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>
27 #include <linux/ckrm.h>
29 #include <asm/uaccess.h>
31 #include <asm/unistd.h>
33 #ifndef SET_UNALIGN_CTL
34 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
36 #ifndef GET_UNALIGN_CTL
37 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
40 # define SET_FPEMU_CTL(a,b) (-EINVAL)
43 # define GET_FPEMU_CTL(a,b) (-EINVAL)
46 # define SET_FPEXC_CTL(a,b) (-EINVAL)
49 # define GET_FPEXC_CTL(a,b) (-EINVAL)
53 * this is where the system-wide overflow UID and GID are defined, for
54 * architectures that now have 32-bit UID/GID but didn't in the past
57 int overflowuid = DEFAULT_OVERFLOWUID;
58 int overflowgid = DEFAULT_OVERFLOWGID;
61 EXPORT_SYMBOL(overflowuid);
62 EXPORT_SYMBOL(overflowgid);
66 * the same as above, but for filesystems which can only store a 16-bit
67 * UID and GID. as such, this is needed on all architectures
70 int fs_overflowuid = DEFAULT_FS_OVERFLOWUID;
71 int fs_overflowgid = DEFAULT_FS_OVERFLOWUID;
73 EXPORT_SYMBOL(fs_overflowuid);
74 EXPORT_SYMBOL(fs_overflowgid);
77 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
84 * Notifier list for kernel code which wants to be called
85 * at shutdown. This is used to stop any idling DMA operations
89 static struct notifier_block *reboot_notifier_list;
90 rwlock_t notifier_lock = RW_LOCK_UNLOCKED;
93 * notifier_chain_register - Add notifier to a notifier chain
94 * @list: Pointer to root list pointer
95 * @n: New entry in notifier chain
97 * Adds a notifier to a notifier chain.
99 * Currently always returns zero.
102 int notifier_chain_register(struct notifier_block **list, struct notifier_block *n)
104 write_lock(¬ifier_lock);
107 if(n->priority > (*list)->priority)
109 list= &((*list)->next);
113 write_unlock(¬ifier_lock);
117 EXPORT_SYMBOL(notifier_chain_register);
120 * notifier_chain_unregister - Remove notifier from a notifier chain
121 * @nl: Pointer to root list pointer
122 * @n: New entry in notifier chain
124 * Removes a notifier from a notifier chain.
126 * Returns zero on success, or %-ENOENT on failure.
129 int notifier_chain_unregister(struct notifier_block **nl, struct notifier_block *n)
131 write_lock(¬ifier_lock);
137 write_unlock(¬ifier_lock);
142 write_unlock(¬ifier_lock);
146 EXPORT_SYMBOL(notifier_chain_unregister);
149 * notifier_call_chain - Call functions in a notifier chain
150 * @n: Pointer to root pointer of notifier chain
151 * @val: Value passed unmodified to notifier function
152 * @v: Pointer passed unmodified to notifier function
154 * Calls each function in a notifier chain in turn.
156 * If the return value of the notifier can be and'd
157 * with %NOTIFY_STOP_MASK, then notifier_call_chain
158 * will return immediately, with the return value of
159 * the notifier function which halted execution.
160 * Otherwise, the return value is the return value
161 * of the last notifier function called.
164 int notifier_call_chain(struct notifier_block **n, unsigned long val, void *v)
167 struct notifier_block *nb = *n;
171 ret=nb->notifier_call(nb,val,v);
172 if(ret&NOTIFY_STOP_MASK)
181 EXPORT_SYMBOL(notifier_call_chain);
184 * register_reboot_notifier - Register function to be called at reboot time
185 * @nb: Info about notifier function to be called
187 * Registers a function with the list of functions
188 * to be called at reboot time.
190 * Currently always returns zero, as notifier_chain_register
191 * always returns zero.
194 int register_reboot_notifier(struct notifier_block * nb)
196 return notifier_chain_register(&reboot_notifier_list, nb);
199 EXPORT_SYMBOL(register_reboot_notifier);
202 * unregister_reboot_notifier - Unregister previously registered reboot notifier
203 * @nb: Hook to be unregistered
205 * Unregisters a previously registered reboot
208 * Returns zero on success, or %-ENOENT on failure.
