4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/config.h>
8 #include <linux/compat.h>
9 #include <linux/module.h>
11 #include <linux/utsname.h>
12 #include <linux/mman.h>
13 #include <linux/smp_lock.h>
14 #include <linux/notifier.h>
15 #include <linux/kmod.h>
16 #include <linux/reboot.h>
17 #include <linux/prctl.h>
18 #include <linux/init.h>
19 #include <linux/highuid.h>
21 #include <linux/kernel.h>
22 #include <linux/kexec.h>
23 #include <linux/workqueue.h>
24 #include <linux/device.h>
25 #include <linux/times.h>
26 #include <linux/security.h>
27 #include <linux/dcookies.h>
28 #include <linux/suspend.h>
29 #include <linux/ckrm.h>
30 #include <linux/vs_base.h>
31 #include <linux/vs_cvirt.h>
33 #include <asm/uaccess.h>
35 #include <asm/unistd.h>
37 #ifndef SET_UNALIGN_CTL
38 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
40 #ifndef GET_UNALIGN_CTL
41 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
44 # define SET_FPEMU_CTL(a,b) (-EINVAL)
47 # define GET_FPEMU_CTL(a,b) (-EINVAL)
50 # define SET_FPEXC_CTL(a,b) (-EINVAL)
53 # define GET_FPEXC_CTL(a,b) (-EINVAL)
57 * this is where the system-wide overflow UID and GID are defined, for
58 * architectures that now have 32-bit UID/GID but didn't in the past
61 int overflowuid = DEFAULT_OVERFLOWUID;
62 int overflowgid = DEFAULT_OVERFLOWGID;
65 EXPORT_SYMBOL(overflowuid);
66 EXPORT_SYMBOL(overflowgid);
70 * the same as above, but for filesystems which can only store a 16-bit
71 * UID and GID. as such, this is needed on all architectures
74 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
75 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
77 EXPORT_SYMBOL(fs_overflowuid);
78 EXPORT_SYMBOL(fs_overflowgid);
81 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
88 * Notifier list for kernel code which wants to be called
89 * at shutdown. This is used to stop any idling DMA operations
93 static struct notifier_block *reboot_notifier_list;
94 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
97 * notifier_chain_register - Add notifier to a notifier chain
98 * @list: Pointer to root list pointer
99 * @n: New entry in notifier chain
101 * Adds a notifier to a notifier chain.
103 * Currently always returns zero.
106 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
108 write_lock(¬ifier_lock);
111 if(n->priority > (*list)->priority)
113 list= &((*list)->next);
117 write_unlock(¬ifier_lock);
121 EXPORT_SYMBOL(notifier_chain_register);
124 * notifier_chain_unregister - Remove notifier from a notifier chain
125 * @nl: Pointer to root list pointer
126 * @n: New entry in notifier chain
128 * Removes a notifier from a notifier chain.
130 * Returns zero on success, or %-ENOENT on failure.
133 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
135 write_lock(¬ifier_lock);
141 write_unlock(¬ifier_lock);
146 write_unlock(¬ifier_lock);
150 EXPORT_SYMBOL(notifier_chain_unregister);
153 * notifier_call_chain - Call functions in a notifier chain
154 * @n: Pointer to root pointer of notifier chain
155 * @val: Value passed unmodified to notifier function
156 * @v: Pointer passed unmodified to notifier function
158 * Calls each function in a notifier chain in turn.
160 * If the return value of the notifier can be and'd
161 * with %NOTIFY_STOP_MASK, then notifier_call_chain
162 * will return immediately, with the return value of
163 * the notifier function which halted execution.
164 * Otherwise, the return value is the return value
165 * of the last notifier function called.
168 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
171 struct notifier_block *nb = *n;
175 ret=nb->notifier_call(nb,val,v);
176 if(ret&NOTIFY_STOP_MASK)
185 EXPORT_SYMBOL(notifier_call_chain);
188 * register_reboot_notifier - Register function to be called at reboot time
189 * @nb: Info about notifier function to be called
191 * Registers a function with the list of functions
192 * to be called at reboot time.
194 * Currently always returns zero, as notifier_chain_register
195 * always returns zero.
198 int register_reboot_notifier(struct notifier_block * nb)
200 return notifier_chain_register(&reboot_notifier_list, nb);
203 EXPORT_SYMBOL(register_reboot_notifier);
206 * unregister_reboot_notifier - Unregister previously registered reboot notifier
207 * @nb: Hook to be unregistered
209 * Unregisters a previously registered reboot
212 * Returns zero on success, or %-ENOENT on failure.
