Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * linux/kernel/sys.c |
| 3 | * |
| 4 | * Copyright (C) 1991, 1992 Linus Torvalds |
| 5 | */ |
| 6 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 7 | #include <linux/module.h> |
| 8 | #include <linux/mm.h> |
| 9 | #include <linux/utsname.h> |
| 10 | #include <linux/mman.h> |
| 11 | #include <linux/smp_lock.h> |
| 12 | #include <linux/notifier.h> |
| 13 | #include <linux/reboot.h> |
| 14 | #include <linux/prctl.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 15 | #include <linux/highuid.h> |
| 16 | #include <linux/fs.h> |
Daniel Walker | 3e88c55 | 2007-05-10 22:22:53 -0700 | [diff] [blame] | 17 | #include <linux/resource.h> |
Eric W. Biederman | dc009d9 | 2005-06-25 14:57:52 -0700 | [diff] [blame] | 18 | #include <linux/kernel.h> |
| 19 | #include <linux/kexec.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 20 | #include <linux/workqueue.h> |
Randy.Dunlap | c59ede7 | 2006-01-11 12:17:46 -0800 | [diff] [blame] | 21 | #include <linux/capability.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 22 | #include <linux/device.h> |
| 23 | #include <linux/key.h> |
| 24 | #include <linux/times.h> |
| 25 | #include <linux/posix-timers.h> |
| 26 | #include <linux/security.h> |
| 27 | #include <linux/dcookies.h> |
| 28 | #include <linux/suspend.h> |
| 29 | #include <linux/tty.h> |
Jesper Juhl | 7ed20e1 | 2005-05-01 08:59:14 -0700 | [diff] [blame] | 30 | #include <linux/signal.h> |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 31 | #include <linux/cn_proc.h> |
Andi Kleen | 3cfc348 | 2006-09-26 10:52:28 +0200 | [diff] [blame] | 32 | #include <linux/getcpu.h> |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 33 | #include <linux/task_io_accounting_ops.h> |
Andrea Arcangeli | 1d9d02f | 2007-07-15 23:41:32 -0700 | [diff] [blame] | 34 | #include <linux/seccomp.h> |
Mark Lord | 4047727 | 2007-10-01 01:20:10 -0700 | [diff] [blame] | 35 | #include <linux/cpu.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 36 | |
| 37 | #include <linux/compat.h> |
| 38 | #include <linux/syscalls.h> |
Keshavamurthy Anil S | 00d7c05 | 2005-12-12 00:37:33 -0800 | [diff] [blame] | 39 | #include <linux/kprobes.h> |
Cedric Le Goater | acce292 | 2007-07-15 23:40:59 -0700 | [diff] [blame] | 40 | #include <linux/user_namespace.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 41 | |
| 42 | #include <asm/uaccess.h> |
| 43 | #include <asm/io.h> |
| 44 | #include <asm/unistd.h> |
| 45 | |
| 46 | #ifndef SET_UNALIGN_CTL |
| 47 | # define SET_UNALIGN_CTL(a,b) (-EINVAL) |
| 48 | #endif |
| 49 | #ifndef GET_UNALIGN_CTL |
| 50 | # define GET_UNALIGN_CTL(a,b) (-EINVAL) |
| 51 | #endif |
| 52 | #ifndef SET_FPEMU_CTL |
| 53 | # define SET_FPEMU_CTL(a,b) (-EINVAL) |
| 54 | #endif |
| 55 | #ifndef GET_FPEMU_CTL |
| 56 | # define GET_FPEMU_CTL(a,b) (-EINVAL) |
| 57 | #endif |
| 58 | #ifndef SET_FPEXC_CTL |
| 59 | # define SET_FPEXC_CTL(a,b) (-EINVAL) |
| 60 | #endif |
| 61 | #ifndef GET_FPEXC_CTL |
| 62 | # define GET_FPEXC_CTL(a,b) (-EINVAL) |
| 63 | #endif |
Anton Blanchard | 651d765 | 2006-06-07 16:10:19 +1000 | [diff] [blame] | 64 | #ifndef GET_ENDIAN |
| 65 | # define GET_ENDIAN(a,b) (-EINVAL) |
| 66 | #endif |
| 67 | #ifndef SET_ENDIAN |
| 68 | # define SET_ENDIAN(a,b) (-EINVAL) |
| 69 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 70 | |
| 71 | /* |
| 72 | * this is where the system-wide overflow UID and GID are defined, for |
| 73 | * architectures that now have 32-bit UID/GID but didn't in the past |
| 74 | */ |
| 75 | |
| 76 | int overflowuid = DEFAULT_OVERFLOWUID; |
| 77 | int overflowgid = DEFAULT_OVERFLOWGID; |
| 78 | |
| 79 | #ifdef CONFIG_UID16 |
| 80 | EXPORT_SYMBOL(overflowuid); |
| 81 | EXPORT_SYMBOL(overflowgid); |
| 82 | #endif |
| 83 | |
| 84 | /* |
| 85 | * the same as above, but for filesystems which can only store a 16-bit |
| 86 | * UID and GID. as such, this is needed on all architectures |
| 87 | */ |
| 88 | |
| 89 | int fs_overflowuid = DEFAULT_FS_OVERFLOWUID; |
| 90 | int fs_overflowgid = DEFAULT_FS_OVERFLOWUID; |
| 91 | |