211 int unregister_reboot_notifier(struct notifier_block * nb)
213 return notifier_chain_unregister(&reboot_notifier_list, nb);
216 EXPORT_SYMBOL(unregister_reboot_notifier);
218 asmlinkage long sys_ni_syscall(void)
223 cond_syscall(sys_nfsservctl)
224 cond_syscall(sys_quotactl)
225 cond_syscall(sys_acct)
226 cond_syscall(sys_lookup_dcookie)
227 cond_syscall(sys_swapon)
228 cond_syscall(sys_swapoff)
229 cond_syscall(sys_init_module)
230 cond_syscall(sys_delete_module)
231 cond_syscall(sys_socketpair)
232 cond_syscall(sys_bind)
233 cond_syscall(sys_listen)
234 cond_syscall(sys_accept)
235 cond_syscall(sys_connect)
236 cond_syscall(sys_getsockname)
237 cond_syscall(sys_getpeername)
238 cond_syscall(sys_sendto)
239 cond_syscall(sys_send)
240 cond_syscall(sys_recvfrom)
241 cond_syscall(sys_recv)
242 cond_syscall(sys_socket)
243 cond_syscall(sys_setsockopt)
244 cond_syscall(sys_getsockopt)
245 cond_syscall(sys_shutdown)
246 cond_syscall(sys_sendmsg)
247 cond_syscall(sys_recvmsg)
248 cond_syscall(sys_socketcall)
249 cond_syscall(sys_futex)
250 cond_syscall(compat_sys_futex)
251 cond_syscall(sys_epoll_create)
252 cond_syscall(sys_epoll_ctl)
253 cond_syscall(sys_epoll_wait)
254 cond_syscall(sys_semget)
255 cond_syscall(sys_semop)
256 cond_syscall(sys_semtimedop)
257 cond_syscall(sys_semctl)
258 cond_syscall(sys_msgget)
259 cond_syscall(sys_msgsnd)
260 cond_syscall(sys_msgrcv)
261 cond_syscall(sys_msgctl)
262 cond_syscall(sys_shmget)
263 cond_syscall(sys_shmdt)
264 cond_syscall(sys_shmctl)
265 cond_syscall(sys_mq_open)
266 cond_syscall(sys_mq_unlink)
267 cond_syscall(sys_mq_timedsend)
268 cond_syscall(sys_mq_timedreceive)
269 cond_syscall(sys_mq_notify)
270 cond_syscall(sys_mq_getsetattr)
271 cond_syscall(compat_sys_mq_open)
272 cond_syscall(compat_sys_mq_timedsend)
273 cond_syscall(compat_sys_mq_timedreceive)
274 cond_syscall(compat_sys_mq_notify)
275 cond_syscall(compat_sys_mq_getsetattr)
276 cond_syscall(sys_mbind)
277 cond_syscall(sys_get_mempolicy)
278 cond_syscall(sys_set_mempolicy)
279 cond_syscall(compat_get_mempolicy)
281 /* arch-specific weak syscall entries */
282 cond_syscall(sys_pciconfig_read)
283 cond_syscall(sys_pciconfig_write)
284 cond_syscall(sys_pciconfig_iobase)
286 static int set_one_prio(struct task_struct *p, int niceval, int error)
290 if (p->uid != current->euid &&
291 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
295 if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
299 no_nice = security_task_setnice(p, niceval);
306 set_user_nice(p, niceval);
311 asmlinkage long sys_setpriority(int which, int who, int niceval)
313 struct task_struct *g, *p;
314 struct user_struct *user;
319 if (which > 2 || which < 0)
322 /* normalize: avoid signed division (rounding problems) */
329 read_lock(&tasklist_lock);
334 p = find_task_by_pid(who);
336 error = set_one_prio(p, niceval, error);
340 who = process_group(current);
341 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid)
342 error = set_one_prio(p, niceval, error);
346 user = current->user;
348 user = find_user(vx_current_xid(), who);
355 error = set_one_prio(p, niceval, error);
356 while_each_thread(g, p);
358 free_uid(user); /* For find_user() */
362 read_unlock(&tasklist_lock);
368 * Ugh. To avoid negative return values, "getpriority()" will
369 * not return the normal nice-value, but a negated value that
370 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
371 * to stay compatible.
373 asmlinkage long sys_getpriority(int which, int who)
375 struct task_struct *g, *p;
378 struct user_struct *user;
379 long niceval, retval = -ESRCH;
381 if (which > 2 || which < 0)
384 read_lock(&tasklist_lock);
389 p = find_task_by_pid(who);
391 niceval = 20 - task_nice(p);
392 if (niceval > retval)
398 who = process_group(current);
399 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid) {
400 niceval = 20 - task_nice(p);
401 if (niceval > retval)
407 user = current->user;
409 user = find_user(vx_current_xid(), who);
416 niceval = 20 - task_nice(p);
417 if (niceval > retval)
420 while_each_thread(g, p);
422 free_uid(user); /* for find_user() */
426 read_unlock(&tasklist_lock);
432 * vshelper path is set via /proc/sys
433 * invoked by vserver sys_reboot(), with
434 * the following arguments
436 * argv [0] = vshelper_path;
437 * argv [1] = action: "restart", "halt", "poweroff", ...
