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)
276 /* arch-specific weak syscall entries */
277 cond_syscall(sys_pciconfig_read)
278 cond_syscall(sys_pciconfig_write)
279 cond_syscall(sys_pciconfig_iobase)
281 static int set_one_prio(struct task_struct *p, int niceval, int error)
285 if (p->uid != current->euid &&
286 p->uid != current->uid && !capable(CAP_SYS_NICE)) {
290 if (niceval < task_nice(p) && !capable(CAP_SYS_NICE)) {
294 no_nice = security_task_setnice(p, niceval);
301 set_user_nice(p, niceval);
306 asmlinkage long sys_setpriority(int which, int who, int niceval)
308 struct task_struct *g, *p;
309 struct user_struct *user;
314 if (which > 2 || which < 0)
317 /* normalize: avoid signed division (rounding problems) */
324 read_lock(&tasklist_lock);
329 p = find_task_by_pid(who);
331 error = set_one_prio(p, niceval, error);
335 who = process_group(current);
336 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid)
337 error = set_one_prio(p, niceval, error);
341 user = current->user;
343 user = find_user(vx_current_xid(), who);
350 error = set_one_prio(p, niceval, error);
351 while_each_thread(g, p);
355 read_unlock(&tasklist_lock);
361 * Ugh. To avoid negative return values, "getpriority()" will
362 * not return the normal nice-value, but a negated value that
363 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
364 * to stay compatible.
366 asmlinkage long sys_getpriority(int which, int who)
368 struct task_struct *g, *p;
371 struct user_struct *user;
372 long niceval, retval = -ESRCH;
374 if (which > 2 || which < 0)
377 read_lock(&tasklist_lock);
382 p = find_task_by_pid(who);
384 niceval = 20 - task_nice(p);
385 if (niceval > retval)
391 who = process_group(current);
392 for_each_task_pid(who, PIDTYPE_PGID, p, l, pid) {
393 niceval = 20 - task_nice(p);
394 if (niceval > retval)
400 user = current->user;
402 user = find_user(vx_current_xid(), who);
409 niceval = 20 - task_nice(p);
410 if (niceval > retval)
413 while_each_thread(g, p);
417 read_unlock(&tasklist_lock);
423 * vshelper path is set via /proc/sys
424 * invoked by vserver sys_reboot(), with
425 * the following arguments
427 * argv [0] = vshelper_path;
428 * argv [1] = action: "restart", "halt", "poweroff", ...
429 * argv [2] = context identifier
430 * argv [3] = additional argument (restart2)
432 * envp [*] = type-specific parameters
434 char vshelper_path[255] = "/sbin/vshelper";
436 long vs_reboot(unsigned int cmd, void * arg)
438 char id_buf[8], cmd_buf[32];
439 char uid_buf[32], pid_buf[32];
442 char *argv[] = {vshelper_path, NULL, id_buf, NULL, 0};
443 char *envp[] = {"HOME=/", "TERM=linux",
444 "PATH=/sbin:/usr/sbin:/bin:/usr/bin",
445 uid_buf, pid_buf, cmd_buf, 0};
447 snprintf(id_buf, sizeof(id_buf)-1, "%d", vx_current_xid());
449 snprintf(cmd_buf, sizeof(cmd_buf)-1, "VS_CMD=%08x", cmd);
450 snprintf(uid_buf, sizeof(uid_buf)-1, "VS_UID=%d", current->uid);
451 snprintf(pid_buf, sizeof(pid_buf)-1, "VS_PID=%d", current->pid);
454 case LINUX_REBOOT_CMD_RESTART:
458 case LINUX_REBOOT_CMD_HALT:
462 case LINUX_REBOOT_CMD_POWER_OFF:
463 argv[1] = "poweroff";
466 case LINUX_REBOOT_CMD_SW_SUSPEND:
470 case LINUX_REBOOT_CMD_RESTART2:
471 if (strncpy_from_user(&buffer[0], (char *)arg, sizeof(buffer) - 1) < 0)
475 argv[1] = "restart2";
479 /* maybe we should wait ? */
480 if (call_usermodehelper(*argv, argv, envp, 0)) {
482 "vs_reboot(): failed to exec (%s %s %s %s)\n",
483 vshelper_path, argv[1], argv[2], argv[3]);
490 * Reboot system call: for obvious reasons only root may call it,
491 * and even root needs to set up some magic numbers in the registers
492 * so that some mistake won't make this reboot the whole machine.