| 92 | EXPORT_SYMBOL(fs_overflowuid); |
| 93 | EXPORT_SYMBOL(fs_overflowgid); |
| 94 | |
| 95 | /* |
| 96 | * this indicates whether you can reboot with ctrl-alt-del: the default is yes |
| 97 | */ |
| 98 | |
| 99 | int C_A_D = 1; |
Cedric Le Goater | 9ec5209 | 2006-10-02 02:19:00 -0700 | [diff] [blame] | 100 | struct pid *cad_pid; |
| 101 | EXPORT_SYMBOL(cad_pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 102 | |
| 103 | /* |
Rafael J. Wysocki | bd804eb | 2007-07-19 01:47:40 -0700 | [diff] [blame] | 104 | * If set, this is used for preparing the system to power off. |
| 105 | */ |
| 106 | |
| 107 | void (*pm_power_off_prepare)(void); |
Rafael J. Wysocki | bd804eb | 2007-07-19 01:47:40 -0700 | [diff] [blame] | 108 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 109 | static int set_one_prio(struct task_struct *p, int niceval, int error) |
| 110 | { |
| 111 | int no_nice; |
| 112 | |
| 113 | if (p->uid != current->euid && |
| 114 | p->euid != current->euid && !capable(CAP_SYS_NICE)) { |
| 115 | error = -EPERM; |
| 116 | goto out; |
| 117 | } |
Matt Mackall | e43379f | 2005-05-01 08:59:00 -0700 | [diff] [blame] | 118 | if (niceval < task_nice(p) && !can_nice(p, niceval)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 119 | error = -EACCES; |
| 120 | goto out; |
| 121 | } |
| 122 | no_nice = security_task_setnice(p, niceval); |
| 123 | if (no_nice) { |
| 124 | error = no_nice; |
| 125 | goto out; |
| 126 | } |
| 127 | if (error == -ESRCH) |
| 128 | error = 0; |
| 129 | set_user_nice(p, niceval); |
| 130 | out: |
| 131 | return error; |
| 132 | } |
| 133 | |
| 134 | asmlinkage long sys_setpriority(int which, int who, int niceval) |
| 135 | { |
| 136 | struct task_struct *g, *p; |
| 137 | struct user_struct *user; |
| 138 | int error = -EINVAL; |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 139 | struct pid *pgrp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 140 | |
Daniel Walker | 3e88c55 | 2007-05-10 22:22:53 -0700 | [diff] [blame] | 141 | if (which > PRIO_USER || which < PRIO_PROCESS) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 142 | goto out; |
| 143 | |
| 144 | /* normalize: avoid signed division (rounding problems) */ |
| 145 | error = -ESRCH; |
| 146 | if (niceval < -20) |
| 147 | niceval = -20; |
| 148 | if (niceval > 19) |
| 149 | niceval = 19; |
| 150 | |
| 151 | read_lock(&tasklist_lock); |
| 152 | switch (which) { |
| 153 | case PRIO_PROCESS: |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 154 | if (who) |
Pavel Emelyanov | 228ebcb | 2007-10-18 23:40:16 -0700 | [diff] [blame] | 155 | p = find_task_by_vpid(who); |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 156 | else |
| 157 | p = current; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 158 | if (p) |
| 159 | error = set_one_prio(p, niceval, error); |
| 160 | break; |
| 161 | case PRIO_PGRP: |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 162 | if (who) |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 163 | pgrp = find_vpid(who); |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 164 | else |
| 165 | pgrp = task_pgrp(current); |
| 166 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 167 | error = set_one_prio(p, niceval, error); |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 168 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 169 | break; |
| 170 | case PRIO_USER: |
| 171 | user = current->user; |
| 172 | if (!who) |
| 173 | who = current->uid; |
| 174 | else |
| 175 | if ((who != current->uid) && !(user = find_user(who))) |
| 176 | goto out_unlock; /* No processes for this user */ |
| 177 | |
| 178 | do_each_thread(g, p) |
| 179 | if (p->uid == who) |
| 180 | error = set_one_prio(p, niceval, error); |
| 181 | while_each_thread(g, p); |
| 182 | if (who != current->uid) |
| 183 | free_uid(user); /* For find_user() */ |
| 184 | break; |
| 185 | } |
| 186 | out_unlock: |
| 187 | read_unlock(&tasklist_lock); |
| 188 | out: |
| 189 | return error; |
| 190 | } |
| 191 | |
| 192 | /* |
| 193 | * Ugh. To avoid negative return values, "getpriority()" will |