438 * argv [2] = context identifier
439 * argv [3] = additional argument (restart2)
441 * envp [*] = type-specific parameters
443 char vshelper_path[255] = "/sbin/vshelper";
445 long vs_reboot(unsigned int cmd, void * arg)
447 char id_buf[8], cmd_buf[32];
448 char uid_buf[32], pid_buf[32];
451 char *argv[] = {vshelper_path, NULL, id_buf, NULL, 0};
452 char *envp[] = {"HOME=/", "TERM=linux",
453 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
454 uid_buf, pid_buf, cmd_buf, 0};
456 snprintf(id_buf, sizeof(id_buf)-1, "%d", vx_current_xid());
458 snprintf(cmd_buf, sizeof(cmd_buf)-1, "VS_CMD=%08x", cmd);
459 snprintf(uid_buf, sizeof(uid_buf)-1, "VS_UID=%d", current->uid);
460 snprintf(pid_buf, sizeof(pid_buf)-1, "VS_PID=%d", current->pid);
463 case LINUX_REBOOT_CMD_RESTART:
467 case LINUX_REBOOT_CMD_HALT:
471 case LINUX_REBOOT_CMD_POWER_OFF:
472 argv[1] = "poweroff";
475 case LINUX_REBOOT_CMD_SW_SUSPEND:
479 case LINUX_REBOOT_CMD_RESTART2:
480 if (strncpy_from_user(&buffer[0], (char *)arg, sizeof(buffer) - 1) < 0)
484 argv[1] = "restart2";
488 /* maybe we should wait ? */
489 if (call_usermodehelper(*argv, argv, envp, 0)) {
491 "vs_reboot(): failed to exec (%s %s %s %s)\n",
492 vshelper_path, argv[1], argv[2], argv[3]);
499 * Reboot system call: for obvious reasons only root may call it,
500 * and even root needs to set up some magic numbers in the registers
501 * so that some mistake won't make this reboot the whole machine.
502 * You can also set the meaning of the ctrl-alt-del-key here.
504 * reboot doesn't sync: do that yourself before calling this.
506 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
510 /* We only trust the superuser with rebooting the system. */
511 if (!capable(CAP_SYS_BOOT))
514 /* For safety, we require "magic" arguments. */
515 if (magic1 != LINUX_REBOOT_MAGIC1 ||
516 (magic2 != LINUX_REBOOT_MAGIC2 &&
517 magic2 != LINUX_REBOOT_MAGIC2A &&
518 magic2 != LINUX_REBOOT_MAGIC2B &&
519 magic2 != LINUX_REBOOT_MAGIC2C))
522 if (!vx_check(0, VX_ADMIN|VX_WATCH))
523 return vs_reboot(cmd, arg);
527 case LINUX_REBOOT_CMD_RESTART:
528 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
529 system_state = SYSTEM_RESTART;
531 printk(KERN_EMERG "Restarting system.\n");
532 machine_restart(NULL);
535 case LINUX_REBOOT_CMD_CAD_ON:
539 case LINUX_REBOOT_CMD_CAD_OFF:
543 case LINUX_REBOOT_CMD_HALT:
544 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
545 system_state = SYSTEM_HALT;
547 printk(KERN_EMERG "System halted.\n");
553 case LINUX_REBOOT_CMD_POWER_OFF:
554 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
555 system_state = SYSTEM_POWER_OFF;
557 printk(KERN_EMERG "Power down.\n");
563 case LINUX_REBOOT_CMD_RESTART2:
564 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
568 buffer[sizeof(buffer) - 1] = '\0';
570 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
571 system_state = SYSTEM_RESTART;
573 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
574 machine_restart(buffer);
577 #ifdef CONFIG_SOFTWARE_SUSPEND
578 case LINUX_REBOOT_CMD_SW_SUSPEND:
580 int ret = software_suspend();
594 static void deferred_cad(void *dummy)
596 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
597 machine_restart(NULL);
601 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
602 * As it's called within an interrupt, it may NOT sync: the only choice
603 * is whether to reboot at once, or just ignore the ctrl-alt-del.