493 * You can also set the meaning of the ctrl-alt-del-key here.
495 * reboot doesn't sync: do that yourself before calling this.
497 asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg)
501 /* We only trust the superuser with rebooting the system. */
502 if (!capable(CAP_SYS_BOOT))
505 /* For safety, we require "magic" arguments. */
506 if (magic1 != LINUX_REBOOT_MAGIC1 ||
507 (magic2 != LINUX_REBOOT_MAGIC2 &&
508 magic2 != LINUX_REBOOT_MAGIC2A &&
509 magic2 != LINUX_REBOOT_MAGIC2B &&
510 magic2 != LINUX_REBOOT_MAGIC2C))
513 if (!vx_check(0, VX_ADMIN|VX_WATCH))
514 return vs_reboot(cmd, arg);
518 case LINUX_REBOOT_CMD_RESTART:
519 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
520 system_state = SYSTEM_SHUTDOWN;
522 printk(KERN_EMERG "Restarting system.\n");
523 machine_restart(NULL);
526 case LINUX_REBOOT_CMD_CAD_ON:
530 case LINUX_REBOOT_CMD_CAD_OFF:
534 case LINUX_REBOOT_CMD_HALT:
535 notifier_call_chain(&reboot_notifier_list, SYS_HALT, NULL);
536 system_state = SYSTEM_SHUTDOWN;
538 printk(KERN_EMERG "System halted.\n");
544 case LINUX_REBOOT_CMD_POWER_OFF:
545 notifier_call_chain(&reboot_notifier_list, SYS_POWER_OFF, NULL);
546 system_state = SYSTEM_SHUTDOWN;
548 printk(KERN_EMERG "Power down.\n");
554 case LINUX_REBOOT_CMD_RESTART2:
555 if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) {
559 buffer[sizeof(buffer) - 1] = '\0';
561 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, buffer);
562 system_state = SYSTEM_SHUTDOWN;
564 printk(KERN_EMERG "Restarting system with command '%s'.\n", buffer);
565 machine_restart(buffer);
568 #ifdef CONFIG_SOFTWARE_SUSPEND
569 case LINUX_REBOOT_CMD_SW_SUSPEND:
571 int ret = software_suspend();
585 static void deferred_cad(void *dummy)
587 notifier_call_chain(&reboot_notifier_list, SYS_RESTART, NULL);
588 machine_restart(NULL);
592 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
593 * As it's called within an interrupt, it may NOT sync: the only choice
594 * is whether to reboot at once, or just ignore the ctrl-alt-del.
596 void ctrl_alt_del(void)
598 static DECLARE_WORK(cad_work, deferred_cad, NULL);
601 schedule_work(&cad_work);
603 kill_proc(cad_pid, SIGINT, 1);
608 * Unprivileged users may change the real gid to the effective gid
609 * or vice versa. (BSD-style)
611 * If you set the real gid at all, or set the effective gid to a value not
612 * equal to the real gid, then the saved gid is set to the new effective gid.
614 * This makes it possible for a setgid program to completely drop its
615 * privileges, which is often a useful assertion to make when you are doing
616 * a security audit over a program.
618 * The general idea is that a program which uses just setregid() will be
619 * 100% compatible with BSD. A program which uses just setgid() will be
620 * 100% compatible with POSIX with saved IDs.
622 * SMP: There are not races, the GIDs are checked only by filesystem
623 * operations (as far as semantic preservation is concerned).
