[PATCH] Don't attempt to power off if power off is not implemented
[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 #include <linux/kprobes.h>
36
37 #include <asm/uaccess.h>
38 #include <asm/io.h>
39 #include <asm/unistd.h>
40
41 #ifndef SET_UNALIGN_CTL
42 # define SET_UNALIGN_CTL(a,b)   (-EINVAL)
43 #endif
44 #ifndef GET_UNALIGN_CTL
45 # define GET_UNALIGN_CTL(a,b)   (-EINVAL)
46 #endif
47 #ifndef SET_FPEMU_CTL
48 # define SET_FPEMU_CTL(a,b)     (-EINVAL)
49 #endif
50 #ifndef GET_FPEMU_CTL
51 # define GET_FPEMU_CTL(a,b)     (-EINVAL)
52 #endif
53 #ifndef SET_FPEXC_CTL
54 # define SET_FPEXC_CTL(a,b)     (-EINVAL)
55 #endif
56 #ifndef GET_FPEXC_CTL
57 # define GET_FPEXC_CTL(a,b)     (-EINVAL)
58 #endif
59
60 /*
61  * this is where the system-wide overflow UID and GID are defined, for
62  * architectures that now have 32-bit UID/GID but didn't in the past
63  */
64
65 int overflowuid = DEFAULT_OVERFLOWUID;
66 int overflowgid = DEFAULT_OVERFLOWGID;
67
68 #ifdef CONFIG_UID16
69 EXPORT_SYMBOL(overflowuid);
70 EXPORT_SYMBOL(overflowgid);
71 #endif
72
73 /*
74  * the same as above, but for filesystems which can only store a 16-bit
75  * UID and GID. as such, this is needed on all architectures
76  */
77
78 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
79 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
80
81 EXPORT_SYMBOL(fs_overflowuid);
82 EXPORT_SYMBOL(fs_overflowgid);
83
84 /*
85  * this indicates whether you can reboot with ctrl-alt-del: the default is yes
86  */
87
88 int C_A_D = 1;
89 int cad_pid = 1;
90
91 /*
92  *      Notifier list for kernel code which wants to be called
93  *      at shutdown. This is used to stop any idling DMA operations
94  *      and the like. 
95  */
96
97 static struct notifier_block *reboot_notifier_list;
98 static DEFINE_RWLOCK(notifier_lock);
99
100 /**
101  *      notifier_chain_register - Add notifier to a notifier chain
102  *      @list: Pointer to root list pointer
103  *      @n: New entry in notifier chain
104  *
105  *      Adds a notifier to a notifier chain.
106  *
107  *      Currently always returns zero.
108  */
109  
110 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
111 {
112         write_lock(&notifier_lock);
113         while(*list)
114         {
115                 if(n->priority > (*list)->priority)
116                         break;
117                 list= &((*list)->next);
118         }
119         n->next = *list;
120         *list=n;
121         write_unlock(&notifier_lock);
122         return 0;
123 }
124
125 EXPORT_SYMBOL(notifier_chain_register);
126
127 /**
128  *      notifier_chain_unregister - Remove notifier from a notifier chain
129  *      @nl: Pointer to root list pointer
130  *      @n: New entry in notifier chain
131  *
132  *      Removes a notifier from a notifier chain.
133  *
134  *      Returns zero on success, or %-ENOENT on failure.
135  */
136  
137 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
138 {
139         write_lock(&notifier_lock);
140         while((*nl)!=NULL)
141         {
142                 if((*nl)==n)
143                 {
144                         *nl=n->next;
145                         write_unlock(&notifier_lock);
146                         return 0;
147                 }
148                 nl=&((*nl)->next);
149         }
150         write_unlock(&notifier_lock);
151         return -ENOENT;
152 }
153
154 EXPORT_SYMBOL(notifier_chain_unregister);
155
156 /**
157  *      notifier_call_chain - Call functions in a notifier chain
158  *      @n: Pointer to root pointer of notifier chain
159  *      @val: Value passed unmodified to notifier function
160  *      @v: Pointer passed unmodified to notifier function
161  *
162  *      Calls each function in a notifier chain in turn.
163  *
164  *      If the return value of the notifier can be and'd
165  *      with %NOTIFY_STOP_MASK, then notifier_call_chain
166  *      will return immediately, with the return value of
167  *      the notifier function which halted execution.
168  *      Otherwise, the return value is the return value
169  *      of the last notifier function called.
170  */
171  
172 int __kprobes notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
173 {
174         int ret=NOTIFY_DONE;
175         struct notifier_block *nb = *n;
176
177         while(nb)
178         {
179                 ret=nb->notifier_call(nb,val,v);
180                 if(ret&NOTIFY_STOP_MASK)
181                 {
182                         return ret;
183                 }
184                 nb=nb->next;
185         }
186         return ret;
187 }
188
189 EXPORT_SYMBOL(notifier_call_chain);
190
191 /**
192  *      register_reboot_notifier - Register function to be called at reboot time
193  *      @nb: Info about notifier function to be called
194  *
195  *      Registers a function with the list of functions
196  *      to be called at reboot time.
197  *
198  *      Currently always returns zero, as notifier_chain_register
199  *      always returns zero.
200  */
201  
202 int register_reboot_notifier(struct notifier_block * nb)
203 {
204         return notifier_chain_register(&reboot_notifier_list, nb);
205 }
206
207 EXPORT_SYMBOL(register_reboot_notifier);
208
209 /**
210  *      unregister_reboot_notifier - Unregister previously registered reboot notifier
211  *      @nb: Hook to be unregistered
212  *
213  *      Unregisters a previously registered reboot
214  *      notifier function.
215  *
216  *      Returns zero on success, or %-ENOENT on failure.
217  */
218  
219 int unregister_reboot_notifier(struct notifier_block * nb)
220 {
221         return notifier_chain_unregister(&reboot_notifier_list, nb);
222 }
223
224 EXPORT_SYMBOL(unregister_reboot_notifier);
225
226 static int set_one_prio(struct task_struct *p, int niceval, int error)
227 {
228         int no_nice;
229
230         if (p->uid != current->euid &&
231                 p->euid != current->euid && !capable(CAP_SYS_NICE)) {
232                 error = -EPERM;
233                 goto out;
234         }
235         if (niceval < task_nice(p) && !can_nice(p, niceval)) {
236                 error = -EACCES;
237                 goto out;
238         }
239         no_nice = security_task_setnice(p, niceval);
240         if (no_nice) {
241                 error = no_nice;
242                 goto out;
243         }
244         if (error == -ESRCH)
245                 error = 0;
246         set_user_nice(p, niceval);
247 out:
248         return error;
249 }
250
251 asmlinkage long sys_setpriority(int which, int who, int niceval)
252 {
253         struct task_struct *g, *p;
254         struct user_struct *user;
255         int error = -EINVAL;
256
257         if (which > 2 || which < 0)
258                 goto out;
259
260         /* normalize: avoid signed division (rounding problems) */
261         error = -ESRCH;
262         if (niceval < -20)
263                 niceval = -20;
264         if (niceval > 19)
265                 niceval = 19;
266
267         read_lock(&tasklist_lock);
268         switch (which) {
269                 case PRIO_PROCESS:
270                         if (!who)
271                                 who = current->pid;
272                         p = find_task_by_pid(who);
273                         if (p)
274                                 error = set_one_prio(p, niceval, error);
275                         break;
276                 case PRIO_PGRP:
277                         if (!who)
278                                 who = process_group(current);
279                         do_each_task_pid(who, PIDTYPE_PGID, p) {
280                                 error = set_one_prio(p, niceval, error);
281                         } while_each_task_pid(who, PIDTYPE_PGID, p);
282                         break;
283                 case PRIO_USER:
284                         user = current->user;
285                         if (!who)
286                                 who = current->uid;
287                         else
288                                 if ((who != current->uid) && !(user = find_user(who)))
289                                         goto out_unlock;        /* No processes for this user */
290
291                         do_each_thread(g, p)
292                                 if (p->uid == who)
293                                         error = set_one_prio(p, niceval, error);
294                         while_each_thread(g, p);
295                         if (who != current->uid)
296                                 free_uid(user);         /* For find_user() */
297                         break;
298         }
299 out_unlock:
300         read_unlock(&tasklist_lock);
301 out:
302         return error;
303 }
304
305 /*
306  * Ugh. To avoid negative return values, "getpriority()" will
307  * not return the normal nice-value, but a negated value that
308  * has been offset by 20 (ie it returns 40..1 instead of -20..19)
309  * to stay compatible.
