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)
278 cond_syscall(compat_get_mempolicy)
280 /* arch-specific weak syscall entries */
281 cond_syscall(sys_pciconfig_read)
282 cond_syscall(sys_pciconfig_write)
283 cond_syscall(sys_pciconfig_iobase)
285 static int set_one_prio(struct task_struct *p, int niceval, int error)
289 if (p->uid != current->euid &&
290 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
294 if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
298 no_nice = security_task_setnice(p, niceval);
305 set_user_nice(p, niceval);
310 asmlinkage long sys_setpriority(int which, int who, int niceval)
312 struct task_struct *g, *p;
313 struct user_struct *user;
318 if (which > 2 || which < 0)
321 /* normalize: avoid signed division (rounding problems) */
328 read_lock(&tasklist_lock);
333 p = find_task_by_pid(who);
335 error = set_one_prio(p, niceval, error);
339 who = process_group(current);
340 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid)
341 error = set_one_prio(p, niceval, error);
345 user = current->user;
347 user = find_user(vx_current_xid(), who);
354 error = set_one_prio(p, niceval, error);
355 while_each_thread(g, p);
357 free_uid(user); /* For find_user() */
361 read_unlock(&tasklist_lock);
367 * Ugh. To avoid negative return values, "getpriority()" will
368 * not return the normal nice-value, but a negated value that
369 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
370 * to stay compatible.
372 asmlinkage long sys_getpriority(int which, int who)
374 struct task_struct *g, *p;
377 struct user_struct *user;
378 long niceval, retval = -ESRCH;
380 if (which > 2 || which < 0)
383 read_lock(&tasklist_lock);
388 p = find_task_by_pid(who);
390 niceval = 20 - task_nice(p);
391 if (niceval > retval)
397 who = process_group(current);
398 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid) {
399 niceval = 20 - task_nice(p);
400 if (niceval > retval)
406 user = current->user;
408 user = find_user(vx_current_xid(), who);
415 niceval = 20 - task_nice(p);
416 if (niceval > retval)
419 while_each_thread(g, p);
421 free_uid(user); /* for find_user() */
425 read_unlock(&tasklist_lock);
431 * vshelper path is set via /proc/sys
432 * invoked by vserver sys_reboot(), with
433 * the following arguments
435 * argv [0] = vshelper_path;
436 * argv [1] = action: "restart", "halt", "poweroff", ...
437 * argv [2] = context identifier
438 * argv [3] = additional argument (restart2)
440 * envp [*] = type-specific parameters
442 char vshelper_path[255] = "/sbin/vshelper";
444 long vs_reboot(unsigned int cmd, void * arg)
446 char id_buf[8], cmd_buf[32];
447 char uid_buf[32], pid_buf[32];
450 char *argv[] = {vshelper_path, NULL, id_buf, NULL, 0};
451 char *envp[] = {"HOME=/", "TERM=linux",
452 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
453 uid_buf, pid_buf, cmd_buf, 0};
455 snprintf(id_buf, sizeof(id_buf)-1, "%d", vx_current_xid());
457 snprintf(cmd_buf, sizeof(cmd_buf)-1, "VS_CMD=%08x", cmd);
458 snprintf(uid_buf, sizeof(uid_buf)-1, "VS_UID=%d", current->uid);
459 snprintf(pid_buf, sizeof(pid_buf)-1, "VS_PID=%d", current->pid);
462 case LINUX_REBOOT_CMD_RESTART:
466 case LINUX_REBOOT_CMD_HALT:
470 case LINUX_REBOOT_CMD_POWER_OFF:
471 argv[1] = "poweroff";
474 case LINUX_REBOOT_CMD_SW_SUSPEND:
478 case LINUX_REBOOT_CMD_RESTART2:
479 if (strncpy_from_user(&buffer[0], (char *)arg, sizeof(buffer) - 1) < 0)
483 argv[1] = "restart2";
487 /* maybe we should wait ? */
488 if (call_usermodehelper(*argv, argv, envp, 0)) {
490 "vs_reboot(): failed to exec (%s %s %s %s)\n",
491 vshelper_path, argv[1], argv[2], argv[3]);
498 * Reboot system call: for obvious reasons only root may call it,
499 * and even root needs to set up some magic numbers in the registers
500 * so that some mistake won't make this reboot the whole machine.