625 asmlinkage long sys_setregid(gid_t rgid, gid_t egid)
627 int old_rgid = current->gid;
628 int old_egid = current->egid;
629 int new_rgid = old_rgid;
630 int new_egid = old_egid;
633 retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE);
637 if (rgid != (gid_t) -1) {
638 if ((old_rgid == rgid) ||
639 (current->egid==rgid) ||
645 if (egid != (gid_t) -1) {
646 if ((old_rgid == egid) ||
647 (current->egid == egid) ||
648 (current->sgid == egid) ||
655 if (new_egid != old_egid)
657 current->mm->dumpable = 0;
660 if (rgid != (gid_t) -1 ||
661 (egid != (gid_t) -1 && egid != old_rgid))
662 current->sgid = new_egid;
663 current->fsgid = new_egid;
664 current->egid = new_egid;
665 current->gid = new_rgid;
670 * setgid() is implemented like SysV w/ SAVED_IDS
672 * SMP: Same implicit races as above.
674 asmlinkage long sys_setgid(gid_t gid)
676 int old_egid = current->egid;
679 retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID);
683 if (capable(CAP_SETGID))
687 current->mm->dumpable=0;
690 current->gid = current->egid = current->sgid = current->fsgid = gid;
692 else if ((gid == current->gid) || (gid == current->sgid))
696 current->mm->dumpable=0;
699 current->egid = current->fsgid = gid;
706 static int set_user(uid_t new_ruid, int dumpclear)
708 struct user_struct *new_user;
710 new_user = alloc_uid(vx_current_xid(), new_ruid);
714 if (atomic_read(&new_user->processes) >=
715 current->rlim[RLIMIT_NPROC].rlim_cur &&
716 new_user != &root_user) {
721 switch_uid(new_user);
725 current->mm->dumpable = 0;
728 current->uid = new_ruid;
733 * Unprivileged users may change the real uid to the effective uid
734 * or vice versa. (BSD-style)
736 * If you set the real uid at all, or set the effective uid to a value not
737 * equal to the real uid, then the saved uid is set to the new effective uid.
739 * This makes it possible for a setuid program to completely drop its
740 * privileges, which is often a useful assertion to make when you are doing
741 * a security audit over a program.
743 * The general idea is that a program which uses just setreuid() will be
744 * 100% compatible with BSD. A program which uses just setuid() will be
745 * 100% compatible with POSIX with saved IDs.
747 asmlinkage long sys_setreuid(uid_t ruid, uid_t euid)
749 int old_ruid, old_euid, old_suid, new_ruid, new_euid;
752 retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE);
756 new_ruid = old_ruid = current->uid;
757 new_euid = old_euid = current->euid;
758 old_suid = current->suid;
760 if (ruid != (uid_t) -1) {
762 if ((old_ruid != ruid) &&
763 (current->euid != ruid) &&
764 !capable(CAP_SETUID))
768 if (euid != (uid_t) -1) {
770 if ((old_ruid != euid) &&
771 (current->euid != euid) &&
772 (current->suid != euid) &&
773 !capable(CAP_SETUID))
777 if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0)
780 if (new_euid != old_euid)
782 current->mm->dumpable=0;
785 current->fsuid = current->euid = new_euid;
786 if (ruid != (uid_t) -1 ||
787 (euid != (uid_t) -1 && euid != old_ruid))
788 current->suid = current->euid;
789 current->fsuid = current->euid;
791 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE);
797 * setuid() is implemented like SysV with SAVED_IDS
799 * Note that SAVED_ID's is deficient in that a setuid root program
800 * like sendmail, for example, cannot set its uid to be a normal
801 * user and then switch back, because if you're root, setuid() sets
802 * the saved uid too. If you don't like this, blame the bright people
803 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
804 * will allow a root program to temporarily drop privileges and be able to
805 * regain them by swapping the real and effective uid.
