1e1f41b3fdf610c0e80d8330514975e722b22fcc
[linux-2.6.git] / kernel / sys.c
1 /*
2  *  linux/kernel/sys.c
3  *
4  *  Copyright (C) 1991, 1992  Linus Torvalds
5  */
6
7 #include <linux/config.h>
8 #include <linux/module.h>
9 #include <linux/mm.h>
10 #include <linux/utsname.h>
11 #include <linux/mman.h>
12 #include <linux/smp_lock.h>
13 #include <linux/notifier.h>
14 #include <linux/reboot.h>
15 #include <linux/prctl.h>
16 #include <linux/init.h>
17 #include <linux/highuid.h>
18 #include <linux/fs.h>
19 #include <linux/kernel.h>
20 #include <linux/kexec.h>
21 #include <linux/workqueue.h>
22 #include <linux/device.h>
23 #include <linux/key.h>
24 #include <linux/times.h>
25 #include <linux/posix-timers.h>
26 #include <linux/security.h>
27 #include <linux/dcookies.h>
28 #include <linux/suspend.h>
29 #include <linux/tty.h>
30 #include <linux/signal.h>
31 #include <linux/cn_proc.h>
32
33 #include <linux/compat.h>
34 #include <linux/syscalls.h>
35
36 #include <asm/uaccess.h>
37 #include <asm/io.h>
38 #include <asm/unistd.h>
39
40 #ifndef SET_UNALIGN_CTL
41 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
42 #endif
43 #ifndef GET_UNALIGN_CTL
44 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
45 #endif
46 #ifndef SET_FPEMU_CTL
47 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
48 #endif
49 #ifndef GET_FPEMU_CTL
50 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
51 #endif
52 #ifndef SET_FPEXC_CTL
53 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
54 #endif
55 #ifndef GET_FPEXC_CTL
56 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
57 #endif
58
59 /*
60  * this is where the system-wide overflow UID and GID are defined, for
61  * architectures that now have 32-bit UID/GID but didn't in the past
62  */
63
64 int overflowuid = DEFAULT_OVERFLOWUID;
65 int overflowgid = DEFAULT_OVERFLOWGID;
66
67 #ifdef CONFIG_UID16
68 EXPORT_SYMBOL(overflowuid);
69 EXPORT_SYMBOL(overflowgid);
70 #endif
71
72 /*
73  * the same as above, but for filesystems which can only store a 16-bit
74  * UID and GID. as such, this is needed on all architectures
75  */
76
77 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
78 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
79
80 EXPORT_SYMBOL(fs_overflowuid);
81 EXPORT_SYMBOL(fs_overflowgid);
82
83 /*
84  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
85  */
86
87 int C_A_D = 1;
88 int cad_pid = 1;
89
90 /*
91  *      Notifier list for kernel code which wants to be called
92  *      at shutdown. This is used to stop any idling DMA operations
93  *      and the like. 
94  */
95
96 static struct notifier_block *reboot_notifier_list;
97 static DEFINE_RWLOCK(notifier_lock);
98
99 /**
100  *      notifier_chain_register - Add notifier to a notifier chain
101  *      @list: Pointer to root list pointer
102  *      @n: New entry in notifier chain
103  *
104  *      Adds a notifier to a notifier chain.
105  *
106  *      Currently always returns zero.
107  */
108  
109 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
110 {
111         write_lock(&notifier_lock);
112         while(*list)
113         {
114                 if(n->priority > (*list)->priority)
115                         break;
116                 list= &((*list)->next);
117         }
118         n->next = *list;
119         *list=n;
120         write_unlock(&notifier_lock);
121         return 0;
122 }
123
124 EXPORT_SYMBOL(notifier_chain_register);
125
126 /**
127  *      notifier_chain_unregister - Remove notifier from a notifier chain
128  *      @nl: Pointer to root list pointer
129  *      @n: New entry in notifier chain
130  *
131  *      Removes a notifier from a notifier chain.
132  *
133  *      Returns zero on success, or %-ENOENT on failure.
134  */
135  
136 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
137 {
138         write_lock(&notifier_lock);
139         while((*nl)!=NULL)
140         {
141                 if((*nl)==n)
142                 {
143                         *nl=n->next;
144                         write_unlock(&notifier_lock);
145                         return 0;
146                 }
147                 nl=&((*nl)->next);
148         }
149         write_unlock(&notifier_lock);
150         return -ENOENT;
151 }
152
153 EXPORT_SYMBOL(notifier_chain_unregister);
154
155 /**
156  *      notifier_call_chain - Call functions in a notifier chain
157  *      @n: Pointer to root pointer of notifier chain
158  *      @val: Value passed unmodified to notifier function
159  *      @v: Pointer passed unmodified to notifier function
160  *
161  *      Calls each function in a notifier chain in turn.
162  *
163  *      If the return value of the notifier can be and'd
164  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
165  *      will return immediately, with the return value of
166  *      the notifier function which halted execution.
167  *      Otherwise, the return value is the return value
168  *      of the last notifier function called.
169  */
170  
171 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
172 {
173         int ret=NOTIFY_DONE;
174         struct notifier_block *nb = *n;
175
176         while(nb)
177         {
178                 ret=nb->notifier_call(nb,val,v);
179                 if(ret&NOTIFY_STOP_MASK)
180                 {
181                         return ret;
182                 }
183                 nb=nb->next;
184         }
185         return ret;
186 }
187
188 EXPORT_SYMBOL(notifier_call_chain);
189
190 /**
191  *      register_reboot_notifier - Register function to be called at reboot time
192  *      @nb: Info about notifier function to be called
193  *
194  *      Registers a function with the list of functions
195  *      to be called at reboot time.
196  *
197  *      Currently always returns zero, as notifier_chain_register
198  *      always returns zero.
199  */
200  
201 int register_reboot_notifier(struct notifier_block * nb)
202 {
203         return notifier_chain_register(&reboot_notifier_list, nb);
204 }
205
206 EXPORT_SYMBOL(register_reboot_notifier);
207
208 /**
209  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
210  *      @nb: Hook to be unregistered
211  *
212  *      Unregisters a previously registered reboot
213  *      notifier function.
214  *
215  *      Returns zero on success, or %-ENOENT on failure.