501 * You can also set the meaning of the ctrl-alt-del-key here.
503 * reboot doesn't sync: do that yourself before calling this.
505 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
509 /* We only trust the superuser with rebooting the system. */
510 if (!capable(CAP_SYS_BOOT))
513 /* For safety, we require "magic" arguments. */
514 if (magic1 != LINUX_REBOOT_MAGIC1 ||
515 (magic2 != LINUX_REBOOT_MAGIC2 &&
516 magic2 != LINUX_REBOOT_MAGIC2A &&
517 magic2 != LINUX_REBOOT_MAGIC2B &&
518 magic2 != LINUX_REBOOT_MAGIC2C))
521 if (!vx_check(0, VX_ADMIN|VX_WATCH))
522 return vs_reboot(cmd, arg);
526 case LINUX_REBOOT_CMD_RESTART:
527 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
528 system_state = SYSTEM_RESTART;
530 printk(KERN_EMERG "Restarting system.\n");
531 machine_restart(NULL);
534 case LINUX_REBOOT_CMD_CAD_ON:
538 case LINUX_REBOOT_CMD_CAD_OFF:
542 case LINUX_REBOOT_CMD_HALT:
543 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
544 system_state = SYSTEM_HALT;
546 printk(KERN_EMERG "System halted.\n");
552 case LINUX_REBOOT_CMD_POWER_OFF:
553 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
554 system_state = SYSTEM_POWER_OFF;
556 printk(KERN_EMERG "Power down.\n");
562 case LINUX_REBOOT_CMD_RESTART2:
563 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
567 buffer[sizeof(buffer) - 1] = '\0';
569 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
570 system_state = SYSTEM_RESTART;
572 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
573 machine_restart(buffer);
576 #ifdef CONFIG_SOFTWARE_SUSPEND
577 case LINUX_REBOOT_CMD_SW_SUSPEND:
579 int ret = software_suspend();
593 static void deferred_cad(void *dummy)
595 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
596 machine_restart(NULL);
600 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
601 * As it's called within an interrupt, it may NOT sync: the only choice
602 * is whether to reboot at once, or just ignore the ctrl-alt-del.
604 void ctrl_alt_del(void)
606 static DECLARE_WORK(cad_work, deferred_cad, NULL);
609 schedule_work(&cad_work);
611 kill_proc(cad_pid, SIGINT, 1);
616 * Unprivileged users may change the real gid to the effective gid
617 * or vice versa. (BSD-style)
619 * If you set the real gid at all, or set the effective gid to a value not
620 * equal to the real gid, then the saved gid is set to the new effective gid.
622 * This makes it possible for a setgid program to completely drop its
623 * privileges, which is often a useful assertion to make when you are doing
624 * a security audit over a program.
626 * The general idea is that a program which uses just setregid() will be
627 * 100% compatible with BSD. A program which uses just setgid() will be
628 * 100% compatible with POSIX with saved IDs.
630 * SMP: There are not races, the GIDs are checked only by filesystem
631 * operations (as far as semantic preservation is concerned).
633 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
635 int old_rgid = current->gid;
636 int old_egid = current->egid;
637 int new_rgid = old_rgid;
638 int new_egid = old_egid;
641 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
645 if (rgid != (gid_t) -1) {
646 if ((old_rgid == rgid) ||
647 (current->egid==rgid) ||
653 if (egid != (gid_t) -1) {
654 if ((old_rgid == egid) ||
655 (current->egid == egid) ||
656 (current->sgid == egid) ||
663 if (new_egid != old_egid)
665 current->mm->dumpable = 0;
668 if (rgid != (gid_t) -1 ||
669 (egid != (gid_t) -1 && egid != old_rgid))
670 current->sgid = new_egid;
671 current->fsgid = new_egid;
672 current->egid = new_egid;
673 current->gid = new_rgid;
678 * setgid() is implemented like SysV w/ SAVED_IDS
680 * SMP: Same implicit races as above.