807 asmlinkage long sys_setuid(uid_t uid)
809 int old_euid = current->euid;
810 int old_ruid, old_suid, new_ruid, new_suid;
813 retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID);
817 old_ruid = new_ruid = current->uid;
818 old_suid = current->suid;
821 if (capable(CAP_SETUID)) {
822 if (uid != old_ruid && set_user(uid, old_euid != uid) < 0)
825 } else if ((uid != current->uid) && (uid != new_suid))
830 current->mm->dumpable = 0;
833 current->fsuid = current->euid = uid;
834 current->suid = new_suid;
836 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID);
841 * This function implements a generic ability to update ruid, euid,
842 * and suid. This allows you to implement the 4.4 compatible seteuid().
844 asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid)
846 int old_ruid = current->uid;
847 int old_euid = current->euid;
848 int old_suid = current->suid;
851 retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES);
855 if (!capable(CAP_SETUID)) {
856 if ((ruid != (uid_t) -1) && (ruid != current->uid) &&
857 (ruid != current->euid) && (ruid != current->suid))
859 if ((euid != (uid_t) -1) && (euid != current->uid) &&
860 (euid != current->euid) && (euid != current->suid))
862 if ((suid != (uid_t) -1) && (suid != current->uid) &&
863 (suid != current->euid) && (suid != current->suid))
866 if (ruid != (uid_t) -1) {
867 if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0)
870 if (euid != (uid_t) -1) {
871 if (euid != current->euid)
873 current->mm->dumpable = 0;
876 current->euid = euid;
878 current->fsuid = current->euid;
879 if (suid != (uid_t) -1)
880 current->suid = suid;
882 return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES);
885 asmlinkage long sys_getresuid(uid_t *ruid, uid_t *euid, uid_t *suid)
889 if (!(retval = put_user(current->uid, ruid)) &&
890 !(retval = put_user(current->euid, euid)))
891 retval = put_user(current->suid, suid);
897 * Same as above, but for rgid, egid, sgid.
899 asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid)
903 retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES);
907 if (!capable(CAP_SETGID)) {
908 if ((rgid != (gid_t) -1) && (rgid != current->gid) &&
909 (rgid != current->egid) && (rgid != current->sgid))
911 if ((egid != (gid_t) -1) && (egid != current->gid) &&
912 (egid != current->egid) && (egid != current->sgid))
914 if ((sgid != (gid_t) -1) && (sgid != current->gid) &&
915 (sgid != current->egid) && (sgid != current->sgid))
918 if (egid != (gid_t) -1) {
919 if (egid != current->egid)
921 current->mm->dumpable = 0;
924 current->egid = egid;
926 current->fsgid = current->egid;
927 if (rgid != (gid_t) -1)
929 if (sgid != (gid_t) -1)
930 current->sgid = sgid;
934 asmlinkage long sys_getresgid(gid_t *rgid, gid_t *egid, gid_t *sgid)
938 if (!(retval = put_user(current->gid, rgid)) &&
939 !(retval = put_user(current->egid, egid)))
940 retval = put_user(current->sgid, sgid);
947 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
948 * is used for "access()" and for the NFS daemon (letting nfsd stay at
949 * whatever uid it wants to). It normally shadows "euid", except when
950 * explicitly set by setfsuid() or for access..
952 asmlinkage long sys_setfsuid(uid_t uid)
956 old_fsuid = current->fsuid;
957 if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS))
960 if (uid == current->uid || uid == current->euid ||
961 uid == current->suid || uid == current->fsuid ||
964 if (uid != old_fsuid)
966 current->mm->dumpable = 0;
969 current->fsuid = uid;
972 security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS);
978 * Samma på svenska..
980 asmlinkage long sys_setfsgid(gid_t gid)
984 old_fsgid = current->fsgid;
985 if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS))
988 if (gid == current->gid || gid == current->egid ||
989 gid == current->sgid || gid == current->fsgid ||
992 if (gid != old_fsgid)
994 current->mm->dumpable = 0;
997 current->fsgid = gid;
1002 asmlinkage long sys_times(struct tms __user * tbuf)
1005 * In the SMP world we might just be unlucky and have one of
1006 * the times increment as we use it. Since the value is an
1007 * atomically safe type this is just fine. Conceptually its
1008 * as if the syscall took an instant longer to occur.