216  */
217  
218 int unregister_reboot_notifier(struct notifier_block * nb)
219 {
220         return notifier_chain_unregister(&reboot_notifier_list, nb);
221 }
222
223 EXPORT_SYMBOL(unregister_reboot_notifier);
224
225 static int set_one_prio(struct task_struct *p, int niceval, int error)
226 {
227         int no_nice;
228
229         if (p->uid != current->euid &&
230                 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
231                 error = -EPERM;
232                 goto out;
233         }
234         if (niceval < task_nice(p) && !can_nice(p, niceval)) {
235                 error = -EACCES;
236                 goto out;
237         }
238         no_nice = security_task_setnice(p, niceval);
239         if (no_nice) {
240                 error = no_nice;
241                 goto out;
242         }
243         if (error == -ESRCH)
244                 error = 0;
245         set_user_nice(p, niceval);
246 out:
247         return error;
248 }
249
250 asmlinkage long sys_setpriority(int which, int who, int niceval)
251 {
252         struct task_struct *g, *p;
253         struct user_struct *user;
254         int error = -EINVAL;
255
256         if (which > 2 || which < 0)
257                 goto out;
258
259         /* normalize: avoid signed division (rounding problems) */
260         error = -ESRCH;
261         if (niceval < -20)
262                 niceval = -20;
263         if (niceval > 19)
264                 niceval = 19;
265
266         read_lock(&tasklist_lock);
267         switch (which) {
268                 case PRIO_PROCESS:
269                         if (!who)
270                                 who = current->pid;
271                         p = find_task_by_pid(who);
272                         if (p)
273                                 error = set_one_prio(p, niceval, error);
274                         break;
275                 case PRIO_PGRP:
276                         if (!who)
277                                 who = process_group(current);
278                         do_each_task_pid(who, PIDTYPE_PGID, p) {
279                                 error = set_one_prio(p, niceval, error);
280                         } while_each_task_pid(who, PIDTYPE_PGID, p);
281                         break;
282                 case PRIO_USER:
283                         user = current->user;
284                         if (!who)
285                                 who = current->uid;
286                         else
287                                 if ((who != current->uid) && !(user = find_user(who)))
288                                         goto out_unlock;        /* No processes for this user */
289
290                         do_each_thread(g, p)
291                                 if (p->uid == who)
292                                         error = set_one_prio(p, niceval, error);
293                         while_each_thread(g, p);
294                         if (who != current->uid)
295                                 free_uid(user);         /* For find_user() */
296                         break;
297         }
298 out_unlock:
299         read_unlock(&tasklist_lock);
300 out:
301         return error;
302 }
303
304 /*
305  * Ugh. To avoid negative return values, "getpriority()" will
306  * not return the normal nice-value, but a negated value that
307  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
308  * to stay compatible.
309  */
310 asmlinkage long sys_getpriority(int which, int who)
311 {
312         struct task_struct *g, *p;
313         struct user_struct *user;
314         long niceval, retval = -ESRCH;
315
316         if (which > 2 || which < 0)
317                 return -EINVAL;
318
319         read_lock(&tasklist_lock);
320         switch (which) {
321                 case PRIO_PROCESS:
322                         if (!who)
323                                 who = current->pid;
324                         p = find_task_by_pid(who);
325                         if (p) {
326                                 niceval = 20 - task_nice(p);
327                                 if (niceval > retval)
328                                         retval = niceval;
329                         }
330                         break;
331                 case PRIO_PGRP:
332                         if (!who)
333                                 who = process_group(current);
334                         do_each_task_pid(who, PIDTYPE_PGID, p) {
335                                 niceval = 20 - task_nice(p);
336                                 if (niceval > retval)
337                                         retval = niceval;
338                         } while_each_task_pid(who, PIDTYPE_PGID, p);
339                         break;
340                 case PRIO_USER:
341                         user = current->user;
342                         if (!who)
343                                 who = current->uid;
344                         else
345                                 if ((who != current->uid) && !(user = find_user(who)))
346                                         goto out_unlock;        /* No processes for this user */
347
348                         do_each_thread(g, p)
349                                 if (p->uid == who) {
350                                         niceval = 20 - task_nice(p);
351                                         if (niceval > retval)
352                                                 retval = niceval;
353                                 }
354                         while_each_thread(g, p);
355                         if (who != current->uid)
356                                 free_uid(user);         /* for find_user() */
357                         break;
358         }
359 out_unlock:
360         read_unlock(&tasklist_lock);
361
362         return retval;
363 }
364
365 /**
366  *      emergency_restart - reboot the system
367  *
368  *      Without shutting down any hardware or taking any locks
369  *      reboot the system.  This is called when we know we are in
370  *      trouble so this is our best effort to reboot.  This is
371  *      safe to call in interrupt context.
372  */
373 void emergency_restart(void)
374 {
375         machine_emergency_restart();
376 }
377 EXPORT_SYMBOL_GPL(emergency_restart);
378
379 /**
380  *      kernel_restart - reboot the system
381  *
382  *      Shutdown everything and perform a clean reboot.
383  *      This is not safe to call in interrupt context.
384  */
385 void kernel_restart_prepare(char *cmd)
386 {
387         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
388         system_state = SYSTEM_RESTART;
389         device_shutdown();
390 }
391 void kernel_restart(char *cmd)
392 {
393         kernel_restart_prepare(cmd);
394         if (!cmd) {
395                 printk(KERN_EMERG "Restarting system.\n");
396         } else {
397                 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
398         }
399         printk(".\n");
400         machine_restart(cmd);
401 }
402 EXPORT_SYMBOL_GPL(kernel_restart);
403
404 /**
405  *      kernel_kexec - reboot the system
406  *
407  *      Move into place and start executing a preloaded standalone
408  *      executable.  If nothing was preloaded return an error.
409  */
410 void kernel_kexec(void)
411 {
412 #ifdef CONFIG_KEXEC
413         struct kimage *image;
414         image = xchg(&kexec_image, 0);
415         if (!image) {
416                 return;
417         }
418         kernel_restart_prepare(NULL);
419         printk(KERN_EMERG "Starting new kernel\n");
420         machine_shutdown();
421         machine_kexec(image);
422 #endif
423 }
424 EXPORT_SYMBOL_GPL(kernel_kexec);
425
426 /**
427  *      kernel_halt - halt the system
428  *
429  *      Shutdown everything and perform a clean system halt.
430  */
431 void kernel_halt_prepare(void)
432 {
433         notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
434         system_state = SYSTEM_HALT;
435         device_shutdown();
436 }
437 void kernel_halt(void)
438 {
439         kernel_halt_prepare();
440         printk(KERN_EMERG "System halted.\n");
441         machine_halt();
442 }
443 EXPORT_SYMBOL_GPL(kernel_halt);
444
445 /**
446  *      kernel_power_off - power_off the system
447  *
448  *      Shutdown everything and perform a clean system power_off.
449  */
450 void kernel_power_off_prepare(void)
451 {
452         notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
453         system_state = SYSTEM_POWER_OFF;
454         device_shutdown();
455 }
456 void kernel_power_off(void)
457 {
458         kernel_power_off_prepare();
459         printk(KERN_EMERG "Power down.\n");
460         machine_power_off();
461 }
462 EXPORT_SYMBOL_GPL(kernel_power_off);
463
464 /*
465  * Reboot system call: for obvious reasons only root may call it,
466  * and even root needs to set up some magic numbers in the registers
467  * so that some mistake won't make this reboot the whole machine.
468  * You can also set the meaning of the ctrl-alt-del-key here.
469  *
470  * reboot doesn't sync: do that yourself before calling this.