682 asmlinkage long sys_setgid(gid_t gid)
684 int old_egid = current->egid;
687 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
691 if (capable(CAP_SETGID))
695 current->mm->dumpable=0;
698 current->gid = current->egid = current->sgid = current->fsgid = gid;
700 else if ((gid == current->gid) || (gid == current->sgid))
704 current->mm->dumpable=0;
707 current->egid = current->fsgid = gid;
714 static int set_user(uid_t new_ruid, int dumpclear)
716 struct user_struct *new_user;
718 new_user = alloc_uid(vx_current_xid(), new_ruid);
722 if (atomic_read(&new_user->processes) >=
723 current->rlim[RLIMIT_NPROC].rlim_cur &&
724 new_user != &root_user) {
729 switch_uid(new_user);
733 current->mm->dumpable = 0;
736 current->uid = new_ruid;
741 * Unprivileged users may change the real uid to the effective uid
742 * or vice versa. (BSD-style)
744 * If you set the real uid at all, or set the effective uid to a value not
745 * equal to the real uid, then the saved uid is set to the new effective uid.
747 * This makes it possible for a setuid program to completely drop its
748 * privileges, which is often a useful assertion to make when you are doing
749 * a security audit over a program.
751 * The general idea is that a program which uses just setreuid() will be
752 * 100% compatible with BSD. A program which uses just setuid() will be
753 * 100% compatible with POSIX with saved IDs.
755 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
757 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
760 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
764 new_ruid = old_ruid = current->uid;
765 new_euid = old_euid = current->euid;
766 old_suid = current->suid;
768 if (ruid != (uid_t) -1) {
770 if ((old_ruid != ruid) &&
771 (current->euid != ruid) &&
772 !capable(CAP_SETUID))
776 if (euid != (uid_t) -1) {
778 if ((old_ruid != euid) &&
779 (current->euid != euid) &&
780 (current->suid != euid) &&
781 !capable(CAP_SETUID))
785 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
788 if (new_euid != old_euid)
790 current->mm->dumpable=0;
793 current->fsuid = current->euid = new_euid;
794 if (ruid != (uid_t) -1 ||
795 (euid != (uid_t) -1 && euid != old_ruid))
796 current->suid = current->euid;
797 current->fsuid = current->euid;
799 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
805 * setuid() is implemented like SysV with SAVED_IDS
807 * Note that SAVED_ID's is deficient in that a setuid root program
808 * like sendmail, for example, cannot set its uid to be a normal
809 * user and then switch back, because if you're root, setuid() sets
810 * the saved uid too. If you don't like this, blame the bright people
811 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
812 * will allow a root program to temporarily drop privileges and be able to
813 * regain them by swapping the real and effective uid.
815 asmlinkage long sys_setuid(uid_t uid)
817 int old_euid = current->euid;
818 int old_ruid, old_suid, new_ruid, new_suid;
821 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
825 old_ruid = new_ruid = current->uid;
826 old_suid = current->suid;
829 if (capable(CAP_SETUID)) {
830 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
833 } else if ((uid != current->uid) && (uid != new_suid))
838 current->mm->dumpable = 0;
841 current->fsuid = current->euid = uid;
842 current->suid = new_suid;
844 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
849 * This function implements a generic ability to update ruid, euid,
850 * and suid. This allows you to implement the 4.4 compatible seteuid().
852 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
854 int old_ruid = current->uid;
855 int old_euid = current->euid;
856 int old_suid = current->suid;
859 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
863 if (!capable(CAP_SETUID)) {
864 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
865 (ruid != current->euid) && (ruid != current->suid))
867 if ((euid != (uid_t) -1) && (euid != current->uid) &&
868 (euid != current->euid) && (euid != current->suid))
870 if ((suid != (uid_t) -1) && (suid != current->uid) &&
871 (suid != current->euid) && (suid != current->suid))
874 if (ruid != (uid_t) -1) {
875 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
878 if (euid != (uid_t) -1) {
879 if (euid != current->euid)
881 current->mm->dumpable = 0;
884 current->euid = euid;
886 current->fsuid = current->euid;
887 if (suid != (uid_t) -1)
888 current->suid = suid;
890 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
893 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
897 if (!(retval = put_user(current->uid, ruid)) &&
898 !(retval = put_user(current->euid, euid)))
899 retval = put_user(current->suid, suid);
905 * Same as above, but for rgid, egid, sgid.