605 void ctrl_alt_del(void)
607 static DECLARE_WORK(cad_work, deferred_cad, NULL);
610 schedule_work(&cad_work);
612 kill_proc(cad_pid, SIGINT, 1);
617 * Unprivileged users may change the real gid to the effective gid
618 * or vice versa. (BSD-style)
620 * If you set the real gid at all, or set the effective gid to a value not
621 * equal to the real gid, then the saved gid is set to the new effective gid.
623 * This makes it possible for a setgid program to completely drop its
624 * privileges, which is often a useful assertion to make when you are doing
625 * a security audit over a program.
627 * The general idea is that a program which uses just setregid() will be
628 * 100% compatible with BSD. A program which uses just setgid() will be
629 * 100% compatible with POSIX with saved IDs.
631 * SMP: There are not races, the GIDs are checked only by filesystem
632 * operations (as far as semantic preservation is concerned).
634 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
636 int old_rgid = current->gid;
637 int old_egid = current->egid;
638 int new_rgid = old_rgid;
639 int new_egid = old_egid;
642 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
646 if (rgid != (gid_t) -1) {
647 if ((old_rgid == rgid) ||
648 (current->egid==rgid) ||
654 if (egid != (gid_t) -1) {
655 if ((old_rgid == egid) ||
656 (current->egid == egid) ||
657 (current->sgid == egid) ||
664 if (new_egid != old_egid)
666 current->mm->dumpable = 0;
669 if (rgid != (gid_t) -1 ||
670 (egid != (gid_t) -1 && egid != old_rgid))
671 current->sgid = new_egid;
672 current->fsgid = new_egid;
673 current->egid = new_egid;
674 current->gid = new_rgid;
682 * setgid() is implemented like SysV w/ SAVED_IDS
684 * SMP: Same implicit races as above.
686 asmlinkage long sys_setgid(gid_t gid)
688 int old_egid = current->egid;
691 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
695 if (capable(CAP_SETGID))
699 current->mm->dumpable=0;
702 current->gid = current->egid = current->sgid = current->fsgid = gid;
704 else if ((gid == current->gid) || (gid == current->sgid))
708 current->mm->dumpable=0;
711 current->egid = current->fsgid = gid;
721 static int set_user(uid_t new_ruid, int dumpclear)
723 struct user_struct *new_user;
725 new_user = alloc_uid(vx_current_xid(), new_ruid);
729 if (atomic_read(&new_user->processes) >=
730 current->rlim[RLIMIT_NPROC].rlim_cur &&
731 new_user != &root_user) {
736 switch_uid(new_user);
740 current->mm->dumpable = 0;
743 current->uid = new_ruid;
748 * Unprivileged users may change the real uid to the effective uid
749 * or vice versa. (BSD-style)
751 * If you set the real uid at all, or set the effective uid to a value not
752 * equal to the real uid, then the saved uid is set to the new effective uid.
754 * This makes it possible for a setuid program to completely drop its
755 * privileges, which is often a useful assertion to make when you are doing
756 * a security audit over a program.
758 * The general idea is that a program which uses just setreuid() will be
759 * 100% compatible with BSD. A program which uses just setuid() will be
760 * 100% compatible with POSIX with saved IDs.
762 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
764 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
767 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
771 new_ruid = old_ruid = current->uid;
772 new_euid = old_euid = current->euid;
773 old_suid = current->suid;
775 if (ruid != (uid_t) -1) {
777 if ((old_ruid != ruid) &&
778 (current->euid != ruid) &&
779 !capable(CAP_SETUID))
783 if (euid != (uid_t) -1) {
785 if ((old_ruid != euid) &&
786 (current->euid != euid) &&
787 (current->suid != euid) &&
788 !capable(CAP_SETUID))
792 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
795 if (new_euid != old_euid)
797 current->mm->dumpable=0;
800 current->fsuid = current->euid = new_euid;
801 if (ruid != (uid_t) -1 ||
802 (euid != (uid_t) -1 && euid != old_ruid))
803 current->suid = current->euid;
804 current->fsuid = current->euid;
808 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
814 * setuid() is implemented like SysV with SAVED_IDS
816 * Note that SAVED_ID's is deficient in that a setuid root program
817 * like sendmail, for example, cannot set its uid to be a normal
818 * user and then switch back, because if you're root, setuid() sets
819 * the saved uid too. If you don't like this, blame the bright people
820 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
821 * will allow a root program to temporarily drop privileges and be able to
822 * regain them by swapping the real and effective uid.