471  */
472 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
473 {
474         char buffer[256];
475
476         /* We only trust the superuser with rebooting the system. */
477         if (!capable(CAP_SYS_BOOT))
478                 return -EPERM;
479
480         /* For safety, we require "magic" arguments. */
481         if (magic1 != LINUX_REBOOT_MAGIC1 ||
482             (magic2 != LINUX_REBOOT_MAGIC2 &&
483                         magic2 != LINUX_REBOOT_MAGIC2A &&
484                         magic2 != LINUX_REBOOT_MAGIC2B &&
485                         magic2 != LINUX_REBOOT_MAGIC2C))
486                 return -EINVAL;
487
488         lock_kernel();
489         switch (cmd) {
490         case LINUX_REBOOT_CMD_RESTART:
491                 kernel_restart(NULL);
492                 break;
493
494         case LINUX_REBOOT_CMD_CAD_ON:
495                 C_A_D = 1;
496                 break;
497
498         case LINUX_REBOOT_CMD_CAD_OFF:
499                 C_A_D = 0;
500                 break;
501
502         case LINUX_REBOOT_CMD_HALT:
503                 kernel_halt();
504                 unlock_kernel();
505                 do_exit(0);
506                 break;
507
508         case LINUX_REBOOT_CMD_POWER_OFF:
509                 kernel_power_off();
510                 unlock_kernel();
511                 do_exit(0);
512                 break;
513
514         case LINUX_REBOOT_CMD_RESTART2:
515                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
516                         unlock_kernel();
517                         return -EFAULT;
518                 }
519                 buffer[sizeof(buffer) - 1] = '\0';
520
521                 kernel_restart(buffer);
522                 break;
523
524         case LINUX_REBOOT_CMD_KEXEC:
525                 kernel_kexec();
526                 unlock_kernel();
527                 return -EINVAL;
528
529 #ifdef CONFIG_SOFTWARE_SUSPEND
530         case LINUX_REBOOT_CMD_SW_SUSPEND:
531                 {
532                         int ret = software_suspend();
533                         unlock_kernel();
534                         return ret;
535                 }
536 #endif
537
538         default:
539                 unlock_kernel();
540                 return -EINVAL;
541         }
542         unlock_kernel();
543         return 0;
544 }
545
546 static void deferred_cad(void *dummy)
547 {
548         kernel_restart(NULL);
549 }
550
551 /*
552  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
553  * As it's called within an interrupt, it may NOT sync: the only choice
554  * is whether to reboot at once, or just ignore the ctrl-alt-del.
555  */
556 void ctrl_alt_del(void)
557 {
558         static DECLARE_WORK(cad_work, deferred_cad, NULL);
559
560         if (C_A_D)
561                 schedule_work(&cad_work);
562         else
563                 kill_proc(cad_pid, SIGINT, 1);
564 }
565         
566
567 /*
568  * Unprivileged users may change the real gid to the effective gid
569  * or vice versa.  (BSD-style)
570  *
571  * If you set the real gid at all, or set the effective gid to a value not
572  * equal to the real gid, then the saved gid is set to the new effective gid.
573  *
574  * This makes it possible for a setgid program to completely drop its
575  * privileges, which is often a useful assertion to make when you are doing
576  * a security audit over a program.
577  *
578  * The general idea is that a program which uses just setregid() will be
579  * 100% compatible with BSD.  A program which uses just setgid() will be
580  * 100% compatible with POSIX with saved IDs. 
581  *
582  * SMP: There are not races, the GIDs are checked only by filesystem
583  *      operations (as far as semantic preservation is concerned).
584  */
585 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
586 {
587         int old_rgid = current->gid;
588         int old_egid = current->egid;
589         int new_rgid = old_rgid;
590         int new_egid = old_egid;
591         int retval;
592
593         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
594         if (retval)
595                 return retval;
596
597         if (rgid != (gid_t) -1) {
598                 if ((old_rgid == rgid) ||
599                     (current->egid==rgid) ||
600                     capable(CAP_SETGID))
601                         new_rgid = rgid;
602                 else
603                         return -EPERM;
604         }
605         if (egid != (gid_t) -1) {
606                 if ((old_rgid == egid) ||
607                     (current->egid == egid) ||
608                     (current->sgid == egid) ||
609                     capable(CAP_SETGID))
610                         new_egid = egid;
611                 else {
612                         return -EPERM;
613                 }
614         }
615         if (new_egid != old_egid)
616         {
617                 current->mm->dumpable = suid_dumpable;
618                 smp_wmb();
619         }
620         if (rgid != (gid_t) -1 ||
621             (egid != (gid_t) -1 && egid != old_rgid))
622                 current->sgid = new_egid;
623         current->fsgid = new_egid;
624         current->egid = new_egid;
625         current->gid = new_rgid;
626         key_fsgid_changed(current);
627         proc_id_connector(current, PROC_EVENT_GID);
628         return 0;
629 }
630
631 /*
632  * setgid() is implemented like SysV w/ SAVED_IDS 
633  *
634  * SMP: Same implicit races as above.
635  */
636 asmlinkage long sys_setgid(gid_t gid)
637 {
638         int old_egid = current->egid;
639         int retval;
640
641         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
642         if (retval)
643                 return retval;
644
645         if (capable(CAP_SETGID))
646         {
647                 if(old_egid != gid)
648                 {
649                         current->mm->dumpable = suid_dumpable;
650                         smp_wmb();
651                 }
652                 current->gid = current->egid = current->sgid = current->fsgid = gid;
653         }
654         else if ((gid == current->gid) || (gid == current->sgid))
655         {
656                 if(old_egid != gid)
657                 {
658                         current->mm->dumpable = suid_dumpable;
659                         smp_wmb();
660                 }
661                 current->egid = current->fsgid = gid;
662         }
663         else
664                 return -EPERM;
665
666         key_fsgid_changed(current);
667         proc_id_connector(current, PROC_EVENT_GID);
668         return 0;
669 }
670   
671 static int set_user(uid_t new_ruid, int dumpclear)
672 {
673         struct user_struct *new_user;
674
675         new_user = alloc_uid(new_ruid);
676         if (!new_user)
677                 return -EAGAIN;
678
679         if (atomic_read(&new_user->processes) >=
680                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
681                         new_user != &root_user) {
682                 free_uid(new_user);
683                 return -EAGAIN;
684         }
685
686         switch_uid(new_user);
687
688         if(dumpclear)
689         {
690                 current->mm->dumpable = suid_dumpable;
691                 smp_wmb();
692         }
693         current->uid = new_ruid;
694         return 0;
695 }
696
697 /*
698  * Unprivileged users may change the real uid to the effective uid
699  * or vice versa.  (BSD-style)
700  *
701  * If you set the real uid at all, or set the effective uid to a value not
702  * equal to the real uid, then the saved uid is set to the new effective uid.
703  *
704  * This makes it possible for a setuid program to completely drop its
705  * privileges, which is often a useful assertion to make when you are doing
706  * a security audit over a program.
707  *
708  * The general idea is that a program which uses just setreuid() will be
709  * 100% compatible with BSD.  A program which uses just setuid() will be
710  * 100% compatible with POSIX with saved IDs. 