907 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
911 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
915 if (!capable(CAP_SETGID)) {
916 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
917 (rgid != current->egid) && (rgid != current->sgid))
919 if ((egid != (gid_t) -1) && (egid != current->gid) &&
920 (egid != current->egid) && (egid != current->sgid))
922 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
923 (sgid != current->egid) && (sgid != current->sgid))
926 if (egid != (gid_t) -1) {
927 if (egid != current->egid)
929 current->mm->dumpable = 0;
932 current->egid = egid;
934 current->fsgid = current->egid;
935 if (rgid != (gid_t) -1)
937 if (sgid != (gid_t) -1)
938 current->sgid = sgid;
942 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
946 if (!(retval = put_user(current->gid, rgid)) &&
947 !(retval = put_user(current->egid, egid)))
948 retval = put_user(current->sgid, sgid);
955 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
956 * is used for "access()" and for the NFS daemon (letting nfsd stay at
957 * whatever uid it wants to). It normally shadows "euid", except when
958 * explicitly set by setfsuid() or for access..
960 asmlinkage long sys_setfsuid(uid_t uid)
964 old_fsuid = current->fsuid;
965 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
968 if (uid == current->uid || uid == current->euid ||
969 uid == current->suid || uid == current->fsuid ||
972 if (uid != old_fsuid)
974 current->mm->dumpable = 0;
977 current->fsuid = uid;
980 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
986 * Samma på svenska..
988 asmlinkage long sys_setfsgid(gid_t gid)
992 old_fsgid = current->fsgid;
993 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
996 if (gid == current->gid || gid == current->egid ||
997 gid == current->sgid || gid == current->fsgid ||
1000 if (gid != old_fsgid)
1002 current->mm->dumpable = 0;
1005 current->fsgid = gid;
1010 asmlinkage long sys_times(struct tms __user * tbuf)
1013 * In the SMP world we might just be unlucky and have one of
1014 * the times increment as we use it. Since the value is an
1015 * atomically safe type this is just fine. Conceptually its
1016 * as if the syscall took an instant longer to occur.
1020 tmp.tms_utime = jiffies_to_clock_t(current->utime);
1021 tmp.tms_stime = jiffies_to_clock_t(current->stime);
1022 tmp.tms_cutime = jiffies_to_clock_t(current->cutime);
1023 tmp.tms_cstime = jiffies_to_clock_t(current->cstime);
1024 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1027 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1031 * This needs some heavy checking ...
1032 * I just haven't the stomach for it. I also don't fully
1033 * understand sessions/pgrp etc. Let somebody who does explain it.
1035 * OK, I think I have the protection semantics right.... this is really
1036 * only important on a multi-user system anyway, to make sure one user
1037 * can't send a signal to a process owned by another. -TYT, 12/12/91
1039 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1043 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1045 struct task_struct *p;
1055 /* From this point forward we keep holding onto the tasklist lock
1056 * so that our parent does not change from under us. -DaveM
1058 write_lock_irq(&tasklist_lock);
1061 p = find_task_by_pid(pid);
1066 if (!thread_group_leader(p))
1069 if (p->parent == current || p->real_parent == current) {
1071 if (p->signal->session != current->signal->session)
1083 if (p->signal->leader)
1087 struct task_struct *p;
1089 struct list_head *l;
1091 for_each_task_pid(pgid, PIDTYPE_PGID, p, l, pid)
1092 if (p->signal->session == current->signal->session)
1098 err = security_task_setpgid(p, pgid);
1102 if (process_group(p) != pgid) {
1103 detach_pid(p, PIDTYPE_PGID);
1104 p->signal->pgrp = pgid;
1105 attach_pid(p, PIDTYPE_PGID, pgid);
1110 /* All paths lead to here, thus we are safe. -DaveM */
1111 write_unlock_irq(&tasklist_lock);
1115 asmlinkage long sys_getpgid(pid_t pid)
1118 return process_group(current);
1121 struct task_struct *p;
1123 read_lock(&tasklist_lock);
1124 p = find_task_by_pid(pid);
1128 retval = security_task_getpgid(p);
1130 retval = process_group(p);
1132 read_unlock(&tasklist_lock);
1137 #ifdef __ARCH_WANT_SYS_GETPGRP
1139 asmlinkage long sys_getpgrp(void)
1141 /* SMP - assuming writes are word atomic this is fine */
1142 return process_group(current);
1147 asmlinkage long sys_getsid(pid_t pid)