1012 tmp.tms_utime = jiffies_to_clock_t(current->utime);
1013 tmp.tms_stime = jiffies_to_clock_t(current->stime);
1014 tmp.tms_cutime = jiffies_to_clock_t(current->cutime);
1015 tmp.tms_cstime = jiffies_to_clock_t(current->cstime);
1016 if (copy_to_user(tbuf, &tmp, sizeof(struct tms)))
1019 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1023 * This needs some heavy checking ...
1024 * I just haven't the stomach for it. I also don't fully
1025 * understand sessions/pgrp etc. Let somebody who does explain it.
1027 * OK, I think I have the protection semantics right.... this is really
1028 * only important on a multi-user system anyway, to make sure one user
1029 * can't send a signal to a process owned by another. -TYT, 12/12/91
1031 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1035 asmlinkage long sys_setpgid(pid_t pid, pid_t pgid)
1037 struct task_struct *p;
1047 /* From this point forward we keep holding onto the tasklist lock
1048 * so that our parent does not change from under us. -DaveM
1050 write_lock_irq(&tasklist_lock);
1053 p = find_task_by_pid(pid);
1058 if (!thread_group_leader(p))
1061 if (p->parent == current || p->real_parent == current) {
1063 if (p->signal->session != current->signal->session)
1075 if (p->signal->leader)
1079 struct task_struct *p;
1081 struct list_head *l;
1083 for_each_task_pid(pgid, PIDTYPE_PGID, p, l, pid)
1084 if (p->signal->session == current->signal->session)
1090 err = security_task_setpgid(p, pgid);
1094 if (process_group(p) != pgid) {
1095 detach_pid(p, PIDTYPE_PGID);
1096 p->signal->pgrp = pgid;
1097 attach_pid(p, PIDTYPE_PGID, pgid);
1102 /* All paths lead to here, thus we are safe. -DaveM */
1103 write_unlock_irq(&tasklist_lock);
1107 asmlinkage long sys_getpgid(pid_t pid)
1110 return process_group(current);
1113 struct task_struct *p;
1115 read_lock(&tasklist_lock);
1116 p = find_task_by_pid(pid);
1120 retval = security_task_getpgid(p);
1122 retval = process_group(p);
1124 read_unlock(&tasklist_lock);
1129 asmlinkage long sys_getpgrp(void)
1131 /* SMP - assuming writes are word atomic this is fine */
1132 return process_group(current);
1135 asmlinkage long sys_getsid(pid_t pid)
1138 return current->signal->session;
1141 struct task_struct *p;
1143 read_lock(&tasklist_lock);
1144 p = find_task_by_pid(pid);
1148 retval = security_task_getsid(p);
1150 retval = p->signal->session;
1152 read_unlock(&tasklist_lock);
1157 asmlinkage long sys_setsid(void)
1162 if (!thread_group_leader(current))
1165 write_lock_irq(&tasklist_lock);
1167 pid = find_pid(PIDTYPE_PGID, current->pid);
1171 current->signal->leader = 1;
1172 __set_special_pids(current->pid, current->pid);
1173 current->signal->tty = NULL;
1174 current->signal->tty_old_pgrp = 0;
1175 err = process_group(current);
1177 write_unlock_irq(&tasklist_lock);
1182 * Supplementary group IDs
1185 /* init to 2 - one for init_task, one to ensure it is never freed */
1186 struct group_info init_groups = { .usage = ATOMIC_INIT(2) };
1188 struct group_info *groups_alloc(int gidsetsize)
1190 struct group_info *group_info;
1194 nblocks = (gidsetsize/NGROUPS_PER_BLOCK) +
1195 (gidsetsize%NGROUPS_PER_BLOCK?1:0);
1196 group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *),
1200 group_info->ngroups = gidsetsize;
1201 group_info->nblocks = nblocks;
1202 atomic_set(&group_info->usage, 1);
1204 if (gidsetsize <= NGROUPS_SMALL) {
1205 group_info->blocks[0] = group_info->small_block;
1207 for (i = 0; i < nblocks; i++) {