711  */
712 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
713 {
714         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
715         int retval;
716
717         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
718         if (retval)
719                 return retval;
720
721         new_ruid = old_ruid = current->uid;
722         new_euid = old_euid = current->euid;
723         old_suid = current->suid;
724
725         if (ruid != (uid_t) -1) {
726                 new_ruid = ruid;
727                 if ((old_ruid != ruid) &&
728                     (current->euid != ruid) &&
729                     !capable(CAP_SETUID))
730                         return -EPERM;
731         }
732
733         if (euid != (uid_t) -1) {
734                 new_euid = euid;
735                 if ((old_ruid != euid) &&
736                     (current->euid != euid) &&
737                     (current->suid != euid) &&
738                     !capable(CAP_SETUID))
739                         return -EPERM;
740         }
741
742         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
743                 return -EAGAIN;
744
745         if (new_euid != old_euid)
746         {
747                 current->mm->dumpable = suid_dumpable;
748                 smp_wmb();
749         }
750         current->fsuid = current->euid = new_euid;
751         if (ruid != (uid_t) -1 ||
752             (euid != (uid_t) -1 && euid != old_ruid))
753                 current->suid = current->euid;
754         current->fsuid = current->euid;
755
756         key_fsuid_changed(current);
757         proc_id_connector(current, PROC_EVENT_UID);
758
759         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
760 }
761
762
763                 
764 /*
765  * setuid() is implemented like SysV with SAVED_IDS 
766  * 
767  * Note that SAVED_ID's is deficient in that a setuid root program
768  * like sendmail, for example, cannot set its uid to be a normal 
769  * user and then switch back, because if you're root, setuid() sets
770  * the saved uid too.  If you don't like this, blame the bright people
771  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
772  * will allow a root program to temporarily drop privileges and be able to
773  * regain them by swapping the real and effective uid.  
774  */
775 asmlinkage long sys_setuid(uid_t uid)
776 {
777         int old_euid = current->euid;
778         int old_ruid, old_suid, new_ruid, new_suid;
779         int retval;
780
781         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
782         if (retval)
783                 return retval;
784
785         old_ruid = new_ruid = current->uid;
786         old_suid = current->suid;
787         new_suid = old_suid;
788         
789         if (capable(CAP_SETUID)) {
790                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
791                         return -EAGAIN;
792                 new_suid = uid;
793         } else if ((uid != current->uid) && (uid != new_suid))
794                 return -EPERM;
795
796         if (old_euid != uid)
797         {
798                 current->mm->dumpable = suid_dumpable;
799                 smp_wmb();
800         }
801         current->fsuid = current->euid = uid;
802         current->suid = new_suid;
803
804         key_fsuid_changed(current);
805         proc_id_connector(current, PROC_EVENT_UID);
806
807         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
808 }
809
810
811 /*
812  * This function implements a generic ability to update ruid, euid,
813  * and suid.  This allows you to implement the 4.4 compatible seteuid().
814  */
815 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
816 {
817         int old_ruid = current->uid;
818         int old_euid = current->euid;
819         int old_suid = current->suid;
820         int retval;
821
822         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
823         if (retval)
824                 return retval;
825
826         if (!capable(CAP_SETUID)) {
827                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
828                     (ruid != current->euid) && (ruid != current->suid))
829                         return -EPERM;
830                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
831                     (euid != current->euid) && (euid != current->suid))
832                         return -EPERM;
833                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
834                     (suid != current->euid) && (suid != current->suid))
835                         return -EPERM;
836         }
837         if (ruid != (uid_t) -1) {
838                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
839                         return -EAGAIN;
840         }
841         if (euid != (uid_t) -1) {
842                 if (euid != current->euid)
843                 {
844                         current->mm->dumpable = suid_dumpable;
845                         smp_wmb();
846                 }
847                 current->euid = euid;
848         }
849         current->fsuid = current->euid;
850         if (suid != (uid_t) -1)
851                 current->suid = suid;
852
853         key_fsuid_changed(current);
854         proc_id_connector(current, PROC_EVENT_UID);
855
856         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
857 }
858
859 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
860 {
861         int retval;
862
863         if (!(retval = put_user(current->uid, ruid)) &&
864             !(retval = put_user(current->euid, euid)))
865                 retval = put_user(current->suid, suid);
866
867         return retval;
868 }
869
870 /*
871  * Same as above, but for rgid, egid, sgid.
872  */
873 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
874 {
875         int retval;
876
877         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
878         if (retval)
879                 return retval;
880
881         if (!capable(CAP_SETGID)) {
882                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
883                     (rgid != current->egid) && (rgid != current->sgid))
884                         return -EPERM;
885                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
886                     (egid != current->egid) && (egid != current->sgid))
887                         return -EPERM;
888                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
889                     (sgid != current->egid) && (sgid != current->sgid))
890                         return -EPERM;
891         }
892         if (egid != (gid_t) -1) {
893                 if (egid != current->egid)
894                 {
895                         current->mm->dumpable = suid_dumpable;
896                         smp_wmb();
897                 }
898                 current->egid = egid;
899         }
900         current->fsgid = current->egid;
901         if (rgid != (gid_t) -1)
902                 current->gid = rgid;
903         if (sgid != (gid_t) -1)
904                 current->sgid = sgid;
905
906         key_fsgid_changed(current);
907         proc_id_connector(current, PROC_EVENT_GID);
908         return 0;
909 }
910
911 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
912 {
913         int retval;
914
915         if (!(retval = put_user(current->gid, rgid)) &&
916             !(retval = put_user(current->egid, egid)))
917                 retval = put_user(current->sgid, sgid);
918
919         return retval;
920 }
921
922
923 /*
924  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
925  * is used for "access()" and for the NFS daemon (letting nfsd stay at
926  * whatever uid it wants to). It normally shadows "euid", except when
927  * explicitly set by setfsuid() or for access..
928  */
929 asmlinkage long sys_setfsuid(uid_t uid)
930 {
931         int old_fsuid;
932
933         old_fsuid = current->fsuid;
934         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
935                 return old_fsuid;
936
937         if (uid == current->uid || uid == current->euid ||
938             uid == current->suid || uid == current->fsuid || 
939             capable(CAP_SETUID))
940         {
941                 if (uid != old_fsuid)
942                 {
943                         current->mm->dumpable = suid_dumpable;
944                         smp_wmb();
945                 }
946                 current->fsuid = uid;
947         }
948
949         key_fsuid_changed(current);
950         proc_id_connector(current, PROC_EVENT_UID);
951
952         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
953
954         return old_fsuid;
955 }
956
957 /*
958  * Samma på svenska..
959  */
960 asmlinkage long sys_setfsgid(gid_t gid)
961 {
962         int old_fsgid;
963
964         old_fsgid = current->fsgid;
965         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
966                 return old_fsgid;
967
968         if (gid == current->gid || gid == current->egid ||
969             gid == current->sgid || gid == current->fsgid || 
970             capable(CAP_SETGID))
971         {
972                 if (gid != old_fsgid)
973                 {
974                         current->mm->dumpable = suid_dumpable;
975                         smp_wmb();
976                 }
977                 current->fsgid = gid;
978                 key_fsgid_changed(current);
979                 proc_id_connector(current, PROC_EVENT_GID);
980         }
981         return old_fsgid;
982 }
983
984 asmlinkage long sys_times(struct tms __user * tbuf)
985 {
986         /*
987          *      In the SMP world we might just be unlucky and have one of
988          *      the times increment as we use it. Since the value is an
989          *      atomically safe type this is just fine. Conceptually its
990          *      as if the syscall took an instant longer to occur.