215 int unregister_reboot_notifier(struct notifier_block * nb)
217 return notifier_chain_unregister(&reboot_notifier_list, nb);
220 EXPORT_SYMBOL(unregister_reboot_notifier);
222 asmlinkage long sys_ni_syscall(void)
227 cond_syscall(sys_nfsservctl)
228 cond_syscall(sys_quotactl)
229 cond_syscall(sys_acct)
230 cond_syscall(sys_lookup_dcookie)
231 cond_syscall(sys_swapon)
232 cond_syscall(sys_swapoff)
233 cond_syscall(sys_kexec_load)
234 cond_syscall(sys_init_module)
235 cond_syscall(sys_delete_module)
236 cond_syscall(sys_socketpair)
237 cond_syscall(sys_bind)
238 cond_syscall(sys_listen)
239 cond_syscall(sys_accept)
240 cond_syscall(sys_connect)
241 cond_syscall(sys_getsockname)
242 cond_syscall(sys_getpeername)
243 cond_syscall(sys_sendto)
244 cond_syscall(sys_send)
245 cond_syscall(sys_recvfrom)
246 cond_syscall(sys_recv)
247 cond_syscall(sys_socket)
248 cond_syscall(sys_setsockopt)
249 cond_syscall(sys_getsockopt)
250 cond_syscall(sys_shutdown)
251 cond_syscall(sys_sendmsg)
252 cond_syscall(sys_recvmsg)
253 cond_syscall(sys_socketcall)
254 cond_syscall(sys_futex)
255 cond_syscall(compat_sys_futex)
256 cond_syscall(sys_epoll_create)
257 cond_syscall(sys_epoll_ctl)
258 cond_syscall(sys_epoll_wait)
259 cond_syscall(sys_semget)
260 cond_syscall(sys_semop)
261 cond_syscall(sys_semtimedop)
262 cond_syscall(sys_semctl)
263 cond_syscall(sys_msgget)
264 cond_syscall(sys_msgsnd)
265 cond_syscall(sys_msgrcv)
266 cond_syscall(sys_msgctl)
267 cond_syscall(sys_shmget)
268 cond_syscall(sys_shmdt)
269 cond_syscall(sys_shmctl)
270 cond_syscall(sys_mq_open)
271 cond_syscall(sys_mq_unlink)
272 cond_syscall(sys_mq_timedsend)
273 cond_syscall(sys_mq_timedreceive)
274 cond_syscall(sys_mq_notify)
275 cond_syscall(sys_mq_getsetattr)
276 cond_syscall(compat_sys_mq_open)
277 cond_syscall(compat_sys_mq_timedsend)
278 cond_syscall(compat_sys_mq_timedreceive)
279 cond_syscall(compat_sys_mq_notify)
280 cond_syscall(compat_sys_mq_getsetattr)
281 cond_syscall(sys_mbind)
282 cond_syscall(sys_get_mempolicy)
283 cond_syscall(sys_set_mempolicy)
284 cond_syscall(compat_mbind)
285 cond_syscall(compat_get_mempolicy)
286 cond_syscall(compat_set_mempolicy)
288 /* arch-specific weak syscall entries */
289 cond_syscall(sys_pciconfig_read)
290 cond_syscall(sys_pciconfig_write)
291 cond_syscall(sys_pciconfig_iobase)
293 static int set_one_prio(struct task_struct *p, int niceval, int error)
297 if (p->uid != current->euid &&
298 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
302 if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
303 if (vx_flags(VXF_IGNEG_NICE, 0))
309 no_nice = security_task_setnice(p, niceval);
316 set_user_nice(p, niceval);
321 asmlinkage long sys_setpriority(int which, int who, int niceval)
323 struct task_struct *g, *p;
324 struct user_struct *user;
327 if (which > 2 || which < 0)
330 /* normalize: avoid signed division (rounding problems) */
337 read_lock(&tasklist_lock);
342 p = find_task_by_pid(who);
344 error = set_one_prio(p, niceval, error);
348 who = process_group(current);
349 do_each_task_pid(who, PIDTYPE_PGID, p) {
350 error = set_one_prio(p, niceval, error);
351 } while_each_task_pid(who, PIDTYPE_PGID, p);
355 user = current->user;
357 user = find_user(vx_current_xid(), who);
364 error = set_one_prio(p, niceval, error);
365 while_each_thread(g, p);
367 free_uid(user); /* For find_user() */
371 read_unlock(&tasklist_lock);
377 * Ugh. To avoid negative return values, "getpriority()" will
378 * not return the normal nice-value, but a negated value that
379 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
380 * to stay compatible.