1209 b = (void *)__get_free_page(GFP_USER);
1211 goto out_undo_partial_alloc;
1212 group_info->blocks[i] = b;
1217 out_undo_partial_alloc:
1219 free_page((unsigned long)group_info->blocks[i]);
1225 EXPORT_SYMBOL(groups_alloc);
1227 void groups_free(struct group_info *group_info)
1229 if (group_info->blocks[0] != group_info->small_block) {
1231 for (i = 0; i < group_info->nblocks; i++)
1232 free_page((unsigned long)group_info->blocks[i]);
1237 EXPORT_SYMBOL(groups_free);
1239 /* export the group_info to a user-space array */
1240 static int groups_to_user(gid_t __user *grouplist,
1241 struct group_info *group_info)
1244 int count = group_info->ngroups;
1246 for (i = 0; i < group_info->nblocks; i++) {
1247 int cp_count = min(NGROUPS_PER_BLOCK, count);
1248 int off = i * NGROUPS_PER_BLOCK;
1249 int len = cp_count * sizeof(*grouplist);
1251 if (copy_to_user(grouplist+off, group_info->blocks[i], len))
1259 /* fill a group_info from a user-space array - it must be allocated already */
1260 static int groups_from_user(struct group_info *group_info,
1261 gid_t __user *grouplist)
1264 int count = group_info->ngroups;
1266 for (i = 0; i < group_info->nblocks; i++) {
1267 int cp_count = min(NGROUPS_PER_BLOCK, count);
1268 int off = i * NGROUPS_PER_BLOCK;
1269 int len = cp_count * sizeof(*grouplist);
1271 if (copy_from_user(group_info->blocks[i], grouplist+off, len))
1279 /* a simple shell-metzner sort */
1280 static void groups_sort(struct group_info *group_info)
1282 int base, max, stride;
1283 int gidsetsize = group_info->ngroups;
1285 for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1)
1290 max = gidsetsize - stride;
1291 for (base = 0; base < max; base++) {
1293 int right = left + stride;
1294 gid_t tmp = GROUP_AT(group_info, right);
1296 while (left >= 0 && GROUP_AT(group_info, left) > tmp) {
1297 GROUP_AT(group_info, right) =
1298 GROUP_AT(group_info, left);
1302 GROUP_AT(group_info, right) = tmp;
1308 /* a simple bsearch */
1309 static int groups_search(struct group_info *group_info, gid_t grp)
1317 right = group_info->ngroups;
1318 while (left < right) {
1319 int mid = (left+right)/2;
1320 int cmp = grp - GROUP_AT(group_info, mid);
1331 /* validate and set current->group_info */
1332 int set_current_groups(struct group_info *group_info)
1335 struct group_info *old_info;
1337 retval = security_task_setgroups(group_info);
1341 groups_sort(group_info);
1342 get_group_info(group_info);
1343 old_info = current->group_info;
1344 current->group_info = group_info;
1345 put_group_info(old_info);
1350 EXPORT_SYMBOL(set_current_groups);
1352 asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist)
1357 * SMP: Nobody else can change our grouplist. Thus we are
1364 get_group_info(current->group_info);
1365 i = current->group_info->ngroups;
1367 if (i > gidsetsize) {
1371 if (groups_to_user(grouplist, current->group_info)) {
1377 put_group_info(current->group_info);
1382 * SMP: Our groups are copy-on-write. We can set them safely
1383 * without another task interfering.
1386 asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist)
1388 struct group_info *group_info;
1391 if (!capable(CAP_SETGID))
1393 if ((unsigned)gidsetsize > NGROUPS_MAX)
1396 group_info = groups_alloc(gidsetsize);
1399 retval = groups_from_user(group_info, grouplist);
1401 put_group_info(group_info);
1405 retval = set_current_groups(group_info);
1406 put_group_info(group_info);
1412 * Check whether we're fsgid/egid or in the supplemental group..