991          */
992         if (tbuf) {
993                 struct tms tmp;
994                 cputime_t utime, stime, cutime, cstime;
995
996 #ifdef CONFIG_SMP
997                 if (thread_group_empty(current)) {
998                         /*
999                          * Single thread case without the use of any locks.
1000                          *
1001                          * We may race with release_task if two threads are
1002                          * executing. However, release task first adds up the
1003                          * counters (__exit_signal) before  removing the task
1004                          * from the process tasklist (__unhash_process).
1005                          * __exit_signal also acquires and releases the
1006                          * siglock which results in the proper memory ordering
1007                          * so that the list modifications are always visible
1008                          * after the counters have been updated.
1009                          *
1010                          * If the counters have been updated by the second thread
1011                          * but the thread has not yet been removed from the list
1012                          * then the other branch will be executing which will
1013                          * block on tasklist_lock until the exit handling of the
1014                          * other task is finished.
1015                          *
1016                          * This also implies that the sighand->siglock cannot
1017                          * be held by another processor. So we can also
1018                          * skip acquiring that lock.
1019                          */
1020                         utime = cputime_add(current->signal->utime, current->utime);
1021                         stime = cputime_add(current->signal->utime, current->stime);
1022                         cutime = current->signal->cutime;
1023                         cstime = current->signal->cstime;
1024                 } else
1025 #endif
1026                 {
1027
1028                         /* Process with multiple threads */
1029                         struct task_struct *tsk = current;
1030                         struct task_struct *t;
1031
1032                         read_lock(&tasklist_lock);
1033                         utime = tsk->signal->utime;
1034                         stime = tsk->signal->stime;
1035                         t = tsk;
1036                         do {
1037                                 utime = cputime_add(utime, t->utime);
1038                                 stime = cputime_add(stime, t->stime);
1039                                 t = next_thread(t);
1040                         } while (t != tsk);
1041
1042                         /*
1043                          * While we have tasklist_lock read-locked, no dying thread
1044                          * can be updating current->signal->[us]time.  Instead,
1045                          * we got their counts included in the live thread loop.
1046                          * However, another thread can come in right now and
1047                          * do a wait call that updates current->signal->c[us]time.
1048                          * To make sure we always see that pair updated atomically,
1049                          * we take the siglock around fetching them.
1050                          */
1051                         spin_lock_irq(&tsk->sighand->siglock);
1052                         cutime = tsk->signal->cutime;
1053                         cstime = tsk->signal->cstime;
1054                         spin_unlock_irq(&tsk->sighand->siglock);
1055                         read_unlock(&tasklist_lock);
1056                 }
1057                 tmp.tms_utime = cputime_to_clock_t(utime);
1058                 tmp.tms_stime = cputime_to_clock_t(stime);
1059                 tmp.tms_cutime = cputime_to_clock_t(cutime);
1060                 tmp.tms_cstime = cputime_to_clock_t(cstime);
1061                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1062                         return -EFAULT;
1063         }
1064         return (long) jiffies_64_to_clock_t(get_jiffies_64());
1065 }
1066
1067 /*
1068  * This needs some heavy checking ...
1069  * I just haven't the stomach for it. I also don't fully
1070  * understand sessions/pgrp etc. Let somebody who does explain it.
1071  *
1072  * OK, I think I have the protection semantics right.... this is really
1073  * only important on a multi-user system anyway, to make sure one user
1074  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1075  *
1076  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1077  * LBT 04.03.94
1078  */
1079
1080 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1081 {
1082         struct task_struct *p;
1083         int err = -EINVAL;
1084
1085         if (!pid)
1086                 pid = current->pid;
1087         if (!pgid)
1088                 pgid = pid;
1089         if (pgid < 0)
1090                 return -EINVAL;
1091
1092         /* From this point forward we keep holding onto the tasklist lock
1093          * so that our parent does not change from under us. -DaveM
1094          */
1095         write_lock_irq(&tasklist_lock);
1096
1097         err = -ESRCH;
1098         p = find_task_by_pid(pid);
1099         if (!p)
1100                 goto out;
1101
1102         err = -EINVAL;
1103         if (!thread_group_leader(p))
1104                 goto out;
1105
1106         if (p->parent == current || p->real_parent == current) {
1107                 err = -EPERM;
1108                 if (p->signal->session != current->signal->session)
1109                         goto out;
1110                 err = -EACCES;
1111                 if (p->did_exec)
1112                         goto out;
1113         } else {
1114                 err = -ESRCH;
1115                 if (p != current)
1116                         goto out;
1117         }
1118
1119         err = -EPERM;
1120         if (p->signal->leader)
1121                 goto out;
1122
1123         if (pgid != pid) {
1124                 struct task_struct *p;
1125
1126                 do_each_task_pid(pgid, PIDTYPE_PGID, p) {
1127                         if (p->signal->session == current->signal->session)
1128                                 goto ok_pgid;
1129                 } while_each_task_pid(pgid, PIDTYPE_PGID, p);
1130                 goto out;
1131         }
1132
1133 ok_pgid:
1134         err = security_task_setpgid(p, pgid);
1135         if (err)
1136                 goto out;
1137
1138         if (process_group(p) != pgid) {
1139                 detach_pid(p, PIDTYPE_PGID);
1140                 p->signal->pgrp = pgid;
1141                 attach_pid(p, PIDTYPE_PGID, pgid);
1142         }
1143
1144         err = 0;
1145 out:
1146         /* All paths lead to here, thus we are safe. -DaveM */
1147         write_unlock_irq(&tasklist_lock);
1148         return err;
1149 }
1150
1151 asmlinkage long sys_getpgid(pid_t pid)
1152 {
1153         if (!pid) {
1154                 return process_group(current);
1155         } else {
1156                 int retval;
1157                 struct task_struct *p;
1158
1159                 read_lock(&tasklist_lock);
1160                 p = find_task_by_pid(pid);
1161
1162                 retval = -ESRCH;
1163                 if (p) {
1164                         retval = security_task_getpgid(p);
1165                         if (!retval)
1166                                 retval = process_group(p);
1167                 }
1168                 read_unlock(&tasklist_lock);
1169                 return retval;
1170         }
1171 }
1172
1173 #ifdef __ARCH_WANT_SYS_GETPGRP
1174
1175 asmlinkage long sys_getpgrp(void)
1176 {
1177         /* SMP - assuming writes are word atomic this is fine */
1178         return process_group(current);
1179 }
1180
1181 #endif
1182
1183 asmlinkage long sys_getsid(pid_t pid)
1184 {
1185         if (!pid) {
1186                 return current->signal->session;
1187         } else {
1188                 int retval;
1189                 struct task_struct *p;
1190
1191                 read_lock(&tasklist_lock);
1192                 p = find_task_by_pid(pid);
1193
1194                 retval = -ESRCH;
1195                 if(p) {
1196                         retval = security_task_getsid(p);
1197                         if (!retval)
1198                                 retval = p->signal->session;
1199                 }
1200                 read_unlock(&tasklist_lock);
1201                 return retval;
1202         }
1203 }
1204
1205 asmlinkage long sys_setsid(void)
1206 {
1207         struct pid *pid;
1208         int err = -EPERM;
1209
1210         if (!thread_group_leader(current))
1211                 return -EINVAL;
1212
1213         down(&tty_sem);
1214         write_lock_irq(&tasklist_lock);
1215
1216         pid = find_pid(PIDTYPE_PGID, current->pid);
1217         if (pid)
1218                 goto out;
1219
1220         current->signal->leader = 1;
1221         __set_special_pids(current->pid, current->pid);
1222         current->signal->tty = NULL;
1223         current->signal->tty_old_pgrp = 0;
1224         err = process_group(current);
1225 out:
1226         write_unlock_irq(&tasklist_lock);
1227         up(&tty_sem);
1228         return err;
1229 }
1230
1231 /*
1232  * Supplementary group IDs
1233  */
1234
1235 /* init to 2 - one for init_task, one to ensure it is never freed */
1236 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1237
1238 struct group_info *groups_alloc(int gidsetsize)
1239 {
1240         struct group_info *group_info;
1241         int nblocks;
1242         int i;
1243
1244         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1245         /* Make sure we always allocate at least one indirect block pointer */
1246         nblocks = nblocks ? : 1;
1247         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1248         if (!group_info)
1249                 return NULL;
1250         group_info->ngroups = gidsetsize;
1251         group_info->nblocks = nblocks;
1252         atomic_set(&group_info->usage, 1);
1253
1254         if (gidsetsize <= NGROUPS_SMALL) {
1255                 group_info->blocks[0] = group_info->small_block;
1256         } else {
1257                 for (i = 0; i < nblocks; i++) {
1258                         gid_t *b;
1259                         b = (void *)__get_free_page(GFP_USER);
1260                         if (!b)
1261                                 goto out_undo_partial_alloc;
1262                         group_info->blocks[i] = b;
1263                 }
1264         }
1265         return group_info;
1266
1267 out_undo_partial_alloc:
1268         while (--i >= 0) {
1269                 free_page((unsigned long)group_info->blocks[i]);
1270         }
1271         kfree(group_info);
1272         return NULL;
1273 }
1274
1275 EXPORT_SYMBOL(groups_alloc);
1276
1277 void groups_free(struct group_info *group_info)
1278 {
1279         if (group_info->blocks[0] != group_info->small_block) {
1280                 int i;
1281                 for (i = 0; i < group_info->nblocks; i++)
1282                         free_page((unsigned long)group_info->blocks[i]);
1283         }
1284         kfree(group_info);
1285 }
1286
1287 EXPORT_SYMBOL(groups_free);
1288
1289 /* export the group_info to a user-space array */
1290 static int groups_to_user(gid_t __user *grouplist,
1291     struct group_info *group_info)
1292 {
1293         int i;
1294         int count = group_info->ngroups;
1295
1296         for (i = 0; i < group_info->nblocks; i++) {
1297                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1298                 int off = i * NGROUPS_PER_BLOCK;
1299                 int len = cp_count * sizeof(*grouplist);
1300
1301                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1302                         return -EFAULT;
1303
1304                 count -= cp_count;
1305         }
1306         return 0;
1307 }
1308
1309 /* fill a group_info from a user-space array - it must be allocated already */
1310 static int groups_from_user(struct group_info *group_info,
1311     gid_t __user *grouplist)
1312  {
1313         int i;
1314         int count = group_info->ngroups;
1315
1316         for (i = 0; i < group_info->nblocks; i++) {
1317                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1318                 int off = i * NGROUPS_PER_BLOCK;
1319                 int len = cp_count * sizeof(*grouplist);
1320
1321                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1322                         return -EFAULT;
1323
1324                 count -= cp_count;
1325         }
1326         return 0;
1327 }
1328
1329 /* a simple Shell sort */
1330 static void groups_sort(struct group_info *group_info)
1331 {
1332         int base, max, stride;
1333         int gidsetsize = group_info->ngroups;
1334
1335         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1336                 ; /* nothing */
1337         stride /= 3;
1338
1339         while (stride) {
1340                 max = gidsetsize - stride;
1341                 for (base = 0; base < max; base++) {
1342                         int left = base;
1343                         int right = left + stride;
1344                         gid_t tmp = GROUP_AT(group_info, right);
1345
1346                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1347                                 GROUP_AT(group_info, right) =
1348                                     GROUP_AT(group_info, left);
1349                                 right = left;
1350                                 left -= stride;
1351                         }
1352                         GROUP_AT(group_info, right) = tmp;
1353                 }
1354                 stride /= 3;
1355         }
1356 }
1357
1358 /* a simple bsearch */
1359 int groups_search(struct group_info *group_info, gid_t grp)
1360 {
1361         int left, right;
1362
1363         if (!group_info)
1364                 return 0;
1365
1366         left = 0;
1367         right = group_info->ngroups;
1368         while (left < right) {
1369                 int mid = (left+right)/2;
1370                 int cmp = grp - GROUP_AT(group_info, mid);
1371                 if (cmp > 0)
1372                         left = mid + 1;
1373                 else if (cmp < 0)
1374                         right = mid;
1375                 else
1376                         return 1;
1377         }
1378         return 0;
1379 }
1380
1381 /* validate and set current->group_info */
1382 int set_current_groups(struct group_info *group_info)
1383 {
1384         int retval;
1385         struct group_info *old_info;
1386
1387         retval = security_task_setgroups(group_info);
1388         if (retval)
1389                 return retval;
1390
1391         groups_sort(group_info);
1392         get_group_info(group_info);
1393
1394         task_lock(current);
1395         old_info = current->group_info;
1396         current->group_info = group_info;
1397         task_unlock(current);
1398
1399         put_group_info(old_info);
1400
1401         return 0;
1402 }
1403
1404 EXPORT_SYMBOL(set_current_groups);
1405
1406 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1407 {
1408         int i = 0;
1409
1410         /*
1411          *      SMP: Nobody else can change our grouplist. Thus we are
1412          *      safe.
1413          */
1414
1415         if (gidsetsize < 0)
1416                 return -EINVAL;
1417
1418         /* no need to grab task_lock here; it cannot change */
1419         get_group_info(current->group_info);
1420         i = current->group_info->ngroups;
1421         if (gidsetsize) {
1422                 if (i > gidsetsize) {
1423                         i = -EINVAL;
1424                         goto out;
1425                 }
1426                 if (groups_to_user(grouplist, current->group_info)) {
1427                         i = -EFAULT;
1428                         goto out;
1429                 }
1430         }
1431 out:
1432         put_group_info(current->group_info);
1433         return i;
1434 }
1435
1436 /*
1437  *      SMP: Our groups are copy-on-write. We can set them safely
1438  *      without another task interfering.