382 asmlinkage long sys_getpriority(int which, int who)
384 struct task_struct *g, *p;
385 struct user_struct *user;
386 long niceval, retval = -ESRCH;
388 if (which > 2 || which < 0)
391 read_lock(&tasklist_lock);
396 p = find_task_by_pid(who);
398 niceval = 20 - task_nice(p);
399 if (niceval > retval)
405 who = process_group(current);
406 do_each_task_pid(who, PIDTYPE_PGID, p) {
407 niceval = 20 - task_nice(p);
408 if (niceval > retval)
410 } while_each_task_pid(who, PIDTYPE_PGID, p);
414 user = current->user;
416 user = find_user(vx_current_xid(), who);
423 niceval = 20 - task_nice(p);
424 if (niceval > retval)
427 while_each_thread(g, p);
429 free_uid(user); /* for find_user() */
433 read_unlock(&tasklist_lock);
438 long vs_reboot(unsigned int, void *);
441 * Reboot system call: for obvious reasons only root may call it,
442 * and even root needs to set up some magic numbers in the registers
443 * so that some mistake won't make this reboot the whole machine.
444 * You can also set the meaning of the ctrl-alt-del-key here.
446 * reboot doesn't sync: do that yourself before calling this.
448 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
452 /* We only trust the superuser with rebooting the system. */
453 if (!capable(CAP_SYS_BOOT))
456 /* For safety, we require "magic" arguments. */
457 if (magic1 != LINUX_REBOOT_MAGIC1 ||
458 (magic2 != LINUX_REBOOT_MAGIC2 &&
459 magic2 != LINUX_REBOOT_MAGIC2A &&
460 magic2 != LINUX_REBOOT_MAGIC2B &&
461 magic2 != LINUX_REBOOT_MAGIC2C))
464 if (!vx_check(0, VX_ADMIN|VX_WATCH))
465 return vs_reboot(cmd, arg);
469 case LINUX_REBOOT_CMD_RESTART:
470 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
471 system_state = SYSTEM_RESTART;
473 printk(KERN_EMERG "Restarting system.\n");
474 machine_restart(NULL);
477 case LINUX_REBOOT_CMD_CAD_ON:
481 case LINUX_REBOOT_CMD_CAD_OFF:
485 case LINUX_REBOOT_CMD_HALT:
486 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
487 system_state = SYSTEM_HALT;
489 printk(KERN_EMERG "System halted.\n");
495 case LINUX_REBOOT_CMD_POWER_OFF:
496 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
497 system_state = SYSTEM_POWER_OFF;
499 printk(KERN_EMERG "Power down.\n");
505 case LINUX_REBOOT_CMD_RESTART2:
506 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
510 buffer[sizeof(buffer) - 1] = '\0';
512 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
513 system_state = SYSTEM_RESTART;
515 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
516 machine_restart(buffer);
520 case LINUX_REBOOT_CMD_KEXEC:
522 struct kimage *image;
523 image = xchg(&kexec_image, 0);
528 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
529 system_state = SYSTEM_RESTART;
531 system_state = SYSTEM_BOOTING;
532 printk(KERN_EMERG "Starting new kernel\n");
534 machine_kexec(image);
538 #ifdef CONFIG_SOFTWARE_SUSPEND
539 case LINUX_REBOOT_CMD_SW_SUSPEND:
541 int ret = software_suspend();
555 static void deferred_cad(void *dummy)
557 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
558 machine_restart(NULL);
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.
566 void ctrl_alt_del(void)
568 static DECLARE_WORK(cad_work, deferred_cad, NULL);
571 schedule_work(&cad_work);
573 kill_proc(cad_pid, SIGINT, 1);
578 * Unprivileged users may change the real gid to the effective gid
579 * or vice versa. (BSD-style)
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.
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.
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.
592 * SMP: There are not races, the GIDs are checked only by filesystem
593 * operations (as far as semantic preservation is concerned).
595 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
597 int old_rgid = current->gid;
598 int old_egid = current->egid;
599 int new_rgid = old_rgid;
600 int new_egid = old_egid;
603 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
607 if (rgid != (gid_t) -1) {
608 if ((old_rgid == rgid) ||
609 (current->egid==rgid) ||
615 if (egid != (gid_t) -1) {
616 if ((old_rgid == egid) ||
617 (current->egid == egid) ||
618 (current->sgid == egid) ||
625 if (new_egid != old_egid)
627 current->mm->dumpable = suid_dumpable;
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;
643 * setgid() is implemented like SysV w/ SAVED_IDS
645 * SMP: Same implicit races as above.
647 asmlinkage long sys_setgid(gid_t gid)
649 int old_egid = current->egid;
652 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
656 if (capable(CAP_SETGID))
660 current->mm->dumpable = suid_dumpable;
663 current->gid = current->egid = current->sgid = current->fsgid = gid;
665 else if ((gid == current->gid) || (gid == current->sgid))
669 current->mm->dumpable = suid_dumpable;
672 current->egid = current->fsgid = gid;
682 static int set_user(uid_t new_ruid, int dumpclear)
684 struct user_struct *new_user;
686 new_user = alloc_uid(vx_current_xid(), new_ruid);
690 if (atomic_read(&new_user->processes) >=
691 current->rlim[RLIMIT_NPROC].rlim_cur &&
692 new_user != &root_user) {
697 switch_uid(new_user);
701 current->mm->dumpable = suid_dumpable;
704 current->uid = new_ruid;
709 * Unprivileged users may change the real uid to the effective uid
710 * or vice versa. (BSD-style)
712 * If you set the real uid at all, or set the effective uid to a value not
713 * equal to the real uid, then the saved uid is set to the new effective uid.
715 * This makes it possible for a setuid program to completely drop its
716 * privileges, which is often a useful assertion to make when you are doing
717 * a security audit over a program.
719 * The general idea is that a program which uses just setreuid() will be
720 * 100% compatible with BSD. A program which uses just setuid() will be
721 * 100% compatible with POSIX with saved IDs.
723 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
725 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
728 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
732 new_ruid = old_ruid = current->uid;
733 new_euid = old_euid = current->euid;
734 old_suid = current->suid;
736 if (ruid != (uid_t) -1) {
738 if ((old_ruid != ruid) &&
739 (current->euid != ruid) &&
740 !capable(CAP_SETUID))
744 if (euid != (uid_t) -1) {
746 if ((old_ruid != euid) &&
747 (current->euid != euid) &&
748 (current->suid != euid) &&
749 !capable(CAP_SETUID))
753 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
756 if (new_euid != old_euid)
758 current->mm->dumpable = suid_dumpable;
761 current->fsuid = current->euid = new_euid;
762 if (ruid != (uid_t) -1 ||
763 (euid != (uid_t) -1 && euid != old_ruid))
764 current->suid = current->euid;
765 current->fsuid = current->euid;
769 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
775 * setuid() is implemented like SysV with SAVED_IDS
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.
785 asmlinkage long sys_setuid(uid_t uid)
787 int old_euid = current->euid;
788 int old_ruid, old_suid, new_ruid, new_suid;
791 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
795 old_ruid = new_ruid = current->uid;
796 old_suid = current->suid;
799 if (capable(CAP_SETUID)) {
800 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
803 } else if ((uid != current->uid) && (uid != new_suid))
808 current->mm->dumpable = suid_dumpable;
811 current->fsuid = current->euid = uid;
812 current->suid = new_suid;
816 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
821 * This function implements a generic ability to update ruid, euid,
822 * and suid. This allows you to implement the 4.4 compatible seteuid().
824 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
826 int old_ruid = current->uid;
827 int old_euid = current->euid;
828 int old_suid = current->suid;
831 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
835 if (!capable(CAP_SETUID)) {
836 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
837 (ruid != current->euid) && (ruid != current->suid))
839 if ((euid != (uid_t) -1) && (euid != current->uid) &&
840 (euid != current->euid) && (euid != current->suid))
842 if ((suid != (uid_t) -1) && (suid != current->uid) &&
843 (suid != current->euid) && (suid != current->suid))
846 if (ruid != (uid_t) -1) {
847 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
850 if (euid != (uid_t) -1) {
851 if (euid != current->euid)
853 current->mm->dumpable = suid_dumpable;
856 current->euid = euid;
858 current->fsuid = current->euid;
859 if (suid != (uid_t) -1)
860 current->suid = suid;
864 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
867 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
871 if (!(retval = put_user(current->uid, ruid)) &&
872 !(retval = put_user(current->euid, euid)))
873 retval = put_user(current->suid, suid);
879 * Same as above, but for rgid, egid, sgid.
881 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
885 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
889 if (!capable(CAP_SETGID)) {
890 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
891 (rgid != current->egid) && (rgid != current->sgid))
893 if ((egid != (gid_t) -1) && (egid != current->gid) &&
894 (egid != current->egid) && (egid != current->sgid))
896 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
897 (sgid != current->egid) && (sgid != current->sgid))
900 if (egid != (gid_t) -1) {
901 if (egid != current->egid)
903 current->mm->dumpable = suid_dumpable;
906 current->egid = egid;
908 current->fsgid = current->egid;
909 if (rgid != (gid_t) -1)
911 if (sgid != (gid_t) -1)
912 current->sgid = sgid;
919 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
923 if (!(retval = put_user(current->gid, rgid)) &&
924 !(retval = put_user(current->egid, egid)))
925 retval = put_user(current->sgid, sgid);
932 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
933 * is used for "access()" and for the NFS daemon (letting nfsd stay at
934 * whatever uid it wants to). It normally shadows "euid", except when
935 * explicitly set by setfsuid() or for access..
937 asmlinkage long sys_setfsuid(uid_t uid)
941 old_fsuid = current->fsuid;
942 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
945 if (uid == current->uid || uid == current->euid ||
946 uid == current->suid || uid == current->fsuid ||
949 if (uid != old_fsuid)
951 current->mm->dumpable = suid_dumpable;
954 current->fsuid = uid;
957 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
963 * Samma på svenska..
965 asmlinkage long sys_setfsgid(gid_t gid)
969 old_fsgid = current->fsgid;
970 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
973 if (gid == current->gid || gid == current->egid ||
974 gid == current->sgid || gid == current->fsgid ||
977 if (gid != old_fsgid)
979 current->mm->dumpable = suid_dumpable;
982 current->fsgid = gid;
987 asmlinkage long sys_times(struct tms __user * tbuf)
990 * In the SMP world we might just be unlucky and have one of
991 * the times increment as we use it. Since the value is an
992 * atomically safe type this is just fine. Conceptually its
993 * as if the syscall took an instant longer to occur.
997 struct task_struct *tsk = current;
998 struct task_struct *t;
999 unsigned long utime, stime, cutime, cstime;
1001 read_lock(&tasklist_lock);
1002 utime = tsk->signal->utime;
1003 stime = tsk->signal->stime;
1012 * While we have tasklist_lock read-locked, no dying thread
1013 * can be updating current->signal->[us]time. Instead,
1014 * we got their counts included in the live thread loop.
1015 * However, another thread can come in right now and
1016 * do a wait call that updates current->signal->c[us]time.
1017 * To make sure we always see that pair updated atomically,
1018 * we take the siglock around fetching them.
1020 spin_lock_irq(&tsk->sighand->siglock);
1021 cutime = tsk->signal->cutime;
1022 cstime = tsk->signal->cstime;
1023 spin_unlock_irq(&tsk->sighand->siglock);
1024 read_unlock(&tasklist_lock);
1026 tmp.tms_utime = jiffies_to_clock_t(utime);
1027 tmp.tms_stime = jiffies_to_clock_t(stime);
1028 tmp.tms_cutime = jiffies_to_clock_t(cutime);
1029 tmp.tms_cstime = jiffies_to_clock_t(cstime);
1030 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1033 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1037 * This needs some heavy checking ...
1038 * I just haven't the stomach for it. I also don't fully
1039 * understand sessions/pgrp etc. Let somebody who does explain it.
1041 * OK, I think I have the protection semantics right.... this is really
1042 * only important on a multi-user system anyway, to make sure one user
1043 * can't send a signal to a process owned by another. -TYT, 12/12/91
1045 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1049 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1051 struct task_struct *p;
1056 pid = vx_map_pid(current->pid);
1062 rpgid = vx_rmap_pid(pgid);
1064 /* From this point forward we keep holding onto the tasklist lock
1065 * so that our parent does not change from under us. -DaveM
1067 write_lock_irq(&tasklist_lock);
1070 p = find_task_by_pid(pid);
1075 if (!thread_group_leader(p))
1078 if (p->parent == current || p->real_parent == current) {
1080 if (p->signal->session != current->signal->session)
1092 if (p->signal->leader)
1096 struct task_struct *p;
1098 do_each_task_pid(rpgid, PIDTYPE_PGID, p) {
1099 if (p->signal->session == current->signal->session)
1101 } while_each_task_pid(rpgid, PIDTYPE_PGID, p);
1106 err = security_task_setpgid(p, rpgid);
1110 if (process_group(p) != rpgid) {
1111 detach_pid(p, PIDTYPE_PGID);
1112 p->signal->pgrp = rpgid;
1113 attach_pid(p, PIDTYPE_PGID, rpgid);
1118 /* All paths lead to here, thus we are safe. -DaveM */
1119 write_unlock_irq(&tasklist_lock);
1123 asmlinkage long sys_getpgid(pid_t pid)
1126 return vx_rmap_pid(process_group(current));
1129 struct task_struct *p;
1131 read_lock(&tasklist_lock);
1132 p = find_task_by_pid(pid);
1136 retval = security_task_getpgid(p);
1138 retval = vx_rmap_pid(process_group(p));
1140 read_unlock(&tasklist_lock);
1145 #ifdef __ARCH_WANT_SYS_GETPGRP
1147 asmlinkage long sys_getpgrp(void)
1149 /* SMP - assuming writes are word atomic this is fine */
1150 return process_group(current);
1155 asmlinkage long sys_getsid(pid_t pid)
1158 return current->signal->session;
1161 struct task_struct *p;
1163 read_lock(&tasklist_lock);
1164 p = find_task_by_pid(pid);
1168 retval = security_task_getsid(p);
1170 retval = p->signal->session;
1172 read_unlock(&tasklist_lock);
1177 asmlinkage long sys_setsid(void)
1182 if (!thread_group_leader(current))
1185 write_lock_irq(&tasklist_lock);
1187 pid = find_pid(PIDTYPE_PGID, current->pid);
1191 current->signal->leader = 1;
1192 __set_special_pids(current->pid, current->pid);
1193 current->signal->tty = NULL;
1194 current->signal->tty_old_pgrp = 0;
1195 err = process_group(current);
1197 write_unlock_irq(&tasklist_lock);
1202 * Supplementary group IDs
1205 /* init to 2 - one for init_task, one to ensure it is never freed */
1206 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1208 struct group_info *groups_alloc(int gidsetsize)
1210 struct group_info *group_info;
1214 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1215 /* Make sure we always allocate at least one indirect block pointer */
1216 nblocks = nblocks ? : 1;
1217 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1220 group_info->ngroups = gidsetsize;
1221 group_info->nblocks = nblocks;
1222 atomic_set(&group_info->usage, 1);
1224 if (gidsetsize <= NGROUPS_SMALL) {
1225 group_info->blocks[0] = group_info->small_block;
1227 for (i = 0; i < nblocks; i++) {
1229 b = (void *)__get_free_page(GFP_USER);
1231 goto out_undo_partial_alloc;
1232 group_info->blocks[i] = b;
1237 out_undo_partial_alloc:
1239 free_page((unsigned long)group_info->blocks[i]);
1245 EXPORT_SYMBOL(groups_alloc);
1247 void groups_free(struct group_info *group_info)
1249 if (group_info->blocks[0] != group_info->small_block) {
1251 for (i = 0; i < group_info->nblocks; i++)
1252 free_page((unsigned long)group_info->blocks[i]);
1257 EXPORT_SYMBOL(groups_free);
1259 /* export the group_info to a user-space array */
1260 static int groups_to_user(gid_t __user *grouplist,
1261 struct group_info *group_info)
1264 int count = group_info->ngroups;
1266 for (i = 0; i < group_info->nblocks; i++) {
1267 int cp_count = min(NGROUPS_PER_BLOCK, count);
1268 int off = i * NGROUPS_PER_BLOCK;
1269 int len = cp_count * sizeof(*grouplist);
1271 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1279 /* fill a group_info from a user-space array - it must be allocated already */
1280 static int groups_from_user(struct group_info *group_info,
1281 gid_t __user *grouplist)
1284 int count = group_info->ngroups;
1286 for (i = 0; i < group_info->nblocks; i++) {
1287 int cp_count = min(NGROUPS_PER_BLOCK, count);
1288 int off = i * NGROUPS_PER_BLOCK;
1289 int len = cp_count * sizeof(*grouplist);
1291 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1299 /* a simple shell-metzner sort */
1300 static void groups_sort(struct group_info *group_info)
1302 int base, max, stride;
1303 int gidsetsize = group_info->ngroups;
1305 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1310 max = gidsetsize - stride;
1311 for (base = 0; base < max; base++) {
1313 int right = left + stride;
1314 gid_t tmp = GROUP_AT(group_info, right);
1316 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1317 GROUP_AT(group_info, right) =
1318 GROUP_AT(group_info, left);
1322 GROUP_AT(group_info, right) = tmp;
1328 /* a simple bsearch */
1329 static int groups_search(struct group_info *group_info, gid_t grp)
1337 right = group_info->ngroups;
1338 while (left < right) {
1339 int mid = (left+right)/2;
1340 int cmp = grp - GROUP_AT(group_info, mid);
1351 /* validate and set current->group_info */
1352 int set_current_groups(struct group_info *group_info)
1355 struct group_info *old_info;
1357 retval = security_task_setgroups(group_info);
1361 groups_sort(group_info);
1362 get_group_info(group_info);
1365 old_info = current->group_info;
1366 current->group_info = group_info;
1367 task_unlock(current);
1369 put_group_info(old_info);
1374 EXPORT_SYMBOL(set_current_groups);
1376 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1381 * SMP: Nobody else can change our grouplist. Thus we are
1388 /* no need to grab task_lock here; it cannot change */
1389 get_group_info(current->group_info);
1390 i = current->group_info->ngroups;
1392 if (i > gidsetsize) {
1396 if (groups_to_user(grouplist, current->group_info)) {
1402 put_group_info(current->group_info);
1407 * SMP: Our groups are copy-on-write. We can set them safely
1408 * without another task interfering.
1411 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1413 struct group_info *group_info;
1416 if (!capable(CAP_SETGID))
1418 if ((unsigned)gidsetsize > NGROUPS_MAX)
1421 group_info = groups_alloc(gidsetsize);
1424 retval = groups_from_user(group_info, grouplist);
1426 put_group_info(group_info);
1430 retval = set_current_groups(group_info);
1431 put_group_info(group_info);
1437 * Check whether we're fsgid/egid or in the supplemental group..
1439 int in_group_p(gid_t grp)
1442 if (grp != current->fsgid) {
1443 get_group_info(current->group_info);
1444 retval = groups_search(current->group_info, grp);
1445 put_group_info(current->group_info);
1450 EXPORT_SYMBOL(in_group_p);
1452 int in_egroup_p(gid_t grp)
1455 if (grp != current->egid) {
1456 get_group_info(current->group_info);
1457 retval = groups_search(current->group_info, grp);
1458 put_group_info(current->group_info);
1463 EXPORT_SYMBOL(in_egroup_p);
1465 DECLARE_RWSEM(uts_sem);
1467 EXPORT_SYMBOL(uts_sem);
1469 asmlinkage long sys_newuname(struct new_utsname __user * name)
1473 down_read(&uts_sem);
1474 if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1480 asmlinkage long sys_sethostname(char __user *name, int len)
1483 char tmp[__NEW_UTS_LEN];
1485 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1487 if (len < 0 || len > __NEW_UTS_LEN)
1489 down_write(&uts_sem);
1491 if (!copy_from_user(tmp, name, len)) {
1492 char *ptr = vx_new_uts(nodename);
1494 memcpy(ptr, tmp, len);
1502 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1504 asmlinkage long sys_gethostname(char __user *name, int len)
1511 down_read(&uts_sem);
1512 ptr = vx_new_uts(nodename);
1513 i = 1 + strlen(ptr);
1517 if (copy_to_user(name, ptr, i))
1526 * Only setdomainname; getdomainname can be implemented by calling
1529 asmlinkage long sys_setdomainname(char __user *name, int len)
1532 char tmp[__NEW_UTS_LEN];
1534 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1536 if (len < 0 || len > __NEW_UTS_LEN)
1539 down_write(&uts_sem);
1541 if (!copy_from_user(tmp, name, len)) {
1542 char *ptr = vx_new_uts(domainname);
1544 memcpy(ptr, tmp, len);
1552 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1554 if (resource >= RLIM_NLIMITS)
1557 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1561 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1564 * Back compatibility for getrlimit. Needed for some apps.
1567 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1570 if (resource >= RLIM_NLIMITS)
1573 memcpy(&x, current->rlim + resource, sizeof(*rlim));
1574 if(x.rlim_cur > 0x7FFFFFFF)
1575 x.rlim_cur = 0x7FFFFFFF;
1576 if(x.rlim_max > 0x7FFFFFFF)
1577 x.rlim_max = 0x7FFFFFFF;
1578 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1583 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1585 struct rlimit new_rlim, *old_rlim;
1588 if (resource >= RLIM_NLIMITS)
1590 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1592 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1594 old_rlim = current->rlim + resource;
1595 if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1596 (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1597 !capable(CAP_SYS_RESOURCE) && !vx_ccaps(VXC_SET_RLIMIT))
1599 if (resource == RLIMIT_NOFILE) {
1600 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1604 retval = security_task_setrlimit(resource, &new_rlim);
1608 *old_rlim = new_rlim;
1613 * It would make sense to put struct rusage in the task_struct,
1614 * except that would make the task_struct be *really big*. After
1615 * task_struct gets moved into malloc'ed memory, it would
1616 * make sense to do this. It will make moving the rest of the information
1617 * a lot simpler! (Which we're not doing right now because we're not
1618 * measuring them yet).
1620 * This expects to be called with tasklist_lock read-locked or better,
1621 * and the siglock not locked. It may momentarily take the siglock.