1150 return current->signal->session;
1153 struct task_struct *p;
1155 read_lock(&tasklist_lock);
1156 p = find_task_by_pid(pid);
1160 retval = security_task_getsid(p);
1162 retval = p->signal->session;
1164 read_unlock(&tasklist_lock);
1169 asmlinkage long sys_setsid(void)
1174 if (!thread_group_leader(current))
1177 write_lock_irq(&tasklist_lock);
1179 pid = find_pid(PIDTYPE_PGID, current->pid);
1183 current->signal->leader = 1;
1184 __set_special_pids(current->pid, current->pid);
1185 current->signal->tty = NULL;
1186 current->signal->tty_old_pgrp = 0;
1187 err = process_group(current);
1189 write_unlock_irq(&tasklist_lock);
1194 * Supplementary group IDs
1197 /* init to 2 - one for init_task, one to ensure it is never freed */
1198 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1200 struct group_info *groups_alloc(int gidsetsize)
1202 struct group_info *group_info;
1206 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1207 /* Make sure we always allocate at least one indirect block pointer */
1208 nblocks = nblocks ? : 1;
1209 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1212 group_info->ngroups = gidsetsize;
1213 group_info->nblocks = nblocks;
1214 atomic_set(&group_info->usage, 1);
1216 if (gidsetsize <= NGROUPS_SMALL) {
1217 group_info->blocks[0] = group_info->small_block;
1219 for (i = 0; i < nblocks; i++) {
1221 b = (void *)__get_free_page(GFP_USER);
1223 goto out_undo_partial_alloc;
1224 group_info->blocks[i] = b;
1229 out_undo_partial_alloc:
1231 free_page((unsigned long)group_info->blocks[i]);
1237 EXPORT_SYMBOL(groups_alloc);
1239 void groups_free(struct group_info *group_info)
1241 if (group_info->blocks[0] != group_info->small_block) {
1243 for (i = 0; i < group_info->nblocks; i++)
1244 free_page((unsigned long)group_info->blocks[i]);
1249 EXPORT_SYMBOL(groups_free);
1251 /* export the group_info to a user-space array */
1252 static int groups_to_user(gid_t __user *grouplist,
1253 struct group_info *group_info)
1256 int count = group_info->ngroups;
1258 for (i = 0; i < group_info->nblocks; i++) {
1259 int cp_count = min(NGROUPS_PER_BLOCK, count);
1260 int off = i * NGROUPS_PER_BLOCK;
1261 int len = cp_count * sizeof(*grouplist);
1263 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1271 /* fill a group_info from a user-space array - it must be allocated already */
1272 static int groups_from_user(struct group_info *group_info,
1273 gid_t __user *grouplist)
1276 int count = group_info->ngroups;
1278 for (i = 0; i < group_info->nblocks; i++) {
1279 int cp_count = min(NGROUPS_PER_BLOCK, count);
1280 int off = i * NGROUPS_PER_BLOCK;
1281 int len = cp_count * sizeof(*grouplist);
1283 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1291 /* a simple shell-metzner sort */
1292 static void groups_sort(struct group_info *group_info)
1294 int base, max, stride;
1295 int gidsetsize = group_info->ngroups;
1297 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1302 max = gidsetsize - stride;
1303 for (base = 0; base < max; base++) {
1305 int right = left + stride;
1306 gid_t tmp = GROUP_AT(group_info, right);
1308 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1309 GROUP_AT(group_info, right) =
1310 GROUP_AT(group_info, left);
1314 GROUP_AT(group_info, right) = tmp;
1320 /* a simple bsearch */
1321 static int groups_search(struct group_info *group_info, gid_t grp)
1329 right = group_info->ngroups;
1330 while (left < right) {
1331 int mid = (left+right)/2;
1332 int cmp = grp - GROUP_AT(group_info, mid);
1343 /* validate and set current->group_info */
1344 int set_current_groups(struct group_info *group_info)
1347 struct group_info *old_info;
1349 retval = security_task_setgroups(group_info);
1353 groups_sort(group_info);
1354 get_group_info(group_info);
1357 old_info = current->group_info;
1358 current->group_info = group_info;
1359 task_unlock(current);
1361 put_group_info(old_info);
1366 EXPORT_SYMBOL(set_current_groups);
1368 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1373 * SMP: Nobody else can change our grouplist. Thus we are
1380 /* no need to grab task_lock here; it cannot change */
1381 get_group_info(current->group_info);
1382 i = current->group_info->ngroups;
1384 if (i > gidsetsize) {
1388 if (groups_to_user(grouplist, current->group_info)) {
1394 put_group_info(current->group_info);
1399 * SMP: Our groups are copy-on-write. We can set them safely
1400 * without another task interfering.