824 asmlinkage long sys_setuid(uid_t uid)
826 int old_euid = current->euid;
827 int old_ruid, old_suid, new_ruid, new_suid;
830 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
834 old_ruid = new_ruid = current->uid;
835 old_suid = current->suid;
838 if (capable(CAP_SETUID)) {
839 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
842 } else if ((uid != current->uid) && (uid != new_suid))
847 current->mm->dumpable = 0;
850 current->fsuid = current->euid = uid;
851 current->suid = new_suid;
855 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
860 * This function implements a generic ability to update ruid, euid,
861 * and suid. This allows you to implement the 4.4 compatible seteuid().
863 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
865 int old_ruid = current->uid;
866 int old_euid = current->euid;
867 int old_suid = current->suid;
870 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
874 if (!capable(CAP_SETUID)) {
875 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
876 (ruid != current->euid) && (ruid != current->suid))
878 if ((euid != (uid_t) -1) && (euid != current->uid) &&
879 (euid != current->euid) && (euid != current->suid))
881 if ((suid != (uid_t) -1) && (suid != current->uid) &&
882 (suid != current->euid) && (suid != current->suid))
885 if (ruid != (uid_t) -1) {
886 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
889 if (euid != (uid_t) -1) {
890 if (euid != current->euid)
892 current->mm->dumpable = 0;
895 current->euid = euid;
897 current->fsuid = current->euid;
898 if (suid != (uid_t) -1)
899 current->suid = suid;
903 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
906 asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid)
910 if (!(retval = put_user(current->uid, ruid)) &&
911 !(retval = put_user(current->euid, euid)))
912 retval = put_user(current->suid, suid);
918 * Same as above, but for rgid, egid, sgid.
920 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
924 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
928 if (!capable(CAP_SETGID)) {
929 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
930 (rgid != current->egid) && (rgid != current->sgid))
932 if ((egid != (gid_t) -1) && (egid != current->gid) &&
933 (egid != current->egid) && (egid != current->sgid))
935 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
936 (sgid != current->egid) && (sgid != current->sgid))
939 if (egid != (gid_t) -1) {
940 if (egid != current->egid)
942 current->mm->dumpable = 0;
945 current->egid = egid;
947 current->fsgid = current->egid;
948 if (rgid != (gid_t) -1)
950 if (sgid != (gid_t) -1)
951 current->sgid = sgid;
958 asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid)
962 if (!(retval = put_user(current->gid, rgid)) &&
963 !(retval = put_user(current->egid, egid)))
964 retval = put_user(current->sgid, sgid);
971 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
972 * is used for "access()" and for the NFS daemon (letting nfsd stay at
973 * whatever uid it wants to). It normally shadows "euid", except when
974 * explicitly set by setfsuid() or for access..
976 asmlinkage long sys_setfsuid(uid_t uid)
980 old_fsuid = current->fsuid;
981 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
984 if (uid == current->uid || uid == current->euid ||
985 uid == current->suid || uid == current->fsuid ||
988 if (uid != old_fsuid)
990 current->mm->dumpable = 0;
993 current->fsuid = uid;
996 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
1002 * Samma på svenska..
1004 asmlinkage long sys_setfsgid(gid_t gid)
1008 old_fsgid = current->fsgid;
1009 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
1012 if (gid == current->gid || gid == current->egid ||
1013 gid == current->sgid || gid == current->fsgid ||
1014 capable(CAP_SETGID))
1016 if (gid != old_fsgid)
1018 current->mm->dumpable = 0;
1021 current->fsgid = gid;
1026 asmlinkage long sys_times(struct tms __user * tbuf)
1029 * In the SMP world we might just be unlucky and have one of
1030 * the times increment as we use it. Since the value is an
1031 * atomically safe type this is just fine. Conceptually its
1032 * as if the syscall took an instant longer to occur.
1036 tmp.tms_utime = jiffies_to_clock_t(current->utime);
1037 tmp.tms_stime = jiffies_to_clock_t(current->stime);
1038 tmp.tms_cutime = jiffies_to_clock_t(current->cutime);
1039 tmp.tms_cstime = jiffies_to_clock_t(current->cstime);
1040 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1043 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1047 * This needs some heavy checking ...
1048 * I just haven't the stomach for it. I also don't fully
1049 * understand sessions/pgrp etc. Let somebody who does explain it.