1414 int in_group_p(gid_t grp)
1417 if (grp != current->fsgid) {
1418 get_group_info(current->group_info);
1419 retval = groups_search(current->group_info, grp);
1420 put_group_info(current->group_info);
1425 EXPORT_SYMBOL(in_group_p);
1427 int in_egroup_p(gid_t grp)
1430 if (grp != current->egid) {
1431 get_group_info(current->group_info);
1432 retval = groups_search(current->group_info, grp);
1433 put_group_info(current->group_info);
1438 EXPORT_SYMBOL(in_egroup_p);
1440 DECLARE_RWSEM(uts_sem);
1442 EXPORT_SYMBOL(uts_sem);
1444 asmlinkage long sys_newuname(struct new_utsname __user * name)
1448 down_read(&uts_sem);
1449 if (copy_to_user(name, vx_new_utsname(), sizeof *name))
1455 asmlinkage long sys_sethostname(char __user *name, int len)
1458 char tmp[__NEW_UTS_LEN];
1460 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1462 if (len < 0 || len > __NEW_UTS_LEN)
1464 down_write(&uts_sem);
1466 if (!copy_from_user(tmp, name, len)) {
1467 char *ptr = vx_new_uts(nodename);
1469 memcpy(ptr, tmp, len);
1477 asmlinkage long sys_gethostname(char __user *name, int len)
1484 down_read(&uts_sem);
1485 ptr = vx_new_uts(nodename);
1486 i = 1 + strlen(ptr);
1490 if (copy_to_user(name, ptr, i))
1497 * Only setdomainname; getdomainname can be implemented by calling
1500 asmlinkage long sys_setdomainname(char __user *name, int len)
1503 char tmp[__NEW_UTS_LEN];
1505 if (!capable(CAP_SYS_ADMIN) && !vx_ccaps(VXC_SET_UTSNAME))
1507 if (len < 0 || len > __NEW_UTS_LEN)
1510 down_write(&uts_sem);
1512 if (!copy_from_user(tmp, name, len)) {
1513 char *ptr = vx_new_uts(domainname);
1515 memcpy(ptr, tmp, len);
1523 asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1525 if (resource >= RLIM_NLIMITS)
1528 return copy_to_user(rlim, current->rlim + resource, sizeof(*rlim))
1532 #if defined(COMPAT_RLIM_OLD_INFINITY) || !(defined(CONFIG_IA64) || defined(CONFIG_V850))
1535 * Back compatibility for getrlimit. Needed for some apps.
1538 asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim)
1541 if (resource >= RLIM_NLIMITS)
1544 memcpy(&x, current->rlim + resource, sizeof(*rlim));
1545 if(x.rlim_cur > 0x7FFFFFFF)
1546 x.rlim_cur = 0x7FFFFFFF;
1547 if(x.rlim_max > 0x7FFFFFFF)
1548 x.rlim_max = 0x7FFFFFFF;
1549 return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0;
1554 asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim)
1556 struct rlimit new_rlim, *old_rlim;
1559 if (resource >= RLIM_NLIMITS)
1561 if(copy_from_user(&new_rlim, rlim, sizeof(*rlim)))
1563 if (new_rlim.rlim_cur > new_rlim.rlim_max)
1565 old_rlim = current->rlim + resource;
1566 if (((new_rlim.rlim_cur > old_rlim->rlim_max) ||
1567 (new_rlim.rlim_max > old_rlim->rlim_max)) &&
1568 !capable(CAP_SYS_RESOURCE) && vx_ccaps(VXC_SET_RLIMIT))
1570 if (resource == RLIMIT_NOFILE) {
1571 if (new_rlim.rlim_cur > NR_OPEN || new_rlim.rlim_max > NR_OPEN)
1575 retval = security_task_setrlimit(resource, &new_rlim);
1579 *old_rlim = new_rlim;
1584 * It would make sense to put struct rusage in the task_struct,
1585 * except that would make the task_struct be *really big*. After
1586 * task_struct gets moved into malloc'ed memory, it would
1587 * make sense to do this. It will make moving the rest of the information
1588 * a lot simpler! (Which we're not doing right now because we're not
1589 * measuring them yet).