1403 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1405 struct group_info *group_info;
1408 if (!capable(CAP_SETGID))
1410 if ((unsigned)gidsetsize > NGROUPS_MAX)
1413 group_info = groups_alloc(gidsetsize);
1416 retval = groups_from_user(group_info, grouplist);
1418 put_group_info(group_info);
1422 retval = set_current_groups(group_info);
1423 put_group_info(group_info);
1429 * Check whether we're fsgid/egid or in the supplemental group..
1431 int in_group_p(gid_t grp)
1434 if (grp != current->fsgid) {
1435 get_group_info(current->group_info);
1436 retval = groups_search(current->group_info, grp);
1437 put_group_info(current->group_info);
1442 EXPORT_SYMBOL(in_group_p);
1444 int in_egroup_p(gid_t grp)
1447 if (grp != current->egid) {
1448 get_group_info(current->group_info);
1449 retval = groups_search(current->group_info, grp);
1450 put_group_info(current->group_info);
1455 EXPORT_SYMBOL(in_egroup_p);
1457 DECLARE_RWSEM(uts_sem);
1459 EXPORT_SYMBOL(uts_sem);
1461 asmlinkage long sys_newuname(struct new_utsname __user * name)
1465 down_read(&uts_sem);
1466 if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1472 asmlinkage long sys_sethostname(char __user *name, int len)
1475 char tmp[__NEW_UTS_LEN];
1477 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1479 if (len < 0 || len > __NEW_UTS_LEN)
1481 down_write(&uts_sem);
1483 if (!copy_from_user(tmp, name, len)) {
1484 char *ptr = vx_new_uts(nodename);
1486 memcpy(ptr, tmp, len);
1494 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1496 asmlinkage long sys_gethostname(char __user *name, int len)
1503 down_read(&uts_sem);
1504 ptr = vx_new_uts(nodename);
1505 i = 1 + strlen(ptr);
1509 if (copy_to_user(name, ptr, i))
1518 * Only setdomainname; getdomainname can be implemented by calling
1521 asmlinkage long sys_setdomainname(char __user *name, int len)
1524 char tmp[__NEW_UTS_LEN];
1526 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1528 if (len < 0 || len > __NEW_UTS_LEN)
1531 down_write(&uts_sem);
1533 if (!copy_from_user(tmp, name, len)) {
1534 char *ptr = vx_new_uts(domainname);
1536 memcpy(ptr, tmp, len);
1544 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1546 if (resource >= RLIM_NLIMITS)
1549 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1553 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1556 * Back compatibility for getrlimit. Needed for some apps.