1051 * OK, I think I have the protection semantics right.... this is really
1052 * only important on a multi-user system anyway, to make sure one user
1053 * can't send a signal to a process owned by another. -TYT, 12/12/91
1055 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1059 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1061 struct task_struct *p;
1071 /* From this point forward we keep holding onto the tasklist lock
1072 * so that our parent does not change from under us. -DaveM
1074 write_lock_irq(&tasklist_lock);
1077 p = find_task_by_pid(pid);
1082 if (!thread_group_leader(p))
1085 if (p->parent == current || p->real_parent == current) {
1087 if (p->signal->session != current->signal->session)
1099 if (p->signal->leader)
1103 struct task_struct *p;
1105 struct list_head *l;
1107 for_each_task_pid(pgid, PIDTYPE_PGID, p, l, pid)
1108 if (p->signal->session == current->signal->session)
1114 err = security_task_setpgid(p, pgid);
1118 if (process_group(p) != pgid) {
1119 detach_pid(p, PIDTYPE_PGID);
1120 p->signal->pgrp = pgid;
1121 attach_pid(p, PIDTYPE_PGID, pgid);
1126 /* All paths lead to here, thus we are safe. -DaveM */
1127 write_unlock_irq(&tasklist_lock);
1131 asmlinkage long sys_getpgid(pid_t pid)
1134 return process_group(current);
1137 struct task_struct *p;
1139 read_lock(&tasklist_lock);
1140 p = find_task_by_pid(pid);
1144 retval = security_task_getpgid(p);
1146 retval = process_group(p);
1148 read_unlock(&tasklist_lock);
1153 #ifdef __ARCH_WANT_SYS_GETPGRP
1155 asmlinkage long sys_getpgrp(void)
1157 /* SMP - assuming writes are word atomic this is fine */
1158 return process_group(current);
1163 asmlinkage long sys_getsid(pid_t pid)
1166 return current->signal->session;
1169 struct task_struct *p;
1171 read_lock(&tasklist_lock);
1172 p = find_task_by_pid(pid);
1176 retval = security_task_getsid(p);
1178 retval = p->signal->session;
1180 read_unlock(&tasklist_lock);
1185 asmlinkage long sys_setsid(void)
1190 if (!thread_group_leader(current))
1193 write_lock_irq(&tasklist_lock);
1195 pid = find_pid(PIDTYPE_PGID, current->pid);
1199 current->signal->leader = 1;
1200 __set_special_pids(current->pid, current->pid);
1201 current->signal->tty = NULL;
1202 current->signal->tty_old_pgrp = 0;
1203 err = process_group(current);
1205 write_unlock_irq(&tasklist_lock);
1210 * Supplementary group IDs
1213 /* init to 2 - one for init_task, one to ensure it is never freed */
1214 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1216 struct group_info *groups_alloc(int gidsetsize)
1218 struct group_info *group_info;
1222 nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK;
1223 /* Make sure we always allocate at least one indirect block pointer */
1224 nblocks = nblocks ? : 1;
1225 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER);
1228 group_info->ngroups = gidsetsize;
1229 group_info->nblocks = nblocks;
1230 atomic_set(&group_info->usage, 1);
1232 if (gidsetsize <= NGROUPS_SMALL) {
1233 group_info->blocks[0] = group_info->small_block;
1235 for (i = 0; i < nblocks; i++) {
1237 b = (void *)__get_free_page(GFP_USER);
1239 goto out_undo_partial_alloc;
1240 group_info->blocks[i] = b;
1245 out_undo_partial_alloc:
1247 free_page((unsigned long)group_info->blocks[i]);
1253 EXPORT_SYMBOL(groups_alloc);
1255 void groups_free(struct group_info *group_info)
1257 if (group_info->blocks[0] != group_info->small_block) {
1259 for (i = 0; i < group_info->nblocks; i++)
1260 free_page((unsigned long)group_info->blocks[i]);
1265 EXPORT_SYMBOL(groups_free);
1267 /* export the group_info to a user-space array */
1268 static int groups_to_user(gid_t __user *grouplist,
1269 struct group_info *group_info)
1272 int count = group_info->ngroups;
1274 for (i = 0; i < group_info->nblocks; i++) {
1275 int cp_count = min(NGROUPS_PER_BLOCK, count);
1276 int off = i * NGROUPS_PER_BLOCK;
1277 int len = cp_count * sizeof(*grouplist);
1279 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1287 /* fill a group_info from a user-space array - it must be allocated already */
1288 static int groups_from_user(struct group_info *group_info,
1289 gid_t __user *grouplist)
1292 int count = group_info->ngroups;
1294 for (i = 0; i < group_info->nblocks; i++) {
1295 int cp_count = min(NGROUPS_PER_BLOCK, count);