310  */
311 asmlinkage long sys_getpriority(int which, int who)
312 {
313         struct task_struct *g, *p;
314         struct user_struct *user;
315         long niceval, retval = -ESRCH;
316
317         if (which > 2 || which < 0)
318                 return -EINVAL;
319
320         read_lock(&tasklist_lock);
321         switch (which) {
322                 case PRIO_PROCESS:
323                         if (!who)
324                                 who = current->pid;
325                         p = find_task_by_pid(who);
326                         if (p) {
327                                 niceval = 20 - task_nice(p);
328                                 if (niceval > retval)
329                                         retval = niceval;
330                         }
331                         break;
332                 case PRIO_PGRP:
333                         if (!who)
334                                 who = process_group(current);
335                         do_each_task_pid(who, PIDTYPE_PGID, p) {
336                                 niceval = 20 - task_nice(p);
337                                 if (niceval > retval)
338                                         retval = niceval;
339                         } while_each_task_pid(who, PIDTYPE_PGID, p);
340                         break;
341                 case PRIO_USER:
342                         user = current->user;
343                         if (!who)
344                                 who = current->uid;
345                         else
346                                 if ((who != current->uid) && !(user = find_user(who)))
347                                         goto out_unlock;        /* No processes for this user */
348
349                         do_each_thread(g, p)
350                                 if (p->uid == who) {
351                                         niceval = 20 - task_nice(p);
352                                         if (niceval > retval)
353                                                 retval = niceval;
354                                 }
355                         while_each_thread(g, p);
356                         if (who != current->uid)
357                                 free_uid(user);         /* for find_user() */
358                         break;
359         }
360 out_unlock:
361         read_unlock(&tasklist_lock);
362
363         return retval;
364 }
365
366 /**
367  *      emergency_restart - reboot the system
368  *
369  *      Without shutting down any hardware or taking any locks
370  *      reboot the system.  This is called when we know we are in
371  *      trouble so this is our best effort to reboot.  This is
372  *      safe to call in interrupt context.
373  */
374 void emergency_restart(void)
375 {
376         machine_emergency_restart();
377 }
378 EXPORT_SYMBOL_GPL(emergency_restart);
379
380 void kernel_restart_prepare(char *cmd)
381 {
382         notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd);
383         system_state = SYSTEM_RESTART;
384         device_shutdown();
385 }
386
387 /**
388  *      kernel_restart - reboot the system
389  *      @cmd: pointer to buffer containing command to execute for restart
390  *              or %NULL
391  *
392  *      Shutdown everything and perform a clean reboot.
393  *      This is not safe to call in interrupt context.
394  */
395 void kernel_restart(char *cmd)
396 {
397         kernel_restart_prepare(cmd);
398         if (!cmd) {
399                 printk(KERN_EMERG "Restarting system.\n");
400         } else {
401                 printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd);
402         }
403         printk(".\n");
404         machine_restart(cmd);
405 }
406 EXPORT_SYMBOL_GPL(kernel_restart);
407
408 /**
409  *      kernel_kexec - reboot the system
410  *
411  *      Move into place and start executing a preloaded standalone
412  *      executable.  If nothing was preloaded return an error.
413  */
414 void kernel_kexec(void)
415 {
416 #ifdef CONFIG_KEXEC
417         struct kimage *image;
418         image = xchg(&kexec_image, 0);
419         if (!image) {
420                 return;
421         }
422         kernel_restart_prepare(NULL);
423         printk(KERN_EMERG "Starting new kernel\n");
424         machine_shutdown();
425         machine_kexec(image);
426 #endif
427 }
428 EXPORT_SYMBOL_GPL(kernel_kexec);
429
430 /**
431  *      kernel_halt - halt the system
432  *
433  *      Shutdown everything and perform a clean system halt.
434  */
435 void kernel_halt_prepare(void)
436 {
437         notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
438         system_state = SYSTEM_HALT;
439         device_shutdown();
440 }
441 void kernel_halt(void)
442 {
443         kernel_halt_prepare();
444         printk(KERN_EMERG "System halted.\n");
445         machine_halt();
446 }
447 EXPORT_SYMBOL_GPL(kernel_halt);
448
449 /**
450  *      kernel_power_off - power_off the system
451  *
452  *      Shutdown everything and perform a clean system power_off.
453  */
454 void kernel_power_off_prepare(void)
455 {
456         notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
457         system_state = SYSTEM_POWER_OFF;
458         device_shutdown();
459 }
460 void kernel_power_off(void)
461 {
462         kernel_power_off_prepare();
463         printk(KERN_EMERG "Power down.\n");
464         machine_power_off();
465 }
466 EXPORT_SYMBOL_GPL(kernel_power_off);
467
468 /*
469  * Reboot system call: for obvious reasons only root may call it,
470  * and even root needs to set up some magic numbers in the registers
471  * so that some mistake won't make this reboot the whole machine.
472  * You can also set the meaning of the ctrl-alt-del-key here.
473  *
474  * reboot doesn't sync: do that yourself before calling this.
475  */
476 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
477 {
478         char buffer[256];
479
480         /* We only trust the superuser with rebooting the system. */
481         if (!capable(CAP_SYS_BOOT))
482                 return -EPERM;
483
484         /* For safety, we require "magic" arguments. */
485         if (magic1 != LINUX_REBOOT_MAGIC1 ||
486             (magic2 != LINUX_REBOOT_MAGIC2 &&
487                         magic2 != LINUX_REBOOT_MAGIC2A &&
488                         magic2 != LINUX_REBOOT_MAGIC2B &&
489                         magic2 != LINUX_REBOOT_MAGIC2C))
490                 return -EINVAL;
491
492         /* Instead of trying to make the power_off code look like
493          * halt when pm_power_off is not set do it the easy way.
