1 : 0);
2314 
2315         get_avenrun(info->loads, 0, SI_LOAD_SHIFT - FSHIFT);
2316 
2317         info->procs = nr_threads;
2318 
2319         si_meminfo(info);
2320         si_swapinfo(info);
2321 
2322         /*
2323          * If the sum of all the available memory (i.e. ram + swap)
2324          * is less than can be stored in a 32 bit unsigned long then
2325          * we can be binary compatible with 2.2.x kernels.  If not,
2326          * well, in that case 2.2.x was broken anyways...
2327          *
2328          *  -Erik Andersen <andersee@debian.org>
2329          */
2330 
2331         mem_total = info->totalram + info->totalswap;
2332         if (mem_total < info->totalram || mem_total < info->totalswap)
2333                 goto out;
2334         bitcount = 0;
2335         mem_unit = info->mem_unit;
2336         while (mem_unit > 1) {
2337                 bitcount++;
2338                 mem_unit >>= 1;
2339                 sav_total = mem_total;
2340                 mem_total <<= 1;
2341                 if (mem_total < sav_total)
2342                         goto out;
2343         }
2344 
2345         /*
2346          * If mem_total did not overflow, multiply all memory values by
2347          * info->mem_unit and set it to 1.  This leaves things compatible
2348          * with 2.2.x, and also retains compatibility with earlier 2.4.x
2349          * kernels...
2350          */
2351 
2352         info->mem_unit = 1;
2353         info->totalram <<= bitcount;
2354         info->freeram <<= bitcount;
2355         info->sharedram <<= bitcount;
2356         info->bufferram <<= bitcount;
2357         info->totalswap <<= bitcount;
2358         info->freeswap <<= bitcount;
2359         info->totalhigh <<= bitcount;
2360         info->freehigh <<= bitcount;
2361 
2362 out:
2363         return 0;
2364 }
2365 
2366 SYSCALL_DEFINE1(sysinfo, struct sysinfo __user *, info)
2367 {
2368         struct sysinfo val;
2369 
2370         do_sysinfo(&val);
2371 
2372         if (copy_to_user(info, &val, sizeof(struct sysinfo)))
2373                 return -EFAULT;
2374 
2375         return 0;
2376 }
2377 
2378 #ifdef CONFIG_COMPAT
2379 struct compat_sysinfo {
2380         s32 uptime;
2381         u32 loads[3];
2382         u32 totalram;
2383         u32 freeram;
2384         u32 sharedram;
2385         u32 bufferram;
2386         u32 totalswap;
2387         u32 freeswap;
2388         u16 procs;
2389         u16 pad;
2390         u32 totalhigh;
2391         u32 freehigh;
2392         u32 mem_unit;
2393         char _f[20-2*sizeof(u32)-sizeof(int)];
2394 };
2395 
2396 COMPAT_SYSCALL_DEFINE1(sysinfo, struct compat_sysinfo __user *, info)
2397 {
2398         struct sysinfo s;
2399 
2400         do_sysinfo(&s);
2401 
2402         /* Check to see if any memory value is too large for 32-bit and scale
2403          *  down if needed
2404          */
2405         if (upper_32_bits(s.totalram) || upper_32_bits(s.totalswap)) {
2406                 int bitcount = 0;
2407 
2408                 while (s.mem_unit < PAGE_SIZE) {
2409                         s.mem_unit <<= 1;
2410                         bitcount++;
2411                 }
2412 
2413                 s.totalram >>= bitcount;
2414                 s.freeram >>= bitcount;
2415                 s.sharedram >>= bitcount;
2416                 s.bufferram >>= bitcount;
2417                 s.totalswap >>= bitcount;
2418                 s.freeswap >>= bitcount;
2419                 s.totalhigh >>= bitcount;
2420                 s.freehigh >>= bitcount;
2421         }
2422 
2423         if (!access_ok(VERIFY_WRITE, info, sizeof(struct compat_sysinfo)) ||
2424             __put_user(s.uptime, &info->uptime) ||
2425             __put_user(s.loads[0], &info->loads[0]) ||
2426             __put_user(s.loads[1], &info->loads[1]) ||
2427             __put_user(s.loads[2], &info->loads[2]) ||
2428             __put_user(s.totalram, &info->totalram) ||
2429             __put_user(s.freeram, &info->freeram) ||
2430             __put_user(s.sharedram, &info->sharedram) ||
2431             __put_user(s.bufferram, &info->bufferram) ||
2432             __put_user(s.totalswap, &info->totalswap) ||
2433             __put_user(s.freeswap, &info->freeswap) ||
2434             __put_user(s.procs, &info->procs) ||
2435             __put_user(s.totalhigh, &info->totalhigh) ||
2436             __put_user(s.freehigh, &info->freehigh) ||
2437             __put_user(s.mem_unit, &info->mem_unit))
2438                 return -EFAULT;
2439 
2440         return 0;
2441 }
2442 #endif /* CONFIG_COMPAT */
2443 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp