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/workqueue.h>
22 #include <linux/device.h>
23 #include <linux/times.h>
24 #include <linux/security.h>
25 #include <linux/dcookies.h>
26 #include <linux/suspend.h>
28 #include <asm/uaccess.h>
30 #include <asm/unistd.h>
32 #ifndef SET_UNALIGN_CTL
33 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
35 #ifndef GET_UNALIGN_CTL
36 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
39 # define SET_FPEMU_CTL(a,b) (-EINVAL)
42 # define GET_FPEMU_CTL(a,b) (-EINVAL)
45 # define SET_FPEXC_CTL(a,b) (-EINVAL)
48 # define GET_FPEXC_CTL(a,b) (-EINVAL)
52 * this is where the system-wide overflow UID and GID are defined, for
53 * architectures that now have 32-bit UID/GID but didn't in the past
56 int overflowuid = DEFAULT_OVERFLOWUID;
57 int overflowgid = DEFAULT_OVERFLOWGID;
60 EXPORT_SYMBOL(overflowuid);
61 EXPORT_SYMBOL(overflowgid);
65 * the same as above, but for filesystems which can only store a 16-bit
66 * UID and GID. as such, this is needed on all architectures
69 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
70 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
72 EXPORT_SYMBOL(fs_overflowuid);
73 EXPORT_SYMBOL(fs_overflowgid);
76 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
83 * Notifier list for kernel code which wants to be called
84 * at shutdown. This is used to stop any idling DMA operations
88 static struct notifier_block *reboot_notifier_list;
89 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
92 * notifier_chain_register - Add notifier to a notifier chain
93 * @list: Pointer to root list pointer
94 * @n: New entry in notifier chain
96 * Adds a notifier to a notifier chain.
98 * Currently always returns zero.
101 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
103 write_lock(¬ifier_lock);
106 if(n->priority > (*list)->priority)
108 list= &((*list)->next);
112 write_unlock(¬ifier_lock);
116 EXPORT_SYMBOL(notifier_chain_register);
119 * notifier_chain_unregister - Remove notifier from a notifier chain
120 * @nl: Pointer to root list pointer
121 * @n: New entry in notifier chain
123 * Removes a notifier from a notifier chain.
125 * Returns zero on success, or %-ENOENT on failure.
128 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
130 write_lock(¬ifier_lock);
136 write_unlock(¬ifier_lock);
141 write_unlock(¬ifier_lock);
145 EXPORT_SYMBOL(notifier_chain_unregister);
148 * notifier_call_chain - Call functions in a notifier chain
149 * @n: Pointer to root pointer of notifier chain
150 * @val: Value passed unmodified to notifier function
151 * @v: Pointer passed unmodified to notifier function
153 * Calls each function in a notifier chain in turn.
155 * If the return value of the notifier can be and'd
156 * with %NOTIFY_STOP_MASK, then notifier_call_chain
157 * will return immediately, with the return value of
158 * the notifier function which halted execution.
159 * Otherwise, the return value is the return value
160 * of the last notifier function called.
163 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
166 struct notifier_block *nb = *n;
170 ret=nb->notifier_call(nb,val,v);
171 if(ret&NOTIFY_STOP_MASK)
180 EXPORT_SYMBOL(notifier_call_chain);
183 * register_reboot_notifier - Register function to be called at reboot time
184 * @nb: Info about notifier function to be called
186 * Registers a function with the list of functions
187 * to be called at reboot time.
189 * Currently always returns zero, as notifier_chain_register
190 * always returns zero.
193 int register_reboot_notifier(struct notifier_block * nb)
195 return notifier_chain_register(&reboot_notifier_list, nb);
198 EXPORT_SYMBOL(register_reboot_notifier);
201 * unregister_reboot_notifier - Unregister previously registered reboot notifier
202 * @nb: Hook to be unregistered
204 * Unregisters a previously registered reboot
207 * Returns zero on success, or %-ENOENT on failure.
210 int unregister_reboot_notifier(struct notifier_block * nb)
212 return notifier_chain_unregister(&reboot_notifier_list, nb);
215 EXPORT_SYMBOL(unregister_reboot_notifier);
217 asmlinkage long sys_ni_syscall(void)
222 cond_syscall(sys_nfsservctl)
223 cond_syscall(sys_quotactl)
224 cond_syscall(sys_acct)
225 cond_syscall(sys_lookup_dcookie)
226 cond_syscall(sys_swapon)
227 cond_syscall(sys_swapoff)
228 cond_syscall(sys_init_module)
229 cond_syscall(sys_delete_module)
230 cond_syscall(sys_socketpair)
231 cond_syscall(sys_bind)
232 cond_syscall(sys_listen)
233 cond_syscall(sys_accept)
234 cond_syscall(sys_connect)
235 cond_syscall(sys_getsockname)
236 cond_syscall(sys_getpeername)
237 cond_syscall(sys_sendto)
238 cond_syscall(sys_send)
239 cond_syscall(sys_recvfrom)
240 cond_syscall(sys_recv)
241 cond_syscall(sys_socket)
242 cond_syscall(sys_setsockopt)
243 cond_syscall(sys_getsockopt)
244 cond_syscall(sys_shutdown)
245 cond_syscall(sys_sendmsg)
246 cond_syscall(sys_recvmsg)
247 cond_syscall(sys_socketcall)
248 cond_syscall(sys_futex)
249 cond_syscall(compat_sys_futex)
250 cond_syscall(sys_epoll_create)
251 cond_syscall(sys_epoll_ctl)
252 cond_syscall(sys_epoll_wait)
253 cond_syscall(sys_semget)
254 cond_syscall(sys_semop)
255 cond_syscall(sys_semtimedop)
256 cond_syscall(sys_semctl)
257 cond_syscall(sys_msgget)
258 cond_syscall(sys_msgsnd)
259 cond_syscall(sys_msgrcv)
260 cond_syscall(sys_msgctl)
261 cond_syscall(sys_shmget)
262 cond_syscall(sys_shmdt)
263 cond_syscall(sys_shmctl)
264 cond_syscall(sys_mq_open)
265 cond_syscall(sys_mq_unlink)
266 cond_syscall(sys_mq_timedsend)
267 cond_syscall(sys_mq_timedreceive)
268 cond_syscall(sys_mq_notify)
269 cond_syscall(sys_mq_getsetattr)
270 cond_syscall(compat_sys_mq_open)
271 cond_syscall(compat_sys_mq_timedsend)
272 cond_syscall(compat_sys_mq_timedreceive)
273 cond_syscall(compat_sys_mq_notify)
274 cond_syscall(compat_sys_mq_getsetattr)
275 cond_syscall(sys_mbind)
276 cond_syscall(sys_get_mempolicy)
277 cond_syscall(sys_set_mempolicy)
279 /* arch-specific weak syscall entries */
280 cond_syscall(sys_pciconfig_read)
281 cond_syscall(sys_pciconfig_write)
282 cond_syscall(sys_pciconfig_iobase)
284 static int set_one_prio(struct task_struct *p, int niceval, int error)
288 if (p->uid != current->euid &&
289 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
293 if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
297 no_nice = security_task_setnice(p, niceval);
304 set_user_nice(p, niceval);
309 asmlinkage long sys_setpriority(int which, int who, int niceval)
311 struct task_struct *g, *p;
312 struct user_struct *user;
317 if (which > 2 || which < 0)
320 /* normalize: avoid signed division (rounding problems) */
327 read_lock(&tasklist_lock);
332 p = find_task_by_pid(who);
334 error = set_one_prio(p, niceval, error);
338 who = process_group(current);
339 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid)
340 error = set_one_prio(p, niceval, error);
344 user = current->user;
346 user = find_user(vx_current_xid(), who);
353 error = set_one_prio(p, niceval, error);
354 while_each_thread(g, p);
356 free_uid(user); /* For find_user() */
360 read_unlock(&tasklist_lock);
366 * Ugh. To avoid negative return values, "getpriority()" will
367 * not return the normal nice-value, but a negated value that
368 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
369 * to stay compatible.
371 asmlinkage long sys_getpriority(int which, int who)
373 struct task_struct *g, *p;
376 struct user_struct *user;
377 long niceval, retval = -ESRCH;
379 if (which > 2 || which < 0)
382 read_lock(&tasklist_lock);
387 p = find_task_by_pid(who);
389 niceval = 20 - task_nice(p);
390 if (niceval > retval)
396 who = process_group(current);
397 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid) {
398 niceval = 20 - task_nice(p);
399 if (niceval > retval)
405 user = current->user;
407 user = find_user(vx_current_xid(), who);
414 niceval = 20 - task_nice(p);
415 if (niceval > retval)
418 while_each_thread(g, p);
420 free_uid(user); /* for find_user() */
424 read_unlock(&tasklist_lock);
430 * vshelper path is set via /proc/sys
431 * invoked by vserver sys_reboot(), with
432 * the following arguments
434 * argv [0] = vshelper_path;
435 * argv [1] = action: "restart", "halt", "poweroff", ...
436 * argv [2] = context identifier
437 * argv [3] = additional argument (restart2)
439 * envp [*] = type-specific parameters
441 char vshelper_path[255] = "/sbin/vshelper";
443 long vs_reboot(unsigned int cmd, void * arg)
445 char id_buf[8], cmd_buf[32];
446 char uid_buf[32], pid_buf[32];
449 char *argv[] = {vshelper_path, NULL, id_buf, NULL, 0};
450 char *envp[] = {"HOME=/", "TERM=linux",
451 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
452 uid_buf, pid_buf, cmd_buf, 0};
454 snprintf(id_buf, sizeof(id_buf)-1, "%d", vx_current_xid());
456 snprintf(cmd_buf, sizeof(cmd_buf)-1, "VS_CMD=%08x", cmd);
457 snprintf(uid_buf, sizeof(uid_buf)-1, "VS_UID=%d", current->uid);
458 snprintf(pid_buf, sizeof(pid_buf)-1, "VS_PID=%d", current->pid);
461 case LINUX_REBOOT_CMD_RESTART:
465 case LINUX_REBOOT_CMD_HALT:
469 case LINUX_REBOOT_CMD_POWER_OFF:
470 argv[1] = "poweroff";
473 case LINUX_REBOOT_CMD_SW_SUSPEND:
477 case LINUX_REBOOT_CMD_RESTART2:
478 if (strncpy_from_user(&buffer[0], (char *)arg, sizeof(buffer) - 1) < 0)
482 argv[1] = "restart2";
486 /* maybe we should wait ? */
487 if (call_usermodehelper(*argv, argv, envp, 0)) {
489 "vs_reboot(): failed to exec (%s %s %s %s)\n",
490 vshelper_path, argv[1], argv[2], argv[3]);
497 * Reboot system call: for obvious reasons only root may call it,
498 * and even root needs to set up some magic numbers in the registers
499 * so that some mistake won't make this reboot the whole machine.
500 * You can also set the meaning of the ctrl-alt-del-key here.
502 * reboot doesn't sync: do that yourself before calling this.
504 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
508 /* We only trust the superuser with rebooting the system. */
509 if (!capable(CAP_SYS_BOOT))
512 /* For safety, we require "magic" arguments. */
513 if (magic1 != LINUX_REBOOT_MAGIC1 ||
514 (magic2 != LINUX_REBOOT_MAGIC2 &&
515 magic2 != LINUX_REBOOT_MAGIC2A &&
516 magic2 != LINUX_REBOOT_MAGIC2B &&
517 magic2 != LINUX_REBOOT_MAGIC2C))
520 if (!vx_check(0, VX_ADMIN|VX_WATCH))
521 return vs_reboot(cmd, arg);
525 case LINUX_REBOOT_CMD_RESTART:
526 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
527 system_state = SYSTEM_RESTART;
529 printk(KERN_EMERG "Restarting system.\n");
530 machine_restart(NULL);
533 case LINUX_REBOOT_CMD_CAD_ON:
537 case LINUX_REBOOT_CMD_CAD_OFF:
541 case LINUX_REBOOT_CMD_HALT:
542 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
543 system_state = SYSTEM_HALT;
545 printk(KERN_EMERG "System halted.\n");
551 case LINUX_REBOOT_CMD_POWER_OFF:
552 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
553 system_state = SYSTEM_POWER_OFF;
555 printk(KERN_EMERG "Power down.\n");
561 case LINUX_REBOOT_CMD_RESTART2:
562 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
566 buffer[sizeof(buffer) - 1] = '\0';
568 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
569 system_state = SYSTEM_RESTART;
571 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
572 machine_restart(buffer);
575 #ifdef CONFIG_SOFTWARE_SUSPEND
576 case LINUX_REBOOT_CMD_SW_SUSPEND:
578 int ret = software_suspend();
592 static void deferred_cad(void *dummy)
594 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
595 machine_restart(NULL);
599 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
600 * As it's called within an interrupt, it may NOT sync: the only choice
601 * is whether to reboot at once, or just ignore the ctrl-alt-del.
603 void ctrl_alt_del(void)
605 static DECLARE_WORK(cad_work, deferred_cad, NULL);
608 schedule_work(&cad_work);
610 kill_proc(cad_pid, SIGINT, 1);
615 * Unprivileged users may change the real gid to the effective gid
616 * or vice versa. (BSD-style)
618 * If you set the real gid at all, or set the effective gid to a value not
619 * equal to the real gid, then the saved gid is set to the new effective gid.
621 * This makes it possible for a setgid program to completely drop its
622 * privileges, which is often a useful assertion to make when you are doing
623 * a security audit over a program.
625 * The general idea is that a program which uses just setregid() will be
626 * 100% compatible with BSD. A program which uses just setgid() will be
627 * 100% compatible with POSIX with saved IDs.
629 * SMP: There are not races, the GIDs are checked only by filesystem
630 * operations (as far as semantic preservation is concerned).
632 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
634 int old_rgid = current->gid;
635 int old_egid = current->egid;
636 int new_rgid = old_rgid;
637 int new_egid = old_egid;
640 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
644 if (rgid != (gid_t) -1) {
645 if ((old_rgid == rgid) ||
646 (current->egid==rgid) ||
652 if (egid != (gid_t) -1) {
653 if ((old_rgid == egid) ||
654 (current->egid == egid) ||
655 (current->sgid == egid) ||
662 if (new_egid != old_egid)
664 current->mm->dumpable = 0;
667 if (rgid != (gid_t) -1 ||
668 (egid != (gid_t) -1 && egid != old_rgid))
669 current->sgid = new_egid;
670 current->fsgid = new_egid;
671 current->egid = new_egid;
672 current->gid = new_rgid;
677 * setgid() is implemented like SysV w/ SAVED_IDS
679 * SMP: Same implicit races as above.
681 asmlinkage long sys_setgid(gid_t gid)
683 int old_egid = current->egid;
686 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
690 if (capable(CAP_SETGID))
694 current->mm->dumpable=0;
697 current->gid = current->egid = current->sgid = current->fsgid = gid;
699 else if ((gid == current->gid) || (gid == current->sgid))
703 current->mm->dumpable=0;
706 current->egid = current->fsgid = gid;
713 static int set_user(uid_t new_ruid, int dumpclear)
715 struct user_struct *new_user;
717 new_user = alloc_uid(vx_current_xid(), new_ruid);
721 if (atomic_read(&new_user->processes) >=
722 current->rlim[RLIMIT_NPROC].rlim_cur &&
723 new_user != &root_user) {
728 switch_uid(new_user);
732 current->mm->dumpable = 0;
735 current->uid = new_ruid;
740 * Unprivileged users may change the real uid to the effective uid
741 * or vice versa. (BSD-style)
743 * If you set the real uid at all, or set the effective uid to a value not
744 * equal to the real uid, then the saved uid is set to the new effective uid.
746 * This makes it possible for a setuid program to completely drop its
747 * privileges, which is often a useful assertion to make when you are doing
748 * a security audit over a program.
750 * The general idea is that a program which uses just setreuid() will be
751 * 100% compatible with BSD. A program which uses just setuid() will be
752 * 100% compatible with POSIX with saved IDs.
754 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
756 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
759 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
763 new_ruid = old_ruid = current->uid;
764 new_euid = old_euid = current->euid;
765 old_suid = current->suid;
767 if (ruid != (uid_t) -1) {
769 if ((old_ruid != ruid) &&
770 (current->euid != ruid) &&
771 !capable(CAP_SETUID))
775 if (euid != (uid_t) -1) {
777 if ((old_ruid != euid) &&
778 (current->euid != euid) &&
779 (current->suid != euid) &&
780 !capable(CAP_SETUID))
784 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
787 if (new_euid != old_euid)
789 current->mm->dumpable=0;
792 current->fsuid = current->euid = new_euid;
793 if (ruid != (uid_t) -1 ||
794 (euid != (uid_t) -1 && euid != old_ruid))
795 current->suid = current->euid;
796 current->fsuid = current->euid;
798 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
804 * setuid() is implemented like SysV with SAVED_IDS
806 * Note that SAVED_ID's is deficient in that a setuid root program
807 * like sendmail, for example, cannot set its uid to be a normal
808 * user and then switch back, because if you're root, setuid() sets
809 * the saved uid too. If you don't like this, blame the bright people
810 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
811 * will allow a root program to temporarily drop privileges and be able to
812 * regain them by swapping the real and effective uid.
814 asmlinkage long sys_setuid(uid_t uid)
816 int old_euid = current->euid;
817 int old_ruid, old_suid, new_ruid, new_suid;
820 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
824 old_ruid = new_ruid = current->uid;
825 old_suid = current->suid;
828 if (capable(CAP_SETUID)) {
829 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
832 } else if ((uid != current->uid) && (uid != new_suid))
837 current->mm->dumpable = 0;
840 current->fsuid = current->euid = uid;
841 current->suid = new_suid;
843 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
848 * This function implements a generic ability to update ruid, euid,
849 * and suid. This allows you to implement the 4.4 compatible seteuid().
851 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
853 int old_ruid = current->uid;
854 int old_euid = current->euid;
855 int old_suid = current->suid;
858 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
862 if (!capable(CAP_SETUID)) {
863 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
864 (ruid != current->euid) && (ruid != current->suid))
866 if ((euid != (uid_t) -1) && (euid != current->uid) &&
867 (euid != current->euid) && (euid != current->suid))
869 if ((suid != (uid_t) -1) && (suid != current->uid) &&
870 (suid != current->euid) && (suid != current->suid))
873 if (ruid != (uid_t) -1) {
874 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
877 if (euid != (uid_t) -1) {
878 if (euid != current->euid)
880 current->mm->dumpable = 0;
883 current->euid = euid;
885 current->fsuid = current->euid;
886 if (suid != (uid_t) -1)
887 current->suid = suid;
889 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
892 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
896 if (!(retval = put_user(current->uid, ruid)) &&
897 !(retval = put_user(current->euid, euid)))
898 retval = put_user(current->suid, suid);
904 * Same as above, but for rgid, egid, sgid.
906 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
910 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
914 if (!capable(CAP_SETGID)) {
915 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
916 (rgid != current->egid) && (rgid != current->sgid))
918 if ((egid != (gid_t) -1) && (egid != current->gid) &&
919 (egid != current->egid) && (egid != current->sgid))
921 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
922 (sgid != current->egid) && (sgid != current->sgid))
925 if (egid != (gid_t) -1) {
926 if (egid != current->egid)
928 current->mm->dumpable = 0;
931 current->egid = egid;
933 current->fsgid = current->egid;
934 if (rgid != (gid_t) -1)
936 if (sgid != (gid_t) -1)
937 current->sgid = sgid;
941 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
945 if (!(retval = put_user(current->gid, rgid)) &&
946 !(retval = put_user(current->egid, egid)))
947 retval = put_user(current->sgid, sgid);
954 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
955 * is used for "access()" and for the NFS daemon (letting nfsd stay at
956 * whatever uid it wants to). It normally shadows "euid", except when
957 * explicitly set by setfsuid() or for access..
959 asmlinkage long sys_setfsuid(uid_t uid)
963 old_fsuid = current->fsuid;
964 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
967 if (uid == current->uid || uid == current->euid ||
968 uid == current->suid || uid == current->fsuid ||
971 if (uid != old_fsuid)
973 current->mm->dumpable = 0;
976 current->fsuid = uid;
979 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
985 * Samma på svenska..
987 asmlinkage long sys_setfsgid(gid_t gid)
991 old_fsgid = current->fsgid;
992 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
995 if (gid == current->gid || gid == current->egid ||
996 gid == current->sgid || gid == current->fsgid ||
999 if (gid != old_fsgid)
1001 current->mm->dumpable = 0;
1004 current->fsgid = gid;
1009 asmlinkage long sys_times(struct tms __user * tbuf)
1012 * In the SMP world we might just be unlucky and have one of
1013 * the times increment as we use it. Since the value is an
1014 * atomically safe type this is just fine. Conceptually its
1015 * as if the syscall took an instant longer to occur.
1019 tmp.tms_utime = jiffies_to_clock_t(current->utime);
1020 tmp.tms_stime = jiffies_to_clock_t(current->stime);
1021 tmp.tms_cutime = jiffies_to_clock_t(current->cutime);
1022 tmp.tms_cstime = jiffies_to_clock_t(current->cstime);
1023 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1026 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1030 * This needs some heavy checking ...
1031 * I just haven't the stomach for it. I also don't fully
1032 * understand sessions/pgrp etc. Let somebody who does explain it.
1034 * OK, I think I have the protection semantics right.... this is really
1035 * only important on a multi-user system anyway, to make sure one user
1036 * can't send a signal to a process owned by another. -TYT, 12/12/91
1038 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1042 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1044 struct task_struct *p;
1054 /* From this point forward we keep holding onto the tasklist lock
1055 * so that our parent does not change from under us. -DaveM
1057 write_lock_irq(&tasklist_lock);
1060 p = find_task_by_pid(pid);
1065 if (!thread_group_leader(p))
1068 if (p->parent == current || p->real_parent == current) {
1070 if (p->signal->session != current->signal->session)
1082 if (p->signal->leader)
1086 struct task_struct *p;
1088 struct list_head *l;
1090 for_each_task_pid(pgid, PIDTYPE_PGID, p, l, pid)
1091 if (p->signal->session == current->signal->session)
1097 err = security_task_setpgid(p, pgid);
1101 if (process_group(p) != pgid) {
1102 detach_pid(p, PIDTYPE_PGID);
1103 p->signal->pgrp = pgid;
1104 attach_pid(p, PIDTYPE_PGID, pgid);
1109 /* All paths lead to here, thus we are safe. -DaveM */
1110 write_unlock_irq(&tasklist_lock);
1114 asmlinkage long sys_getpgid(pid_t pid)
1117 return process_group(current);
1120 struct task_struct *p;
1122 read_lock(&tasklist_lock);
1123 p = find_task_by_pid(pid);
1127 retval = security_task_getpgid(p);
1129 retval = process_group(p);
1131 read_unlock(&tasklist_lock);
1136 #ifdef __ARCH_WANT_SYS_GETPGRP
1138 asmlinkage long sys_getpgrp(void)
1140 /* SMP - assuming writes are word atomic this is fine */
1141 return process_group(current);
1146 asmlinkage long sys_getsid(pid_t pid)
1149 return current->signal->session;
1152 struct task_struct *p;
1154 read_lock(&tasklist_lock);
1155 p = find_task_by_pid(pid);
1159 retval = security_task_getsid(p);
1161 retval = p->signal->session;
1163 read_unlock(&tasklist_lock);
1168 asmlinkage long sys_setsid(void)
1173 if (!thread_group_leader(current))
1176 write_lock_irq(&tasklist_lock);
1178 pid = find_pid(PIDTYPE_PGID, current->pid);
1182 current->signal->leader = 1;
1183 __set_special_pids(current->pid, current->pid);
1184 current->signal->tty = NULL;
1185 current->signal->tty_old_pgrp = 0;
1186 err = process_group(current);
1188 write_unlock_irq(&tasklist_lock);
1193 * Supplementary group IDs
1196 /* init to 2 - one for init_task, one to ensure it is never freed */
1197 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1199 struct group_info *groups_alloc(int gidsetsize)
1201 struct group_info *group_info;
1205 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1206 /* Make sure we always allocate at least one indirect block pointer */
1207 nblocks = nblocks ? : 1;
1208 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1211 group_info->ngroups = gidsetsize;
1212 group_info->nblocks = nblocks;
1213 atomic_set(&group_info->usage, 1);
1215 if (gidsetsize <= NGROUPS_SMALL) {
1216 group_info->blocks[0] = group_info->small_block;
1218 for (i = 0; i < nblocks; i++) {
1220 b = (void *)__get_free_page(GFP_USER);
1222 goto out_undo_partial_alloc;
1223 group_info->blocks[i] = b;
1228 out_undo_partial_alloc:
1230 free_page((unsigned long)group_info->blocks[i]);
1236 EXPORT_SYMBOL(groups_alloc);
1238 void groups_free(struct group_info *group_info)
1240 if (group_info->blocks[0] != group_info->small_block) {
1242 for (i = 0; i < group_info->nblocks; i++)
1243 free_page((unsigned long)group_info->blocks[i]);
1248 EXPORT_SYMBOL(groups_free);
1250 /* export the group_info to a user-space array */
1251 static int groups_to_user(gid_t __user *grouplist,
1252 struct group_info *group_info)
1255 int count = group_info->ngroups;
1257 for (i = 0; i < group_info->nblocks; i++) {
1258 int cp_count = min(NGROUPS_PER_BLOCK, count);
1259 int off = i * NGROUPS_PER_BLOCK;
1260 int len = cp_count * sizeof(*grouplist);
1262 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1270 /* fill a group_info from a user-space array - it must be allocated already */
1271 static int groups_from_user(struct group_info *group_info,
1272 gid_t __user *grouplist)
1275 int count = group_info->ngroups;
1277 for (i = 0; i < group_info->nblocks; i++) {
1278 int cp_count = min(NGROUPS_PER_BLOCK, count);
1279 int off = i * NGROUPS_PER_BLOCK;
1280 int len = cp_count * sizeof(*grouplist);
1282 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1290 /* a simple shell-metzner sort */
1291 static void groups_sort(struct group_info *group_info)
1293 int base, max, stride;
1294 int gidsetsize = group_info->ngroups;
1296 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1301 max = gidsetsize - stride;
1302 for (base = 0; base < max; base++) {
1304 int right = left + stride;
1305 gid_t tmp = GROUP_AT(group_info, right);
1307 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1308 GROUP_AT(group_info, right) =
1309 GROUP_AT(group_info, left);
1313 GROUP_AT(group_info, right) = tmp;
1319 /* a simple bsearch */
1320 static int groups_search(struct group_info *group_info, gid_t grp)
1328 right = group_info->ngroups;
1329 while (left < right) {
1330 int mid = (left+right)/2;
1331 int cmp = grp - GROUP_AT(group_info, mid);
1342 /* validate and set current->group_info */
1343 int set_current_groups(struct group_info *group_info)
1346 struct group_info *old_info;
1348 retval = security_task_setgroups(group_info);
1352 groups_sort(group_info);
1353 get_group_info(group_info);
1356 old_info = current->group_info;
1357 current->group_info = group_info;
1358 task_unlock(current);
1360 put_group_info(old_info);
1365 EXPORT_SYMBOL(set_current_groups);
1367 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1372 * SMP: Nobody else can change our grouplist. Thus we are
1379 /* no need to grab task_lock here; it cannot change */
1380 get_group_info(current->group_info);
1381 i = current->group_info->ngroups;
1383 if (i > gidsetsize) {
1387 if (groups_to_user(grouplist, current->group_info)) {
1393 put_group_info(current->group_info);
1398 * SMP: Our groups are copy-on-write. We can set them safely
1399 * without another task interfering.
1402 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1404 struct group_info *group_info;
1407 if (!capable(CAP_SETGID))
1409 if ((unsigned)gidsetsize > NGROUPS_MAX)
1412 group_info = groups_alloc(gidsetsize);
1415 retval = groups_from_user(group_info, grouplist);
1417 put_group_info(group_info);
1421 retval = set_current_groups(group_info);
1422 put_group_info(group_info);
1428 * Check whether we're fsgid/egid or in the supplemental group..
1430 int in_group_p(gid_t grp)
1433 if (grp != current->fsgid) {
1434 get_group_info(current->group_info);
1435 retval = groups_search(current->group_info, grp);
1436 put_group_info(current->group_info);
1441 EXPORT_SYMBOL(in_group_p);
1443 int in_egroup_p(gid_t grp)
1446 if (grp != current->egid) {
1447 get_group_info(current->group_info);
1448 retval = groups_search(current->group_info, grp);
1449 put_group_info(current->group_info);
1454 EXPORT_SYMBOL(in_egroup_p);
1456 DECLARE_RWSEM(uts_sem);
1458 EXPORT_SYMBOL(uts_sem);
1460 asmlinkage long sys_newuname(struct new_utsname __user * name)
1464 down_read(&uts_sem);
1465 if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1471 asmlinkage long sys_sethostname(char __user *name, int len)
1474 char tmp[__NEW_UTS_LEN];
1476 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1478 if (len < 0 || len > __NEW_UTS_LEN)
1480 down_write(&uts_sem);
1482 if (!copy_from_user(tmp, name, len)) {
1483 char *ptr = vx_new_uts(nodename);
1485 memcpy(ptr, tmp, len);
1493 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1495 asmlinkage long sys_gethostname(char __user *name, int len)
1502 down_read(&uts_sem);
1503 ptr = vx_new_uts(nodename);
1504 i = 1 + strlen(ptr);
1508 if (copy_to_user(name, ptr, i))
1517 * Only setdomainname; getdomainname can be implemented by calling
1520 asmlinkage long sys_setdomainname(char __user *name, int len)
1523 char tmp[__NEW_UTS_LEN];
1525 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1527 if (len < 0 || len > __NEW_UTS_LEN)
1530 down_write(&uts_sem);
1532 if (!copy_from_user(tmp, name, len)) {
1533 char *ptr = vx_new_uts(domainname);
1535 memcpy(ptr, tmp, len);
1543 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1545 if (resource >= RLIM_NLIMITS)
1548 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1552 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1555 * Back compatibility for getrlimit. Needed for some apps.