494          */
495         if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off)
496                 cmd = LINUX_REBOOT_CMD_HALT;
497
498         lock_kernel();
499         switch (cmd) {
500         case LINUX_REBOOT_CMD_RESTART:
501                 kernel_restart(NULL);
502                 break;
503
504         case LINUX_REBOOT_CMD_CAD_ON:
505                 C_A_D = 1;
506                 break;
507
508         case LINUX_REBOOT_CMD_CAD_OFF:
509                 C_A_D = 0;
510                 break;
511
512         case LINUX_REBOOT_CMD_HALT:
513                 kernel_halt();
514                 unlock_kernel();
515                 do_exit(0);
516                 break;
517
518         case LINUX_REBOOT_CMD_POWER_OFF:
519                 kernel_power_off();
520                 unlock_kernel();
521                 do_exit(0);
522                 break;
523
524         case LINUX_REBOOT_CMD_RESTART2:
525                 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
526                         unlock_kernel();
527                         return -EFAULT;
528                 }
529                 buffer[sizeof(buffer) - 1] = '\0';
530
531                 kernel_restart(buffer);
532                 break;
533
534         case LINUX_REBOOT_CMD_KEXEC:
535                 kernel_kexec();
536                 unlock_kernel();
537                 return -EINVAL;
538
539 #ifdef CONFIG_SOFTWARE_SUSPEND
540         case LINUX_REBOOT_CMD_SW_SUSPEND:
541                 {
542                         int ret = software_suspend();
543                         unlock_kernel();
544                         return ret;
545                 }
546 #endif
547
548         default:
549                 unlock_kernel();
550                 return -EINVAL;
551         }
552         unlock_kernel();
553         return 0;
554 }
555
556 static void deferred_cad(void *dummy)
557 {
558         kernel_restart(NULL);
559 }
560
561 /*
562  * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
563  * As it's called within an interrupt, it may NOT sync: the only choice
564  * is whether to reboot at once, or just ignore the ctrl-alt-del.
565  */
566 void ctrl_alt_del(void)
567 {
568         static DECLARE_WORK(cad_work, deferred_cad, NULL);
569
570         if (C_A_D)
571                 schedule_work(&cad_work);
572         else
573                 kill_proc(cad_pid, SIGINT, 1);
574 }
575         
576
577 /*
578  * Unprivileged users may change the real gid to the effective gid
579  * or vice versa.  (BSD-style)
580  *
581  * If you set the real gid at all, or set the effective gid to a value not
582  * equal to the real gid, then the saved gid is set to the new effective gid.
583  *
584  * This makes it possible for a setgid program to completely drop its
585  * privileges, which is often a useful assertion to make when you are doing
586  * a security audit over a program.
587  *
588  * The general idea is that a program which uses just setregid() will be
589  * 100% compatible with BSD.  A program which uses just setgid() will be
590  * 100% compatible with POSIX with saved IDs. 
591  *
592  * SMP: There are not races, the GIDs are checked only by filesystem
593  *      operations (as far as semantic preservation is concerned).
594  */
595 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
596 {
597         int old_rgid = current->gid;
598         int old_egid = current->egid;
599         int new_rgid = old_rgid;
600         int new_egid = old_egid;
601         int retval;
602
603         retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
604         if (retval)
605                 return retval;
606
607         if (rgid != (gid_t) -1) {
608                 if ((old_rgid == rgid) ||
609                     (current->egid==rgid) ||
610                     capable(CAP_SETGID))
611                         new_rgid = rgid;
612                 else
613                         return -EPERM;
614         }
615         if (egid != (gid_t) -1) {
616                 if ((old_rgid == egid) ||
617                     (current->egid == egid) ||
618                     (current->sgid == egid) ||
619                     capable(CAP_SETGID))
620                         new_egid = egid;
621                 else {
622                         return -EPERM;
623                 }
624         }
625         if (new_egid != old_egid)
626         {
627                 current->mm->dumpable = suid_dumpable;
628                 smp_wmb();
629         }
630         if (rgid != (gid_t) -1 ||
631             (egid != (gid_t) -1 && egid != old_rgid))
632                 current->sgid = new_egid;
633         current->fsgid = new_egid;
634         current->egid = new_egid;
635         current->gid = new_rgid;
636         key_fsgid_changed(current);
637         proc_id_connector(current, PROC_EVENT_GID);
638         return 0;
639 }
640
641 /*
642  * setgid() is implemented like SysV w/ SAVED_IDS 
643  *
644  * SMP: Same implicit races as above.
645  */
646 asmlinkage long sys_setgid(gid_t gid)
647 {
648         int old_egid = current->egid;
649         int retval;
650
651         retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
652         if (retval)
653                 return retval;
654
655         if (capable(CAP_SETGID))
656         {
657                 if(old_egid != gid)
658                 {
659                         current->mm->dumpable = suid_dumpable;
660                         smp_wmb();
661                 }
662                 current->gid = current->egid = current->sgid = current->fsgid = gid;
663         }
664         else if ((gid == current->gid) || (gid == current->sgid))
665         {
666                 if(old_egid != gid)
667                 {
668                         current->mm->dumpable = suid_dumpable;
669                         smp_wmb();
670                 }
671                 current->egid = current->fsgid = gid;
672         }
673         else
674                 return -EPERM;
675
676         key_fsgid_changed(current);
677         proc_id_connector(current, PROC_EVENT_GID);
678         return 0;
679 }
680   
681 static int set_user(uid_t new_ruid, int dumpclear)
682 {
683         struct user_struct *new_user;
684
685         new_user = alloc_uid(new_ruid);
686         if (!new_user)
687                 return -EAGAIN;
688
689         if (atomic_read(&new_user->processes) >=
690                                 current->signal->rlim[RLIMIT_NPROC].rlim_cur &&
691                         new_user != &root_user) {
692                 free_uid(new_user);
693                 return -EAGAIN;
694         }
695
696         switch_uid(new_user);
697
698         if(dumpclear)
699         {
700                 current->mm->dumpable = suid_dumpable;
701                 smp_wmb();
702         }
703         current->uid = new_ruid;
704         return 0;
705 }
706
707 /*
708  * Unprivileged users may change the real uid to the effective uid
709  * or vice versa.  (BSD-style)
710  *
711  * If you set the real uid at all, or set the effective uid to a value not
712  * equal to the real uid, then the saved uid is set to the new effective uid.
713  *
714  * This makes it possible for a setuid program to completely drop its
715  * privileges, which is often a useful assertion to make when you are doing
716  * a security audit over a program.
717  *
718  * The general idea is that a program which uses just setreuid() will be
719  * 100% compatible with BSD.  A program which uses just setuid() will be
720  * 100% compatible with POSIX with saved IDs. 
721  */
722 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
723 {
724         int old_ruid, old_euid, old_suid, new_ruid, new_euid;
725         int retval;
726
727         retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
728         if (retval)
729                 return retval;
730
731         new_ruid = old_ruid = current->uid;
732         new_euid = old_euid = current->euid;
733         old_suid = current->suid;
734
735         if (ruid != (uid_t) -1) {
736                 new_ruid = ruid;
737                 if ((old_ruid != ruid) &&
738                     (current->euid != ruid) &&
739                     !capable(CAP_SETUID))
740                         return -EPERM;
741         }
742
743         if (euid != (uid_t) -1) {
744                 new_euid = euid;
745                 if ((old_ruid != euid) &&
746                     (current->euid != euid) &&
747                     (current->suid != euid) &&
748                     !capable(CAP_SETUID))
749                         return -EPERM;
750         }
751
752         if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
753                 return -EAGAIN;
754
755         if (new_euid != old_euid)
756         {
757                 current->mm->dumpable = suid_dumpable;
758                 smp_wmb();
759         }
760         current->fsuid = current->euid = new_euid;
761         if (ruid != (uid_t) -1 ||
762             (euid != (uid_t) -1 && euid != old_ruid))
763                 current->suid = current->euid;
764         current->fsuid = current->euid;
765
766         key_fsuid_changed(current);
767         proc_id_connector(current, PROC_EVENT_UID);
768
769         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
770 }
771
772
773                 
774 /*
775  * setuid() is implemented like SysV with SAVED_IDS 
776  * 
777  * Note that SAVED_ID's is deficient in that a setuid root program
778  * like sendmail, for example, cannot set its uid to be a normal 
779  * user and then switch back, because if you're root, setuid() sets
780  * the saved uid too.  If you don't like this, blame the bright people
781  * in the POSIX committee and/or USG.  Note that the BSD-style setreuid()
782  * will allow a root program to temporarily drop privileges and be able to
783  * regain them by swapping the real and effective uid.  
784  */
785 asmlinkage long sys_setuid(uid_t uid)
786 {
787         int old_euid = current->euid;
788         int old_ruid, old_suid, new_ruid, new_suid;
789         int retval;
790
791         retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
792         if (retval)
793                 return retval;
794
795         old_ruid = new_ruid = current->uid;
796         old_suid = current->suid;
797         new_suid = old_suid;
798         
799         if (capable(CAP_SETUID)) {
800                 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
801                         return -EAGAIN;
802                 new_suid = uid;
803         } else if ((uid != current->uid) && (uid != new_suid))
804                 return -EPERM;
805
806         if (old_euid != uid)
807         {
808                 current->mm->dumpable = suid_dumpable;
809                 smp_wmb();
810         }
811         current->fsuid = current->euid = uid;
812         current->suid = new_suid;
813
814         key_fsuid_changed(current);
815         proc_id_connector(current, PROC_EVENT_UID);
816
817         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
818 }
819
820
821 /*
822  * This function implements a generic ability to update ruid, euid,
823  * and suid.  This allows you to implement the 4.4 compatible seteuid().
824  */
825 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
826 {
827         int old_ruid = current->uid;
828         int old_euid = current->euid;
829         int old_suid = current->suid;
830         int retval;
831
832         retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
833         if (retval)
834                 return retval;
835
836         if (!capable(CAP_SETUID)) {
837                 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
838                     (ruid != current->euid) && (ruid != current->suid))
839                         return -EPERM;
840                 if ((euid != (uid_t) -1) && (euid != current->uid) &&
841                     (euid != current->euid) && (euid != current->suid))
842                         return -EPERM;
843                 if ((suid != (uid_t) -1) && (suid != current->uid) &&
844                     (suid != current->euid) && (suid != current->suid))
845                         return -EPERM;
846         }
847         if (ruid != (uid_t) -1) {
848                 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
849                         return -EAGAIN;
850         }
851         if (euid != (uid_t) -1) {
852                 if (euid != current->euid)
853                 {
854                         current->mm->dumpable = suid_dumpable;
855                         smp_wmb();
856                 }
857                 current->euid = euid;
858         }
859         current->fsuid = current->euid;
860         if (suid != (uid_t) -1)
861                 current->suid = suid;
862
863         key_fsuid_changed(current);
864         proc_id_connector(current, PROC_EVENT_UID);
865
866         return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
867 }
868
869 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
870 {
871         int retval;
872
873         if (!(retval = put_user(current->uid, ruid)) &&
874             !(retval = put_user(current->euid, euid)))
875                 retval = put_user(current->suid, suid);
876
877         return retval;
878 }
879
880 /*
881  * Same as above, but for rgid, egid, sgid.
882  */
883 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
884 {
885         int retval;
886
887         retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
888         if (retval)
889                 return retval;
890
891         if (!capable(CAP_SETGID)) {
892                 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
893                     (rgid != current->egid) && (rgid != current->sgid))
894                         return -EPERM;
895                 if ((egid != (gid_t) -1) && (egid != current->gid) &&
896                     (egid != current->egid) && (egid != current->sgid))
897                         return -EPERM;
898                 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
899                     (sgid != current->egid) && (sgid != current->sgid))
900                         return -EPERM;
901         }
902         if (egid != (gid_t) -1) {
903                 if (egid != current->egid)
904                 {
905                         current->mm->dumpable = suid_dumpable;
906                         smp_wmb();
907                 }
908                 current->egid = egid;
909         }
910         current->fsgid = current->egid;
911         if (rgid != (gid_t) -1)
912                 current->gid = rgid;
913         if (sgid != (gid_t) -1)
914                 current->sgid = sgid;
915
916         key_fsgid_changed(current);
917         proc_id_connector(current, PROC_EVENT_GID);
918         return 0;
919 }
920
921 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
922 {
923         int retval;
924
925         if (!(retval = put_user(current->gid, rgid)) &&
926             !(retval = put_user(current->egid, egid)))
927                 retval = put_user(current->sgid, sgid);
928
929         return retval;
930 }
931
932
933 /*
934  * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
935  * is used for "access()" and for the NFS daemon (letting nfsd stay at
936  * whatever uid it wants to). It normally shadows "euid", except when
937  * explicitly set by setfsuid() or for access..
938  */
939 asmlinkage long sys_setfsuid(uid_t uid)
940 {
941         int old_fsuid;
942
943         old_fsuid = current->fsuid;
944         if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
945                 return old_fsuid;
946
947         if (uid == current->uid || uid == current->euid ||
948             uid == current->suid || uid == current->fsuid || 
949             capable(CAP_SETUID))
950         {
951                 if (uid != old_fsuid)
952                 {
953                         current->mm->dumpable = suid_dumpable;
954                         smp_wmb();
955                 }
956                 current->fsuid = uid;
957         }
958
959         key_fsuid_changed(current);
960         proc_id_connector(current, PROC_EVENT_UID);
961
962         security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
963
964         return old_fsuid;
965 }
966
967 /*
968  * Samma på svenska..
969  */
970 asmlinkage long sys_setfsgid(gid_t gid)
971 {
972         int old_fsgid;
973
974         old_fsgid = current->fsgid;
975         if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
976                 return old_fsgid;
977
978         if (gid == current->gid || gid == current->egid ||
979             gid == current->sgid || gid == current->fsgid || 
980             capable(CAP_SETGID))
981         {
982                 if (gid != old_fsgid)
983                 {
984                         current->mm->dumpable = suid_dumpable;
985                         smp_wmb();
986                 }
987                 current->fsgid = gid;
988                 key_fsgid_changed(current);
989                 proc_id_connector(current, PROC_EVENT_GID);
990         }
991         return old_fsgid;
992 }
993
994 asmlinkage long sys_times(struct tms __user * tbuf)
995 {
996         /*
997          *      In the SMP world we might just be unlucky and have one of
998          *      the times increment as we use it. Since the value is an
999          *      atomically safe type this is just fine. Conceptually its
1000          *      as if the syscall took an instant longer to occur.
1001          */
1002         if (tbuf) {
1003                 struct tms tmp;
1004                 cputime_t utime, stime, cutime, cstime;
1005
1006 #ifdef CONFIG_SMP
1007                 if (thread_group_empty(current)) {
1008                         /*
1009                          * Single thread case without the use of any locks.
1010                          *
1011                          * We may race with release_task if two threads are
1012                          * executing. However, release task first adds up the
1013                          * counters (__exit_signal) before  removing the task
1014                          * from the process tasklist (__unhash_process).
1015                          * __exit_signal also acquires and releases the
1016                          * siglock which results in the proper memory ordering
1017                          * so that the list modifications are always visible
1018                          * after the counters have been updated.
1019                          *
1020                          * If the counters have been updated by the second thread
1021                          * but the thread has not yet been removed from the list
1022                          * then the other branch will be executing which will
1023                          * block on tasklist_lock until the exit handling of the
1024                          * other task is finished.
1025                          *
1026                          * This also implies that the sighand->siglock cannot
1027                          * be held by another processor. So we can also
1028                          * skip acquiring that lock.
1029                          */
1030                         utime = cputime_add(current->signal->utime, current->utime);
1031                         stime = cputime_add(current->signal->utime, current->stime);
1032                         cutime = current->signal->cutime;
1033                         cstime = current->signal->cstime;
1034                 } else
1035 #endif
1036                 {
1037
1038                         /* Process with multiple threads */
1039                         struct task_struct *tsk = current;
1040                         struct task_struct *t;
1041
1042                         read_lock(&tasklist_lock);
1043                         utime = tsk->signal->utime;
1044                         stime = tsk->signal->stime;
1045                         t = tsk;
1046                         do {
1047                                 utime = cputime_add(utime, t->utime);
1048                                 stime = cputime_add(stime, t->stime);
1049                                 t = next_thread(t);
1050                         } while (t != tsk);
1051
1052                         /*
1053                          * While we have tasklist_lock read-locked, no dying thread
1054                          * can be updating current->signal->[us]time.  Instead,
1055                          * we got their counts included in the live thread loop.
1056                          * However, another thread can come in right now and
1057                          * do a wait call that updates current->signal->c[us]time.
1058                          * To make sure we always see that pair updated atomically,
1059                          * we take the siglock around fetching them.
1060                          */
1061                         spin_lock_irq(&tsk->sighand->siglock);
1062                         cutime = tsk->signal->cutime;
1063                         cstime = tsk->signal->cstime;
1064                         spin_unlock_irq(&tsk->sighand->siglock);
1065                         read_unlock(&tasklist_lock);
1066                 }
1067                 tmp.tms_utime = cputime_to_clock_t(utime);
1068                 tmp.tms_stime = cputime_to_clock_t(stime);
1069                 tmp.tms_cutime = cputime_to_clock_t(cutime);
1070                 tmp.tms_cstime = cputime_to_clock_t(cstime);
1071                 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1072                         return -EFAULT;
1073         }
1074         return (long) jiffies_64_to_clock_t(get_jiffies_64());
1075 }
1076
1077 /*
1078  * This needs some heavy checking ...
1079  * I just haven't the stomach for it. I also don't fully
1080  * understand sessions/pgrp etc. Let somebody who does explain it.
1081  *
1082  * OK, I think I have the protection semantics right.... this is really
1083  * only important on a multi-user system anyway, to make sure one user
1084  * can't send a signal to a process owned by another.  -TYT, 12/12/91
1085  *
1086  * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1087  * LBT 04.03.94
1088  */
1089
1090 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1091 {
1092         struct task_struct *p;
1093         int err = -EINVAL;
1094
1095         if (!pid)
1096                 pid = current->pid;
1097         if (!pgid)
1098                 pgid = pid;
1099         if (pgid < 0)
1100                 return -EINVAL;
1101
1102         /* From this point forward we keep holding onto the tasklist lock
1103          * so that our parent does not change from under us. -DaveM
1104          */
1105         write_lock_irq(&tasklist_lock);
1106
1107         err = -ESRCH;
1108         p = find_task_by_pid(pid);
1109         if (!p)
1110                 goto out;
1111
1112         err = -EINVAL;
1113         if (!thread_group_leader(p))
1114                 goto out;
1115
1116         if (p->parent == current || p->real_parent == current) {
1117                 err = -EPERM;
1118                 if (p->signal->session != current->signal->session)
1119                         goto out;
1120                 err = -EACCES;
1121                 if (p->did_exec)
1122                         goto out;
1123         } else {
1124                 err = -ESRCH;
1125                 if (p != current)
1126                         goto out;
1127         }
1128
1129         err = -EPERM;
1130         if (p->signal->leader)
1131                 goto out;
1132
1133         if (pgid != pid) {
1134                 struct task_struct *p;
1135
1136                 do_each_task_pid(pgid, PIDTYPE_PGID, p) {
1137                         if (p->signal->session == current->signal->session)
1138                                 goto ok_pgid;
1139                 } while_each_task_pid(pgid, PIDTYPE_PGID, p);
1140                 goto out;
1141         }
1142
1143 ok_pgid:
1144         err = security_task_setpgid(p, pgid);
1145         if (err)
1146                 goto out;
1147
1148         if (process_group(p) != pgid) {
1149                 detach_pid(p, PIDTYPE_PGID);
1150                 p->signal->pgrp = pgid;
1151                 attach_pid(p, PIDTYPE_PGID, pgid);
1152         }
1153
1154         err = 0;
1155 out:
1156         /* All paths lead to here, thus we are safe. -DaveM */
1157         write_unlock_irq(&tasklist_lock);
1158         return err;
1159 }
1160
1161 asmlinkage long sys_getpgid(pid_t pid)
1162 {
1163         if (!pid) {
1164                 return process_group(current);
1165         } else {
1166                 int retval;
1167                 struct task_struct *p;
1168
1169                 read_lock(&tasklist_lock);
1170                 p = find_task_by_pid(pid);
1171
1172                 retval = -ESRCH;
1173                 if (p) {
1174                         retval = security_task_getpgid(p);
1175                         if (!retval)
1176                                 retval = process_group(p);
1177                 }
1178                 read_unlock(&tasklist_lock);
1179                 return retval;
1180         }
1181 }
1182
1183 #ifdef __ARCH_WANT_SYS_GETPGRP
1184
1185 asmlinkage long sys_getpgrp(void)
1186 {
1187         /* SMP - assuming writes are word atomic this is fine */
1188         return process_group(current);
1189 }
1190
1191 #endif
1192
1193 asmlinkage long sys_getsid(pid_t pid)
1194 {
1195         if (!pid) {
1196                 return current->signal->session;
1197         } else {
1198                 int retval;
1199                 struct task_struct *p;
1200