1296 int off = i * NGROUPS_PER_BLOCK;
1297 int len = cp_count * sizeof(*grouplist);
1299 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1307 /* a simple shell-metzner sort */
1308 static void groups_sort(struct group_info *group_info)
1310 int base, max, stride;
1311 int gidsetsize = group_info->ngroups;
1313 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1318 max = gidsetsize - stride;
1319 for (base = 0; base < max; base++) {
1321 int right = left + stride;
1322 gid_t tmp = GROUP_AT(group_info, right);
1324 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1325 GROUP_AT(group_info, right) =
1326 GROUP_AT(group_info, left);
1330 GROUP_AT(group_info, right) = tmp;
1336 /* a simple bsearch */
1337 static int groups_search(struct group_info *group_info, gid_t grp)
1345 right = group_info->ngroups;
1346 while (left < right) {
1347 int mid = (left+right)/2;
1348 int cmp = grp - GROUP_AT(group_info, mid);
1359 /* validate and set current->group_info */
1360 int set_current_groups(struct group_info *group_info)
1363 struct group_info *old_info;
1365 retval = security_task_setgroups(group_info);
1369 groups_sort(group_info);
1370 get_group_info(group_info);
1373 old_info = current->group_info;
1374 current->group_info = group_info;
1375 task_unlock(current);
1377 put_group_info(old_info);
1382 EXPORT_SYMBOL(set_current_groups);
1384 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1389 * SMP: Nobody else can change our grouplist. Thus we are
1396 /* no need to grab task_lock here; it cannot change */
1397 get_group_info(current->group_info);
1398 i = current->group_info->ngroups;
1400 if (i > gidsetsize) {
1404 if (groups_to_user(grouplist, current->group_info)) {
1410 put_group_info(current->group_info);
1415 * SMP: Our groups are copy-on-write. We can set them safely
1416 * without another task interfering.
1419 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1421 struct group_info *group_info;
1424 if (!capable(CAP_SETGID))
1426 if ((unsigned)gidsetsize > NGROUPS_MAX)
1429 group_info = groups_alloc(gidsetsize);
1432 retval = groups_from_user(group_info, grouplist);
1434 put_group_info(group_info);
1438 retval = set_current_groups(group_info);
1439 put_group_info(group_info);
1445 * Check whether we're fsgid/egid or in the supplemental group..
1447 int in_group_p(gid_t grp)
1450 if (grp != current->fsgid) {
1451 get_group_info(current->group_info);
1452 retval = groups_search(current->group_info, grp);
1453 put_group_info(current->group_info);
1458 EXPORT_SYMBOL(in_group_p);
1460 int in_egroup_p(gid_t grp)
1463 if (grp != current->egid) {
1464 get_group_info(current->group_info);
1465 retval = groups_search(current->group_info, grp);
1466 put_group_info(current->group_info);
1471 EXPORT_SYMBOL(in_egroup_p);
1473 DECLARE_RWSEM(uts_sem);
1475 EXPORT_SYMBOL(uts_sem);
1477 asmlinkage long sys_newuname(struct new_utsname __user * name)
1481 down_read(&uts_sem);
1482 if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1488 asmlinkage long sys_sethostname(char __user *name, int len)
1491 char tmp[__NEW_UTS_LEN];
1493 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1495 if (len < 0 || len > __NEW_UTS_LEN)
1497 down_write(&uts_sem);
1499 if (!copy_from_user(tmp, name, len)) {
1500 char *ptr = vx_new_uts(nodename);
1502 memcpy(ptr, tmp, len);
1510 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1512 asmlinkage long sys_gethostname(char __user *name, int len)
1519 down_read(&uts_sem);
1520 ptr = vx_new_uts(nodename);
1521 i = 1 + strlen(ptr);
1525 if (copy_to_user(name, ptr, i))
1534 * Only setdomainname; getdomainname can be implemented by calling
1537 asmlinkage long sys_setdomainname(char __user *name, int len)
1540 char tmp[__NEW_UTS_LEN];
1542 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1544 if (len < 0 || len > __NEW_UTS_LEN)
1547 down_write(&uts_sem);
1549 if (!copy_from_user(tmp, name, len)) {
1550 char *ptr = vx_new_uts(domainname);
1552 memcpy(ptr, tmp, len);
1560 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1562 if (resource >= RLIM_NLIMITS)
1565 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1569 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1572 * Back compatibility for getrlimit. Needed for some apps.