1559 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1562 if (resource >= RLIM_NLIMITS)
1565 memcpy(&x, current->rlim + resource, sizeof(*rlim));
1566 if(x.rlim_cur > 0x7FFFFFFF)
1567 x.rlim_cur = 0x7FFFFFFF;
1568 if(x.rlim_max > 0x7FFFFFFF)
1569 x.rlim_max = 0x7FFFFFFF;
1570 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1575 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1577 struct rlimit new_rlim, *old_rlim;
1580 if (resource >= RLIM_NLIMITS)
1582 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1584 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1586 old_rlim = current->rlim + resource;
1587 if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1588 (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1589 !capable(CAP_SYS_RESOURCE) && vx_ccaps(VXC_SET_RLIMIT))
1591 if (resource == RLIMIT_NOFILE) {
1592 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1596 retval = security_task_setrlimit(resource, &new_rlim);
1600 *old_rlim = new_rlim;
1605 * It would make sense to put struct rusage in the task_struct,
1606 * except that would make the task_struct be *really big*. After
1607 * task_struct gets moved into malloc'ed memory, it would
1608 * make sense to do this. It will make moving the rest of the information
1609 * a lot simpler! (Which we're not doing right now because we're not
1610 * measuring them yet).
1612 * This is SMP safe. Either we are called from sys_getrusage on ourselves
1613 * below (we know we aren't going to exit/disappear and only we change our
1614 * rusage counters), or we are called from wait4() on a process which is
1615 * either stopped or zombied. In the zombied case the task won't get
1616 * reaped till shortly after the call to getrusage(), in both cases the
1617 * task being examined is in a frozen state so the counters won't change.
1619 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1623 memset((char *) &r, 0, sizeof(r));
1626 jiffies_to_timeval(p->utime, &r.ru_utime);
1627 jiffies_to_timeval(p->stime, &r.ru_stime);
1628 r.ru_nvcsw = p->nvcsw;
1629 r.ru_nivcsw = p->nivcsw;
1630 r.ru_minflt = p->min_flt;
1631 r.ru_majflt = p->maj_flt;
1633 case RUSAGE_CHILDREN:
1634 jiffies_to_timeval(p->cutime, &r.ru_utime);
1635 jiffies_to_timeval(p->cstime, &r.ru_stime);
1636 r.ru_nvcsw = p->cnvcsw;
1637 r.ru_nivcsw = p->cnivcsw;
1638 r.ru_minflt = p->cmin_flt;
1639 r.ru_majflt = p->cmaj_flt;
1642 jiffies_to_timeval(p->utime + p->cutime, &r.ru_utime);
1643 jiffies_to_timeval(p->stime + p->cstime, &r.ru_stime);
1644 r.ru_nvcsw = p->nvcsw + p->cnvcsw;
1645 r.ru_nivcsw = p->nivcsw + p->cnivcsw;
1646 r.ru_minflt = p->min_flt + p->cmin_flt;
1647 r.ru_majflt = p->maj_flt + p->cmaj_flt;
1650 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1653 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1655 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1657 return getrusage(current, who, ru);
1660 asmlinkage long sys_umask(int mask)
1662 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1666 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1667 unsigned long arg4, unsigned long arg5)
1672 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1677 case PR_SET_PDEATHSIG:
1679 if (sig < 0 || sig > _NSIG) {
1683 current->pdeath_signal = sig;
1685 case PR_GET_PDEATHSIG:
1686 error = put_user(current->pdeath_signal, (int __user *)arg2);
1688 case PR_GET_DUMPABLE:
1689 if (current->mm->dumpable)
1692 case PR_SET_DUMPABLE:
1693 if (arg2 != 0 && arg2 != 1) {
1697 current->mm->dumpable = arg2;
1700 case PR_SET_UNALIGN:
1701 error = SET_UNALIGN_CTL(current, arg2);
1703 case PR_GET_UNALIGN:
1704 error = GET_UNALIGN_CTL(current, arg2);
1707 error = SET_FPEMU_CTL(current, arg2);
1710 error = GET_FPEMU_CTL(current, arg2);
1713 error = SET_FPEXC_CTL(current, arg2);
1716 error = GET_FPEXC_CTL(current, arg2);
1719 error = PR_TIMING_STATISTICAL;
1722 if (arg2 == PR_TIMING_STATISTICAL)
1728 case PR_GET_KEEPCAPS:
1729 if (current->keep_capabilities)
1732 case PR_SET_KEEPCAPS:
1733 if (arg2 != 0 && arg2 != 1) {
1737 current->keep_capabilities = arg2;