1201                 read_lock(&tasklist_lock);
1202                 p = find_task_by_pid(pid);
1203
1204                 retval = -ESRCH;
1205                 if(p) {
1206                         retval = security_task_getsid(p);
1207                         if (!retval)
1208                                 retval = p->signal->session;
1209                 }
1210                 read_unlock(&tasklist_lock);
1211                 return retval;
1212         }
1213 }
1214
1215 asmlinkage long sys_setsid(void)
1216 {
1217         struct pid *pid;
1218         int err = -EPERM;
1219
1220         if (!thread_group_leader(current))
1221                 return -EINVAL;
1222
1223         down(&tty_sem);
1224         write_lock_irq(&tasklist_lock);
1225
1226         pid = find_pid(PIDTYPE_PGID, current->pid);
1227         if (pid)
1228                 goto out;
1229
1230         current->signal->leader = 1;
1231         __set_special_pids(current->pid, current->pid);
1232         current->signal->tty = NULL;
1233         current->signal->tty_old_pgrp = 0;
1234         err = process_group(current);
1235 out:
1236         write_unlock_irq(&tasklist_lock);
1237         up(&tty_sem);
1238         return err;
1239 }
1240
1241 /*
1242  * Supplementary group IDs
1243  */
1244
1245 /* init to 2 - one for init_task, one to ensure it is never freed */
1246 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1247
1248 struct group_info *groups_alloc(int gidsetsize)
1249 {
1250         struct group_info *group_info;
1251         int nblocks;
1252         int i;
1253
1254         nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1255         /* Make sure we always allocate at least one indirect block pointer */
1256         nblocks = nblocks ? : 1;
1257         group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1258         if (!group_info)
1259                 return NULL;
1260         group_info->ngroups = gidsetsize;
1261         group_info->nblocks = nblocks;
1262         atomic_set(&group_info->usage, 1);
1263
1264         if (gidsetsize <= NGROUPS_SMALL) {
1265                 group_info->blocks[0] = group_info->small_block;
1266         } else {
1267                 for (i = 0; i < nblocks; i++) {
1268                         gid_t *b;
1269                         b = (void *)__get_free_page(GFP_USER);
1270                         if (!b)
1271                                 goto out_undo_partial_alloc;
1272                         group_info->blocks[i] = b;
1273                 }
1274         }
1275         return group_info;
1276
1277 out_undo_partial_alloc:
1278         while (--i >= 0) {
1279                 free_page((unsigned long)group_info->blocks[i]);
1280         }
1281         kfree(group_info);
1282         return NULL;
1283 }
1284
1285 EXPORT_SYMBOL(groups_alloc);
1286
1287 void groups_free(struct group_info *group_info)
1288 {
1289         if (group_info->blocks[0] != group_info->small_block) {
1290                 int i;
1291                 for (i = 0; i < group_info->nblocks; i++)
1292                         free_page((unsigned long)group_info->blocks[i]);
1293         }
1294         kfree(group_info);
1295 }
1296
1297 EXPORT_SYMBOL(groups_free);
1298
1299 /* export the group_info to a user-space array */
1300 static int groups_to_user(gid_t __user *grouplist,
1301     struct group_info *group_info)
1302 {
1303         int i;
1304         int count = group_info->ngroups;
1305
1306         for (i = 0; i < group_info->nblocks; i++) {
1307                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1308                 int off = i * NGROUPS_PER_BLOCK;
1309                 int len = cp_count * sizeof(*grouplist);
1310
1311                 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1312                         return -EFAULT;
1313
1314                 count -= cp_count;
1315         }
1316         return 0;
1317 }
1318
1319 /* fill a group_info from a user-space array - it must be allocated already */
1320 static int groups_from_user(struct group_info *group_info,
1321     gid_t __user *grouplist)
1322  {
1323         int i;
1324         int count = group_info->ngroups;
1325
1326         for (i = 0; i < group_info->nblocks; i++) {
1327                 int cp_count = min(NGROUPS_PER_BLOCK, count);
1328                 int off = i * NGROUPS_PER_BLOCK;
1329                 int len = cp_count * sizeof(*grouplist);
1330
1331                 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1332                         return -EFAULT;
1333
1334                 count -= cp_count;
1335         }
1336         return 0;
1337 }
1338
1339 /* a simple Shell sort */
1340 static void groups_sort(struct group_info *group_info)
1341 {
1342         int base, max, stride;
1343         int gidsetsize = group_info->ngroups;
1344
1345         for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1346                 ; /* nothing */
1347         stride /= 3;
1348
1349         while (stride) {
1350                 max = gidsetsize - stride;
1351                 for (base = 0; base < max; base++) {
1352                         int left = base;
1353                         int right = left + stride;
1354                         gid_t tmp = GROUP_AT(group_info, right);
1355
1356                         while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1357                                 GROUP_AT(group_info, right) =
1358                                     GROUP_AT(group_info, left);
1359                                 right = left;
1360                                 left -= stride;
1361                         }
1362                         GROUP_AT(group_info, right) = tmp;
1363                 }
1364                 stride /= 3;
1365         }
1366 }
1367
1368 /* a simple bsearch */
1369 int groups_search(struct group_info *group_info, gid_t grp)
1370 {
1371         int left, right;
1372
1373         if (!group_info)
1374                 return 0;
1375
1376         left = 0;
1377         right = group_info->ngroups;
1378         while (left < right) {
1379                 int mid = (left+right)/2;
1380                 int cmp = grp - GROUP_AT(group_info, mid);
1381                 if (cmp > 0)
1382                         left = mid + 1;
1383                 else if (cmp < 0)
1384                         right = mid;
1385                 else
1386                         return 1;
1387         }
1388         return 0;
1389 }
1390
1391 /* validate and set current->group_info */
1392 int set_current_groups(struct group_info *group_info)
1393 {
1394         int retval;
1395         struct group_info *old_info;
1396
1397         retval = security_task_setgroups(group_info);
1398         if (retval)
1399                 return retval;
1400
1401         groups_sort(group_info);
1402         get_group_info(group_info);
1403
1404         task_lock(current);
1405         old_info = current->group_info;
1406         current->group_info = group_info;
1407         task_unlock(current);
1408
1409         put_group_info(old_info);
1410
1411         return 0;
1412 }
1413
1414 EXPORT_SYMBOL(set_current_groups);
1415
1416 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1417 {
1418         int i = 0;
1419
1420         /*
1421          *      SMP: Nobody else can change our grouplist. Thus we are
1422          *      safe.
1423          */
1424
1425         if (gidsetsize < 0)
1426                 return -EINVAL;
1427
1428         /* no need to grab task_lock here; it cannot change */
1429         get_group_info(current->group_info);
1430         i = current->group_info->ngroups;
1431         if (gidsetsize) {
1432                 if (i > gidsetsize) {
1433                         i = -EINVAL;
1434                         goto out;
1435                 }
1436                 if (groups_to_user(grouplist, current->group_info)) {
1437                         i = -EFAULT;
1438                         goto out;
1439                 }
1440         }
1441 out:
1442         put_group_info(current->group_info);
1443         return i;
1444 }
1445
1446 /*
1447  *      SMP: Our groups are copy-on-write. We can set them safely
1448  *      without another task interfering.