1575 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1578 if (resource >= RLIM_NLIMITS)
1581 memcpy(&x, current->rlim + resource, sizeof(*rlim));
1582 if(x.rlim_cur > 0x7FFFFFFF)
1583 x.rlim_cur = 0x7FFFFFFF;
1584 if(x.rlim_max > 0x7FFFFFFF)
1585 x.rlim_max = 0x7FFFFFFF;
1586 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1591 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1593 struct rlimit new_rlim, *old_rlim;
1596 if (resource >= RLIM_NLIMITS)
1598 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1600 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1602 old_rlim = current->rlim + resource;
1603 if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1604 (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1605 !capable(CAP_SYS_RESOURCE) && vx_ccaps(VXC_SET_RLIMIT))
1607 if (resource == RLIMIT_NOFILE) {
1608 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1612 retval = security_task_setrlimit(resource, &new_rlim);
1616 *old_rlim = new_rlim;
1621 * It would make sense to put struct rusage in the task_struct,
1622 * except that would make the task_struct be *really big*. After
1623 * task_struct gets moved into malloc'ed memory, it would
1624 * make sense to do this. It will make moving the rest of the information
1625 * a lot simpler! (Which we're not doing right now because we're not
1626 * measuring them yet).
1628 * This is SMP safe. Either we are called from sys_getrusage on ourselves
1629 * below (we know we aren't going to exit/disappear and only we change our
1630 * rusage counters), or we are called from wait4() on a process which is
1631 * either stopped or zombied. In the zombied case the task won't get
1632 * reaped till shortly after the call to getrusage(), in both cases the
1633 * task being examined is in a frozen state so the counters won't change.
1635 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1639 memset((char *) &r, 0, sizeof(r));
1642 jiffies_to_timeval(p->utime, &r.ru_utime);
1643 jiffies_to_timeval(p->stime, &r.ru_stime);
1644 r.ru_nvcsw = p->nvcsw;
1645 r.ru_nivcsw = p->nivcsw;
1646 r.ru_minflt = p->min_flt;
1647 r.ru_majflt = p->maj_flt;
1649 case RUSAGE_CHILDREN:
1650 jiffies_to_timeval(p->cutime, &r.ru_utime);
1651 jiffies_to_timeval(p->cstime, &r.ru_stime);
1652 r.ru_nvcsw = p->cnvcsw;
1653 r.ru_nivcsw = p->cnivcsw;
1654 r.ru_minflt = p->cmin_flt;
1655 r.ru_majflt = p->cmaj_flt;
1658 jiffies_to_timeval(p->utime + p->cutime, &r.ru_utime);
1659 jiffies_to_timeval(p->stime + p->cstime, &r.ru_stime);
1660 r.ru_nvcsw = p->nvcsw + p->cnvcsw;
1661 r.ru_nivcsw = p->nivcsw + p->cnivcsw;
1662 r.ru_minflt = p->min_flt + p->cmin_flt;
1663 r.ru_majflt = p->maj_flt + p->cmaj_flt;
1666 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1669 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1671 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1673 return getrusage(current, who, ru);
1676 asmlinkage long sys_umask(int mask)
1678 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1682 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1683 unsigned long arg4, unsigned long arg5)
1688 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1693 case PR_SET_PDEATHSIG:
1695 if (sig < 0 || sig > _NSIG) {
1699 current->pdeath_signal = sig;
1701 case PR_GET_PDEATHSIG:
1702 error = put_user(current->pdeath_signal, (int __user *)arg2);
1704 case PR_GET_DUMPABLE:
1705 if (current->mm->dumpable)
1708 case PR_SET_DUMPABLE:
1709 if (arg2 != 0 && arg2 != 1) {
1713 current->mm->dumpable = arg2;
1716 case PR_SET_UNALIGN:
1717 error = SET_UNALIGN_CTL(current, arg2);
1719 case PR_GET_UNALIGN:
1720 error = GET_UNALIGN_CTL(current, arg2);
1723 error = SET_FPEMU_CTL(current, arg2);
1726 error = GET_FPEMU_CTL(current, arg2);
1729 error = SET_FPEXC_CTL(current, arg2);
1732 error = GET_FPEXC_CTL(current, arg2);
1735 error = PR_TIMING_STATISTICAL;
1738 if (arg2 == PR_TIMING_STATISTICAL)
1744 case PR_GET_KEEPCAPS:
1745 if (current->keep_capabilities)
1748 case PR_SET_KEEPCAPS:
1749 if (arg2 != 0 && arg2 != 1) {
1753 current->keep_capabilities = arg2;