1439  */
1440  
1441 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1442 {
1443         struct group_info *group_info;
1444         int retval;
1445
1446         if (!capable(CAP_SETGID))
1447                 return -EPERM;
1448         if ((unsigned)gidsetsize > NGROUPS_MAX)
1449                 return -EINVAL;
1450
1451         group_info = groups_alloc(gidsetsize);
1452         if (!group_info)
1453                 return -ENOMEM;
1454         retval = groups_from_user(group_info, grouplist);
1455         if (retval) {
1456                 put_group_info(group_info);
1457                 return retval;
1458         }
1459
1460         retval = set_current_groups(group_info);
1461         put_group_info(group_info);
1462
1463         return retval;
1464 }
1465
1466 /*
1467  * Check whether we're fsgid/egid or in the supplemental group..
1468  */
1469 int in_group_p(gid_t grp)
1470 {
1471         int retval = 1;
1472         if (grp != current->fsgid) {
1473                 get_group_info(current->group_info);
1474                 retval = groups_search(current->group_info, grp);
1475                 put_group_info(current->group_info);
1476         }
1477         return retval;
1478 }
1479
1480 EXPORT_SYMBOL(in_group_p);
1481
1482 int in_egroup_p(gid_t grp)
1483 {
1484         int retval = 1;
1485         if (grp != current->egid) {
1486                 get_group_info(current->group_info);
1487                 retval = groups_search(current->group_info, grp);
1488                 put_group_info(current->group_info);
1489         }
1490         return retval;
1491 }
1492
1493 EXPORT_SYMBOL(in_egroup_p);
1494
1495 DECLARE_RWSEM(uts_sem);
1496
1497 EXPORT_SYMBOL(uts_sem);
1498
1499 asmlinkage long sys_newuname(struct new_utsname __user * name)
1500 {
1501         int errno = 0;
1502
1503         down_read(&uts_sem);
1504         if (copy_to_user(name,&system_utsname,sizeof *name))
1505                 errno = -EFAULT;
1506         up_read(&uts_sem);
1507         return errno;
1508 }
1509
1510 asmlinkage long sys_sethostname(char __user *name, int len)
1511 {
1512         int errno;
1513         char tmp[__NEW_UTS_LEN];
1514
1515         if (!capable(CAP_SYS_ADMIN))
1516                 return -EPERM;
1517         if (len < 0 || len > __NEW_UTS_LEN)
1518                 return -EINVAL;
1519         down_write(&uts_sem);
1520         errno = -EFAULT;
1521         if (!copy_from_user(tmp, name, len)) {
1522                 memcpy(system_utsname.nodename, tmp, len);
1523                 system_utsname.nodename[len] = 0;
1524                 errno = 0;
1525         }
1526         up_write(&uts_sem);
1527         return errno;
1528 }
1529
1530 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1531
1532 asmlinkage long sys_gethostname(char __user *name, int len)
1533 {
1534         int i, errno;
1535
1536         if (len < 0)
1537                 return -EINVAL;
1538         down_read(&uts_sem);
1539         i = 1 + strlen(system_utsname.nodename);
1540         if (i > len)
1541                 i = len;
1542         errno = 0;
1543         if (copy_to_user(name, system_utsname.nodename, i))
1544                 errno = -EFAULT;
1545         up_read(&uts_sem);
1546         return errno;
1547 }
1548
1549 #endif
1550
1551 /*
1552  * Only setdomainname; getdomainname can be implemented by calling
1553  * uname()
1554  */
1555 asmlinkage long sys_setdomainname(char __user *name, int len)
1556 {
1557         int errno;
1558         char tmp[__NEW_UTS_LEN];
1559
1560         if (!capable(CAP_SYS_ADMIN))
1561                 return -EPERM;
1562         if (len < 0 || len > __NEW_UTS_LEN)
1563                 return -EINVAL;
1564
1565         down_write(&uts_sem);
1566         errno = -EFAULT;
1567         if (!copy_from_user(tmp, name, len)) {
1568                 memcpy(system_utsname.domainname, tmp, len);
1569                 system_utsname.domainname[len] = 0;
1570                 errno = 0;
1571         }
1572         up_write(&uts_sem);
1573         return errno;
1574 }
1575
1576 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1577 {
1578         if (resource >= RLIM_NLIMITS)
1579                 return -EINVAL;
1580         else {
1581                 struct rlimit value;
1582                 task_lock(current->group_leader);
1583                 value = current->signal->rlim[resource];
1584                 task_unlock(current->group_leader);
1585                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1586         }
1587 }
1588
1589 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1590
1591 /*
1592  *      Back compatibility for getrlimit. Needed for some apps.
1593  */
1594  
1595 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1596 {
1597         struct rlimit x;
1598         if (resource >= RLIM_NLIMITS)
1599                 return -EINVAL;
1600
1601         task_lock(current->group_leader);
1602         x = current->signal->rlim[resource];
1603         task_unlock(current->group_leader);
1604         if(x.rlim_cur > 0x7FFFFFFF)
1605                 x.rlim_cur = 0x7FFFFFFF;
1606         if(x.rlim_max > 0x7FFFFFFF)
1607                 x.rlim_max = 0x7FFFFFFF;
1608         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1609 }
1610
1611 #endif
1612
1613 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1614 {
1615         struct rlimit new_rlim, *old_rlim;
1616         int retval;
1617
1618         if (resource >= RLIM_NLIMITS)
1619                 return -EINVAL;
1620         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1621                 return -EFAULT;
1622        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1623                return -EINVAL;
1624         old_rlim = current->signal->rlim + resource;
1625         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1626             !capable(CAP_SYS_RESOURCE))
1627                 return -EPERM;
1628         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1629                         return -EPERM;
1630
1631         retval = security_task_setrlimit(resource, &new_rlim);
1632         if (retval)
1633                 return retval;
1634
1635         task_lock(current->group_leader);
1636         *old_rlim = new_rlim;
1637         task_unlock(current->group_leader);
1638
1639         if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1640             (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1641              new_rlim.rlim_cur <= cputime_to_secs(
1642                      current->signal->it_prof_expires))) {
1643                 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1644                 read_lock(&tasklist_lock);
1645                 spin_lock_irq(&current->sighand->siglock);
1646                 set_process_cpu_timer(current, CPUCLOCK_PROF,
1647                                       &cputime, NULL);
1648                 spin_unlock_irq(&current->sighand->siglock);
1649                 read_unlock(&tasklist_lock);
1650         }
1651
1652         return 0;
1653 }
1654
1655 /*
1656  * It would make sense to put struct rusage in the task_struct,
1657  * except that would make the task_struct be *really big*.  After
1658  * task_struct gets moved into malloc'ed memory, it would
1659  * make sense to do this.  It will make moving the rest of the information
1660  * a lot simpler!  (Which we're not doing right now because we're not
1661  * measuring them yet).