1449  */
1450  
1451 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1452 {
1453         struct group_info *group_info;
1454         int retval;
1455
1456         if (!capable(CAP_SETGID))
1457                 return -EPERM;
1458         if ((unsigned)gidsetsize > NGROUPS_MAX)
1459                 return -EINVAL;
1460
1461         group_info = groups_alloc(gidsetsize);
1462         if (!group_info)
1463                 return -ENOMEM;
1464         retval = groups_from_user(group_info, grouplist);
1465         if (retval) {
1466                 put_group_info(group_info);
1467                 return retval;
1468         }
1469
1470         retval = set_current_groups(group_info);
1471         put_group_info(group_info);
1472
1473         return retval;
1474 }
1475
1476 /*
1477  * Check whether we're fsgid/egid or in the supplemental group..
1478  */
1479 int in_group_p(gid_t grp)
1480 {
1481         int retval = 1;
1482         if (grp != current->fsgid) {
1483                 get_group_info(current->group_info);
1484                 retval = groups_search(current->group_info, grp);
1485                 put_group_info(current->group_info);
1486         }
1487         return retval;
1488 }
1489
1490 EXPORT_SYMBOL(in_group_p);
1491
1492 int in_egroup_p(gid_t grp)
1493 {
1494         int retval = 1;
1495         if (grp != current->egid) {
1496                 get_group_info(current->group_info);
1497                 retval = groups_search(current->group_info, grp);
1498                 put_group_info(current->group_info);
1499         }
1500         return retval;
1501 }
1502
1503 EXPORT_SYMBOL(in_egroup_p);
1504
1505 DECLARE_RWSEM(uts_sem);
1506
1507 EXPORT_SYMBOL(uts_sem);
1508
1509 asmlinkage long sys_newuname(struct new_utsname __user * name)
1510 {
1511         int errno = 0;
1512
1513         down_read(&uts_sem);
1514         if (copy_to_user(name,&system_utsname,sizeof *name))
1515                 errno = -EFAULT;
1516         up_read(&uts_sem);
1517         return errno;
1518 }
1519
1520 asmlinkage long sys_sethostname(char __user *name, int len)
1521 {
1522         int errno;
1523         char tmp[__NEW_UTS_LEN];
1524
1525         if (!capable(CAP_SYS_ADMIN))
1526                 return -EPERM;
1527         if (len < 0 || len > __NEW_UTS_LEN)
1528                 return -EINVAL;
1529         down_write(&uts_sem);
1530         errno = -EFAULT;
1531         if (!copy_from_user(tmp, name, len)) {
1532                 memcpy(system_utsname.nodename, tmp, len);
1533                 system_utsname.nodename[len] = 0;
1534                 errno = 0;
1535         }
1536         up_write(&uts_sem);
1537         return errno;
1538 }
1539
1540 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1541
1542 asmlinkage long sys_gethostname(char __user *name, int len)
1543 {
1544         int i, errno;
1545
1546         if (len < 0)
1547                 return -EINVAL;
1548         down_read(&uts_sem);
1549         i = 1 + strlen(system_utsname.nodename);
1550         if (i > len)
1551                 i = len;
1552         errno = 0;
1553         if (copy_to_user(name, system_utsname.nodename, i))
1554                 errno = -EFAULT;
1555         up_read(&uts_sem);
1556         return errno;
1557 }
1558
1559 #endif
1560
1561 /*
1562  * Only setdomainname; getdomainname can be implemented by calling
1563  * uname()
1564  */
1565 asmlinkage long sys_setdomainname(char __user *name, int len)
1566 {
1567         int errno;
1568         char tmp[__NEW_UTS_LEN];
1569
1570         if (!capable(CAP_SYS_ADMIN))
1571                 return -EPERM;
1572         if (len < 0 || len > __NEW_UTS_LEN)
1573                 return -EINVAL;
1574
1575         down_write(&uts_sem);
1576         errno = -EFAULT;
1577         if (!copy_from_user(tmp, name, len)) {
1578                 memcpy(system_utsname.domainname, tmp, len);
1579                 system_utsname.domainname[len] = 0;
1580                 errno = 0;
1581         }
1582         up_write(&uts_sem);
1583         return errno;
1584 }
1585
1586 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1587 {
1588         if (resource >= RLIM_NLIMITS)
1589                 return -EINVAL;
1590         else {
1591                 struct rlimit value;
1592                 task_lock(current->group_leader);
1593                 value = current->signal->rlim[resource];
1594                 task_unlock(current->group_leader);
1595                 return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0;
1596         }
1597 }
1598
1599 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1600
1601 /*
1602  *      Back compatibility for getrlimit. Needed for some apps.
1603  */
1604  
1605 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1606 {
1607         struct rlimit x;
1608         if (resource >= RLIM_NLIMITS)
1609                 return -EINVAL;
1610
1611         task_lock(current->group_leader);
1612         x = current->signal->rlim[resource];
1613         task_unlock(current->group_leader);
1614         if(x.rlim_cur > 0x7FFFFFFF)
1615                 x.rlim_cur = 0x7FFFFFFF;
1616         if(x.rlim_max > 0x7FFFFFFF)
1617                 x.rlim_max = 0x7FFFFFFF;
1618         return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1619 }
1620
1621 #endif
1622
1623 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1624 {
1625         struct rlimit new_rlim, *old_rlim;
1626         int retval;
1627
1628         if (resource >= RLIM_NLIMITS)
1629                 return -EINVAL;
1630         if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1631                 return -EFAULT;
1632        if (new_rlim.rlim_cur > new_rlim.rlim_max)
1633                return -EINVAL;
1634         old_rlim = current->signal->rlim + resource;
1635         if ((new_rlim.rlim_max > old_rlim->rlim_max) &&
1636             !capable(CAP_SYS_RESOURCE))
1637                 return -EPERM;
1638         if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN)
1639                         return -EPERM;
1640
1641         retval = security_task_setrlimit(resource, &new_rlim);
1642         if (retval)
1643                 return retval;
1644
1645         task_lock(current->group_leader);
1646         *old_rlim = new_rlim;
1647         task_unlock(current->group_leader);
1648
1649         if (resource == RLIMIT_CPU && new_rlim.rlim_cur != RLIM_INFINITY &&
1650             (cputime_eq(current->signal->it_prof_expires, cputime_zero) ||
1651              new_rlim.rlim_cur <= cputime_to_secs(
1652                      current->signal->it_prof_expires))) {
1653                 cputime_t cputime = secs_to_cputime(new_rlim.rlim_cur);
1654                 read_lock(&tasklist_lock);
1655                 spin_lock_irq(&current->sighand->siglock);
1656                 set_process_cpu_timer(current, CPUCLOCK_PROF,
1657                                       &cputime, NULL);
1658                 spin_unlock_irq(&current->sighand->siglock);
1659                 read_unlock(&tasklist_lock);
1660         }
1661
1662         return 0;
1663 }
1664
1665 /*
1666  * It would make sense to put struct rusage in the task_struct,
1667  * except that would make the task_struct be *really big*.  After
1668  * task_struct gets moved into malloc'ed memory, it would
1669  * make sense to do this.  It will make moving the rest of the information
1670  * a lot simpler!  (Which we're not doing right now because we're not
1671  * measuring them yet).