1591 * This is SMP safe. Either we are called from sys_getrusage on ourselves
1592 * below (we know we aren't going to exit/disappear and only we change our
1593 * rusage counters), or we are called from wait4() on a process which is
1594 * either stopped or zombied. In the zombied case the task won't get
1595 * reaped till shortly after the call to getrusage(), in both cases the
1596 * task being examined is in a frozen state so the counters won't change.
1598 int getrusage(struct task_struct *p, int who, struct rusage __user *ru)
1602 memset((char *) &r, 0, sizeof(r));
1605 jiffies_to_timeval(p->utime, &r.ru_utime);
1606 jiffies_to_timeval(p->stime, &r.ru_stime);
1607 r.ru_nvcsw = p->nvcsw;
1608 r.ru_nivcsw = p->nivcsw;
1609 r.ru_minflt = p->min_flt;
1610 r.ru_majflt = p->maj_flt;
1612 case RUSAGE_CHILDREN:
1613 jiffies_to_timeval(p->cutime, &r.ru_utime);
1614 jiffies_to_timeval(p->cstime, &r.ru_stime);
1615 r.ru_nvcsw = p->cnvcsw;
1616 r.ru_nivcsw = p->cnivcsw;
1617 r.ru_minflt = p->cmin_flt;
1618 r.ru_majflt = p->cmaj_flt;
1621 jiffies_to_timeval(p->utime + p->cutime, &r.ru_utime);
1622 jiffies_to_timeval(p->stime + p->cstime, &r.ru_stime);
1623 r.ru_nvcsw = p->nvcsw + p->cnvcsw;
1624 r.ru_nivcsw = p->nivcsw + p->cnivcsw;
1625 r.ru_minflt = p->min_flt + p->cmin_flt;
1626 r.ru_majflt = p->maj_flt + p->cmaj_flt;
1629 return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0;
1632 asmlinkage long sys_getrusage(int who, struct rusage __user *ru)
1634 if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN)
1636 return getrusage(current, who, ru);
1639 asmlinkage long sys_umask(int mask)
1641 mask = xchg(¤t->fs->umask, mask & S_IRWXUGO);
1645 asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3,
1646 unsigned long arg4, unsigned long arg5)
1651 error = security_task_prctl(option, arg2, arg3, arg4, arg5);
1656 case PR_SET_PDEATHSIG:
1658 if (sig < 0 || sig > _NSIG) {
1662 current->pdeath_signal = sig;
1664 case PR_GET_PDEATHSIG:
1665 error = put_user(current->pdeath_signal, (int __user *)arg2);
1667 case PR_GET_DUMPABLE:
1668 if (current->mm->dumpable)
1671 case PR_SET_DUMPABLE:
1672 if (arg2 != 0 && arg2 != 1) {
1676 current->mm->dumpable = arg2;
1679 case PR_SET_UNALIGN:
1680 error = SET_UNALIGN_CTL(current, arg2);
1682 case PR_GET_UNALIGN:
1683 error = GET_UNALIGN_CTL(current, arg2);
1686 error = SET_FPEMU_CTL(current, arg2);
1689 error = GET_FPEMU_CTL(current, arg2);
1692 error = SET_FPEXC_CTL(current, arg2);
1695 error = GET_FPEXC_CTL(current, arg2);
1698 error = PR_TIMING_STATISTICAL;
1701 if (arg2 == PR_TIMING_STATISTICAL)
1707 case PR_GET_KEEPCAPS:
1708 if (current->keep_capabilities)
1711 case PR_SET_KEEPCAPS:
1712 if (arg2 != 0 && arg2 != 1) {
1716 current->keep_capabilities = arg2;