1623 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1624 * races with threads incrementing their own counters. But since word
1625 * reads are atomic, we either get new values or old values and we don't
1626 * care which for the sums. We always take the siglock to protect reading
1627 * the c* fields from p->signal from races with exit.c updating those
1628 * fields when reaping, so a sample either gets all the additions of a
1629 * given child after it's reaped, or none so this sample is before reaping.
1632 void k_getrusage(struct task_struct *p, int who, struct rusage *r)
1634 struct task_struct *t;
1635 unsigned long flags;
1636 unsigned long utime, stime;
1638 memset((char *) r, 0, sizeof *r);
1640 if (unlikely(!p->signal))
1644 case RUSAGE_CHILDREN:
1645 spin_lock_irqsave(&p->sighand->siglock, flags);
1646 utime = p->signal->cutime;
1647 stime = p->signal->cstime;
1648 r->ru_nvcsw = p->signal->cnvcsw;
1649 r->ru_nivcsw = p->signal->cnivcsw;
1650 r->ru_minflt = p->signal->cmin_flt;
1651 r->ru_majflt = p->signal->cmaj_flt;
1652 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1653 jiffies_to_timeval(utime, &r->ru_utime);
1654 jiffies_to_timeval(stime, &r->ru_stime);
1657 spin_lock_irqsave(&p->sighand->siglock, flags);
1661 spin_lock_irqsave(&p->sighand->siglock, flags);
1662 utime = p->signal->cutime;
1663 stime = p->signal->cstime;
1664 r->ru_nvcsw = p->signal->cnvcsw;
1665 r->ru_nivcsw = p->signal->cnivcsw;
1666 r->ru_minflt = p->signal->cmin_flt;
1667 r->ru_majflt = p->signal->cmaj_flt;
1669 utime += p->signal->utime;
1670 stime += p->signal->stime;
1671 r->ru_nvcsw += p->signal->nvcsw;
1672 r->ru_nivcsw += p->signal->nivcsw;
1673 r->ru_minflt += p->signal->min_flt;
1674 r->ru_majflt += p->signal->maj_flt;
1679 r->ru_nvcsw += t->nvcsw;
1680 r->ru_nivcsw += t->nivcsw;
1681 r->ru_minflt += t->min_flt;
1682 r->ru_majflt += t->maj_flt;
1685 spin_unlock_irqrestore(&p->sighand->siglock, flags);
1686 jiffies_to_timeval(utime, &r->ru_utime);
1687 jiffies_to_timeval(stime, &r->ru_stime);
1694 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1697 read_lock(&tasklist_lock);
1698 k_getrusage(p, who, &r);
1699 read_unlock(&tasklist_lock);
1700 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1703 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1705 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1707 return getrusage(current, who, ru);
1710 asmlinkage long sys_umask(int mask)
1712 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1716 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1717 unsigned long arg4, unsigned long arg5)
1722 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1727 case PR_SET_PDEATHSIG:
1729 if (sig < 0 || sig > _NSIG) {
1733 current->pdeath_signal = sig;
1735 case PR_GET_PDEATHSIG:
1736 error = put_user(current->pdeath_signal, (int __user *)arg2);
1738 case PR_GET_DUMPABLE:
1739 if (current->mm->dumpable)
1742 case PR_SET_DUMPABLE:
1743 if (arg2 < 0 || arg2 > 2) {
1747 current->mm->dumpable = arg2;
1750 case PR_SET_UNALIGN:
1751 error = SET_UNALIGN_CTL(current, arg2);
1753 case PR_GET_UNALIGN:
1754 error = GET_UNALIGN_CTL(current, arg2);
1757 error = SET_FPEMU_CTL(current, arg2);
1760 error = GET_FPEMU_CTL(current, arg2);
1763 error = SET_FPEXC_CTL(current, arg2);
1766 error = GET_FPEXC_CTL(current, arg2);
1769 error = PR_TIMING_STATISTICAL;
1772 if (arg2 == PR_TIMING_STATISTICAL)
1778 case PR_GET_KEEPCAPS:
1779 if (current->keep_capabilities)
1782 case PR_SET_KEEPCAPS:
1783 if (arg2 != 0 && arg2 != 1) {
1787 current->keep_capabilities = arg2;
1790 struct task_struct *me = current;
1791 unsigned char ncomm[sizeof(me->comm)];
1793 ncomm[sizeof(me->comm)-1] = 0;
1794 if (strncpy_from_user(ncomm, (char __user *)arg2,
1795 sizeof(me->comm)-1) < 0)
1797 set_task_comm(me, ncomm);