1672  *
1673  * This expects to be called with tasklist_lock read-locked or better,
1674  * and the siglock not locked.  It may momentarily take the siglock.
1675  *
1676  * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1677  * races with threads incrementing their own counters.  But since word
1678  * reads are atomic, we either get new values or old values and we don't
1679  * care which for the sums.  We always take the siglock to protect reading
1680  * the c* fields from p->signal from races with exit.c updating those
1681  * fields when reaping, so a sample either gets all the additions of a
1682  * given child after it's reaped, or none so this sample is before reaping.
1683  */
1684
1685 static void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1686 {
1687         struct task_struct *t;
1688         unsigned long flags;
1689         cputime_t utime, stime;
1690
1691         memset((char *) r, 0, sizeof *r);
1692
1693         if (unlikely(!p->signal))
1694                 return;
1695
1696         switch (who) {
1697                 case RUSAGE_CHILDREN:
1698                         spin_lock_irqsave(&p->sighand->siglock, flags);
1699                         utime = p->signal->cutime;
1700                         stime = p->signal->cstime;
1701                         r->ru_nvcsw = p->signal->cnvcsw;
1702                         r->ru_nivcsw = p->signal->cnivcsw;
1703                         r->ru_minflt = p->signal->cmin_flt;
1704                         r->ru_majflt = p->signal->cmaj_flt;
1705                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1706                         cputime_to_timeval(utime, &r->ru_utime);
1707                         cputime_to_timeval(stime, &r->ru_stime);
1708                         break;
1709                 case RUSAGE_SELF:
1710                         spin_lock_irqsave(&p->sighand->siglock, flags);
1711                         utime = stime = cputime_zero;
1712                         goto sum_group;
1713                 case RUSAGE_BOTH:
1714                         spin_lock_irqsave(&p->sighand->siglock, flags);
1715                         utime = p->signal->cutime;
1716                         stime = p->signal->cstime;
1717                         r->ru_nvcsw = p->signal->cnvcsw;
1718                         r->ru_nivcsw = p->signal->cnivcsw;
1719                         r->ru_minflt = p->signal->cmin_flt;
1720                         r->ru_majflt = p->signal->cmaj_flt;
1721                 sum_group:
1722                         utime = cputime_add(utime, p->signal->utime);
1723                         stime = cputime_add(stime, p->signal->stime);
1724                         r->ru_nvcsw += p->signal->nvcsw;
1725                         r->ru_nivcsw += p->signal->nivcsw;
1726                         r->ru_minflt += p->signal->min_flt;
1727                         r->ru_majflt += p->signal->maj_flt;
1728                         t = p;
1729                         do {
1730                                 utime = cputime_add(utime, t->utime);
1731                                 stime = cputime_add(stime, t->stime);
1732                                 r->ru_nvcsw += t->nvcsw;
1733                                 r->ru_nivcsw += t->nivcsw;
1734                                 r->ru_minflt += t->min_flt;
1735                                 r->ru_majflt += t->maj_flt;
1736                                 t = next_thread(t);
1737                         } while (t != p);
1738                         spin_unlock_irqrestore(&p->sighand->siglock, flags);
1739                         cputime_to_timeval(utime, &r->ru_utime);
1740                         cputime_to_timeval(stime, &r->ru_stime);
1741                         break;
1742                 default:
1743                         BUG();
1744         }
1745 }
1746
1747 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1748 {
1749         struct rusage r;
1750         read_lock(&tasklist_lock);
1751         k_getrusage(p, who, &r);
1752         read_unlock(&tasklist_lock);
1753         return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1754 }
1755
1756 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1757 {
1758         if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1759                 return -EINVAL;
1760         return getrusage(current, who, ru);
1761 }
1762
1763 asmlinkage long sys_umask(int mask)
1764 {
1765         mask = xchg(&current->fs->umask, mask & S_IRWXUGO);
1766         return mask;
1767 }
1768     
1769 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1770                           unsigned long arg4, unsigned long arg5)
1771 {
1772         long error;
1773
1774         error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1775         if (error)
1776                 return error;
1777
1778         switch (option) {
1779                 case PR_SET_PDEATHSIG:
1780                         if (!valid_signal(arg2)) {
1781                                 error = -EINVAL;
1782                                 break;
1783                         }
1784                         current->pdeath_signal = arg2;
1785                         break;
1786                 case PR_GET_PDEATHSIG:
1787                         error = put_user(current->pdeath_signal, (int __user *)arg2);
1788                         break;
1789                 case PR_GET_DUMPABLE:
1790                         error = current->mm->dumpable;
1791                         break;
1792                 case PR_SET_DUMPABLE:
1793                         if (arg2 < 0 || arg2 > 2) {
1794                                 error = -EINVAL;
1795                                 break;
1796                         }
1797                         current->mm->dumpable = arg2;
1798                         break;
1799
1800                 case PR_SET_UNALIGN:
1801                         error = SET_UNALIGN_CTL(current, arg2);
1802                         break;
1803                 case PR_GET_UNALIGN:
1804                         error = GET_UNALIGN_CTL(current, arg2);
1805                         break;
1806                 case PR_SET_FPEMU:
1807                         error = SET_FPEMU_CTL(current, arg2);
1808                         break;
1809                 case PR_GET_FPEMU:
1810                         error = GET_FPEMU_CTL(current, arg2);
1811                         break;
1812                 case PR_SET_FPEXC:
1813                         error = SET_FPEXC_CTL(current, arg2);
1814                         break;
1815                 case PR_GET_FPEXC:
1816                         error = GET_FPEXC_CTL(current, arg2);
1817                         break;
1818                 case PR_GET_TIMING:
1819                         error = PR_TIMING_STATISTICAL;
1820                         break;
1821                 case PR_SET_TIMING:
1822                         if (arg2 == PR_TIMING_STATISTICAL)
1823                                 error = 0;
1824                         else
1825                                 error = -EINVAL;
1826                         break;
1827
1828                 case PR_GET_KEEPCAPS:
1829                         if (current->keep_capabilities)
1830                                 error = 1;
1831                         break;
1832                 case PR_SET_KEEPCAPS:
1833                         if (arg2 != 0 && arg2 != 1) {
1834                                 error = -EINVAL;
1835                                 break;
1836                         }
1837                         current->keep_capabilities = arg2;
1838                         break;
1839                 case PR_SET_NAME: {
1840                         struct task_struct *me = current;
1841                         unsigned char ncomm[sizeof(me->comm)];
1842
1843                         ncomm[sizeof(me->comm)-1] = 0;
1844                         if (strncpy_from_user(ncomm, (char __user *)arg2,
1845                                                 sizeof(me->comm)-1) < 0)
1846                                 return -EFAULT;
1847                         set_task_comm(me, ncomm);
1848                         return 0;
1849                 }
1850                 case PR_GET_NAME: {
1851                         struct task_struct *me = current;
1852                         unsigned char tcomm[sizeof(me->comm)];
1853
1854                         get_task_comm(tcomm, me);
1855                         if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm)))
1856                                 return -EFAULT;
1857                         return 0;
1858                 }
1859                 default:
1860                         error = -EINVAL;
1861                         break;
1862         }
1863         return error;
1864 }