1662  *
1663  * This expects to be called with tasklist_lock read-locked or better,
1664  * and the siglock not locked.  It may momentarily take the siglock.
1665  *
1666  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1667  * races with threads incrementing their own counters.  But since word
1668  * reads are atomic, we either get new values or old values and we don't
1669  * care which for the sums.  We always take the siglock to protect reading
1670  * the c* fields from p->signal from races with exit.c updating those
1671  * fields when reaping, so a sample either gets all the additions of a
1672  * given child after it's reaped, or none so this sample is before reaping.
1673  */
1674
1675 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1676 {
1677         struct task_struct *t;
1678         unsigned long flags;
1679         cputime_t utime, stime;
1680
1681         memset((char *) r, 0, sizeof *r);
1682
1683         if (unlikely(!p->signal))
1684                 return;
1685
1686         switch (who) {
1687                 case RUSAGE_CHILDREN:
1688                         spin_lock_irqsave(&p->sighand->siglock, flags);
1689                         utime = p->signal->cutime;
1690                         stime = p->signal->cstime;
1691                         r->ru_nvcsw = p->signal->cnvcsw;
1692                         r->ru_nivcsw = p->signal->cnivcsw;
1693                         r->ru_minflt = p->signal->cmin_flt;
1694                         r->ru_majflt = p->signal->cmaj_flt;
1695                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1696                         cputime_to_timeval(utime, &r->ru_utime);
1697                         cputime_to_timeval(stime, &r->ru_stime);
1698                         break;
1699                 case RUSAGE_SELF:
1700                         spin_lock_irqsave(&p->sighand->siglock, flags);
1701                         utime = stime = cputime_zero;
1702                         goto sum_group;
1703                 case RUSAGE_BOTH:
1704                         spin_lock_irqsave(&p->sighand->siglock, flags);
1705                         utime = p->signal->cutime;
1706                         stime = p->signal->cstime;
1707                         r->ru_nvcsw = p->signal->cnvcsw;
1708                         r->ru_nivcsw = p->signal->cnivcsw;
1709                         r->ru_minflt = p->signal->cmin_flt;
1710                         r->ru_majflt = p->signal->cmaj_flt;
1711                 sum_group:
1712                         utime = cputime_add(utime, p->signal->utime);
1713                         stime = cputime_add(stime, p->signal->stime);
1714                         r->ru_nvcsw += p->signal->nvcsw;
1715                         r->ru_nivcsw += p->signal->nivcsw;
1716                         r->ru_minflt += p->signal->min_flt;
1717                         r->ru_majflt += p->signal->maj_flt;
1718                         t = p;
1719                         do {
1720                                 utime = cputime_add(utime, t->utime);
1721                                 stime = cputime_add(stime, t->stime);
1722                                 r->ru_nvcsw += t->nvcsw;
1723                                 r->ru_nivcsw += t->nivcsw;
1724                                 r->ru_minflt += t->min_flt;
1725                                 r->ru_majflt += t->maj_flt;
1726                                 t = next_thread(t);
1727                         } while (t != p);
1728                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1729                         cputime_to_timeval(utime, &r->ru_utime);
1730                         cputime_to_timeval(stime, &r->ru_stime);
1731                         break;
1732                 default:
1733                         BUG();
1734         }
1735 }
1736
1737 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1738 {
1739         struct rusage r;
1740         read_lock(&tasklist_lock);
1741         k_getrusage(p, who, &r);
1742         read_unlock(&tasklist_lock);
1743         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1744 }
1745
1746 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1747 {
1748         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1749                 return -EINVAL;
1750         return getrusage(current, who, ru);
1751 }
1752
1753 asmlinkage long sys_umask(int mask)
1754 {
1755         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1756         return mask;
1757 }
1758     
1759 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1760                           unsigned long arg4, unsigned long arg5)
1761 {
1762         long error;
1763
1764         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1765         if (error)
1766                 return error;
1767
1768         switch (option) {
1769                 case PR_SET_PDEATHSIG:
1770                         if (!valid_signal(arg2)) {
1771                                 error = -EINVAL;
1772                                 break;
1773                         }
1774                         current->pdeath_signal = arg2;
1775                         break;
1776                 case PR_GET_PDEATHSIG:
1777                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1778                         break;
1779                 case PR_GET_DUMPABLE:
1780                         error = current->mm->dumpable;
1781                         break;
1782                 case PR_SET_DUMPABLE:
1783                         if (arg2 < 0 || arg2 > 2) {
1784                                 error = -EINVAL;
1785                                 break;
1786                         }
1787                         current->mm->dumpable = arg2;
1788                         break;
1789
1790                 case PR_SET_UNALIGN:
1791                         error = SET_UNALIGN_CTL(current, arg2);
1792                         break;
1793                 case PR_GET_UNALIGN:
1794                         error = GET_UNALIGN_CTL(current, arg2);
1795                         break;
1796                 case PR_SET_FPEMU:
1797                         error = SET_FPEMU_CTL(current, arg2);
1798                         break;
1799                 case PR_GET_FPEMU:
1800                         error = GET_FPEMU_CTL(current, arg2);
1801                         break;
1802                 case PR_SET_FPEXC:
1803                         error = SET_FPEXC_CTL(current, arg2);
1804                         break;
1805                 case PR_GET_FPEXC:
1806                         error = GET_FPEXC_CTL(current, arg2);
1807                         break;
1808                 case PR_GET_TIMING:
1809                         error = PR_TIMING_STATISTICAL;
1810                         break;
1811                 case PR_SET_TIMING:
1812                         if (arg2 == PR_TIMING_STATISTICAL)
1813                                 error = 0;
1814                         else
1815                                 error = -EINVAL;
1816                         break;
1817
1818                 case PR_GET_KEEPCAPS:
1819                         if (current->keep_capabilities)
1820                                 error = 1;
1821                         break;
1822                 case PR_SET_KEEPCAPS:
1823                         if (arg2 != 0 && arg2 != 1) {
1824                                 error = -EINVAL;
1825                                 break;
1826                         }
1827                         current->keep_capabilities = arg2;
1828                         break;
1829                 case PR_SET_NAME: {
1830                         struct task_struct *me = current;
1831                         unsigned char ncomm[sizeof(me->comm)];
1832
1833                         ncomm[sizeof(me->comm)-1] = 0;
1834                         if (strncpy_from_user(ncomm, (char __user *)arg2,
1835                                                 sizeof(me->comm)-1) < 0)
1836                                 return -EFAULT;
1837                         set_task_comm(me, ncomm);
1838                         return 0;
1839                 }
1840                 case PR_GET_NAME: {
1841                         struct task_struct *me = current;
1842                         unsigned char tcomm[sizeof(me->comm)];
1843
1844                         get_task_comm(tcomm, me);
1845                         if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
1846                                 return -EFAULT;
1847                         return 0;
1848                 }
1849                 default:
1850                         error = -EINVAL;
1851                         break;
1852         }
1853         return error;
1854 }