1558 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1561 if (resource >= RLIM_NLIMITS)
1564 memcpy(&x, current->rlim + resource, sizeof(*rlim));
1565 if(x.rlim_cur > 0x7FFFFFFF)
1566 x.rlim_cur = 0x7FFFFFFF;
1567 if(x.rlim_max > 0x7FFFFFFF)
1568 x.rlim_max = 0x7FFFFFFF;
1569 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1574 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1576 struct rlimit new_rlim, *old_rlim;
1579 if (resource >= RLIM_NLIMITS)
1581 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1583 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1585 old_rlim = current->rlim + resource;
1586 if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1587 (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1588 !capable(CAP_SYS_RESOURCE) && vx_ccaps(VXC_SET_RLIMIT))
1590 if (resource == RLIMIT_NOFILE) {
1591 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1595 retval = security_task_setrlimit(resource, &new_rlim);
1599 *old_rlim = new_rlim;
1604 * It would make sense to put struct rusage in the task_struct,
1605 * except that would make the task_struct be *really big*. After
1606 * task_struct gets moved into malloc'ed memory, it would
1607 * make sense to do this. It will make moving the rest of the information
1608 * a lot simpler! (Which we're not doing right now because we're not
1609 * measuring them yet).
1611 * This is SMP safe. Either we are called from sys_getrusage on ourselves
1612 * below (we know we aren't going to exit/disappear and only we change our
1613 * rusage counters), or we are called from wait4() on a process which is
1614 * either stopped or zombied. In the zombied case the task won't get
1615 * reaped till shortly after the call to getrusage(), in both cases the
1616 * task being examined is in a frozen state so the counters won't change.
1618 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1622 memset((char *) &r, 0, sizeof(r));
1625 jiffies_to_timeval(p->utime, &r.ru_utime);
1626 jiffies_to_timeval(p->stime, &r.ru_stime);
1627 r.ru_nvcsw = p->nvcsw;
1628 r.ru_nivcsw = p->nivcsw;
1629 r.ru_minflt = p->min_flt;
1630 r.ru_majflt = p->maj_flt;
1632 case RUSAGE_CHILDREN:
1633 jiffies_to_timeval(p->cutime, &r.ru_utime);
1634 jiffies_to_timeval(p->cstime, &r.ru_stime);
1635 r.ru_nvcsw = p->cnvcsw;
1636 r.ru_nivcsw = p->cnivcsw;
1637 r.ru_minflt = p->cmin_flt;
1638 r.ru_majflt = p->cmaj_flt;
1641 jiffies_to_timeval(p->utime + p->cutime, &r.ru_utime);
1642 jiffies_to_timeval(p->stime + p->cstime, &r.ru_stime);
1643 r.ru_nvcsw = p->nvcsw + p->cnvcsw;
1644 r.ru_nivcsw = p->nivcsw + p->cnivcsw;
1645 r.ru_minflt = p->min_flt + p->cmin_flt;
1646 r.ru_majflt = p->maj_flt + p->cmaj_flt;
1649 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1652 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1654 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1656 return getrusage(current, who, ru);
1659 asmlinkage long sys_umask(int mask)
1661 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1665 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1666 unsigned long arg4, unsigned long arg5)
1671 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1676 case PR_SET_PDEATHSIG:
1678 if (sig < 0 || sig > _NSIG) {
1682 current->pdeath_signal = sig;
1684 case PR_GET_PDEATHSIG:
1685 error = put_user(current->pdeath_signal, (int __user *)arg2);
1687 case PR_GET_DUMPABLE:
1688 if (current->mm->dumpable)
1691 case PR_SET_DUMPABLE:
1692 if (arg2 != 0 && arg2 != 1) {
1696 current->mm->dumpable = arg2;
1699 case PR_SET_UNALIGN:
1700 error = SET_UNALIGN_CTL(current, arg2);
1702 case PR_GET_UNALIGN:
1703 error = GET_UNALIGN_CTL(current, arg2);
1706 error = SET_FPEMU_CTL(current, arg2);
1709 error = GET_FPEMU_CTL(current, arg2);
1712 error = SET_FPEXC_CTL(current, arg2);
1715 error = GET_FPEXC_CTL(current, arg2);
1718 error = PR_TIMING_STATISTICAL;
1721 if (arg2 == PR_TIMING_STATISTICAL)
1727 case PR_GET_KEEPCAPS:
1728 if (current->keep_capabilities)
1731 case PR_SET_KEEPCAPS:
1732 if (arg2 != 0 && arg2 != 1) {
1736 current->keep_capabilities = arg2;