| 194 | * not return the normal nice-value, but a negated value that |
| 195 | * has been offset by 20 (ie it returns 40..1 instead of -20..19) |
| 196 | * to stay compatible. |
| 197 | */ |
| 198 | asmlinkage long sys_getpriority(int which, int who) |
| 199 | { |
| 200 | struct task_struct *g, *p; |
| 201 | struct user_struct *user; |
| 202 | long niceval, retval = -ESRCH; |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 203 | struct pid *pgrp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 204 | |
Daniel Walker | 3e88c55 | 2007-05-10 22:22:53 -0700 | [diff] [blame] | 205 | if (which > PRIO_USER || which < PRIO_PROCESS) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 206 | return -EINVAL; |
| 207 | |
| 208 | read_lock(&tasklist_lock); |
| 209 | switch (which) { |
| 210 | case PRIO_PROCESS: |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 211 | if (who) |
Pavel Emelyanov | 228ebcb | 2007-10-18 23:40:16 -0700 | [diff] [blame] | 212 | p = find_task_by_vpid(who); |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 213 | else |
| 214 | p = current; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 215 | if (p) { |
| 216 | niceval = 20 - task_nice(p); |
| 217 | if (niceval > retval) |
| 218 | retval = niceval; |
| 219 | } |
| 220 | break; |
| 221 | case PRIO_PGRP: |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 222 | if (who) |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 223 | pgrp = find_vpid(who); |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 224 | else |
| 225 | pgrp = task_pgrp(current); |
| 226 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 227 | niceval = 20 - task_nice(p); |
| 228 | if (niceval > retval) |
| 229 | retval = niceval; |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 230 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 231 | break; |
| 232 | case PRIO_USER: |
| 233 | user = current->user; |
| 234 | if (!who) |
| 235 | who = current->uid; |
| 236 | else |
| 237 | if ((who != current->uid) && !(user = find_user(who))) |
| 238 | goto out_unlock; /* No processes for this user */ |
| 239 | |
| 240 | do_each_thread(g, p) |
| 241 | if (p->uid == who) { |
| 242 | niceval = 20 - task_nice(p); |
| 243 | if (niceval > retval) |
| 244 | retval = niceval; |
| 245 | } |
| 246 | while_each_thread(g, p); |
| 247 | if (who != current->uid) |
| 248 | free_uid(user); /* for find_user() */ |
| 249 | break; |
| 250 | } |
| 251 | out_unlock: |
| 252 | read_unlock(&tasklist_lock); |
| 253 | |
| 254 | return retval; |
| 255 | } |
| 256 | |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 257 | /** |
| 258 | * emergency_restart - reboot the system |
| 259 | * |
| 260 | * Without shutting down any hardware or taking any locks |
| 261 | * reboot the system. This is called when we know we are in |
| 262 | * trouble so this is our best effort to reboot. This is |
| 263 | * safe to call in interrupt context. |
| 264 | */ |
Eric W. Biederman | 7c90347 | 2005-07-26 11:29:55 -0600 | [diff] [blame] | 265 | void emergency_restart(void) |
| 266 | { |
| 267 | machine_emergency_restart(); |
| 268 | } |
| 269 | EXPORT_SYMBOL_GPL(emergency_restart); |
| 270 | |
Adrian Bunk | 83cc5ed | 2006-06-25 05:47:41 -0700 | [diff] [blame] | 271 | static void kernel_restart_prepare(char *cmd) |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 272 | { |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 273 | blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd); |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 274 | system_state = SYSTEM_RESTART; |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 275 | device_shutdown(); |
Rafael J. Wysocki | 58b3b71 | 2007-07-26 16:29:55 +0200 | [diff] [blame] | 276 | sysdev_shutdown(); |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 277 | } |
Randy Dunlap | 1e5d533 | 2005-11-07 01:01:06 -0800 | [diff] [blame] | 278 | |
| 279 | /** |
| 280 | * kernel_restart - reboot the system |
| 281 | * @cmd: pointer to buffer containing command to execute for restart |
Randy Dunlap | b8887e6 | 2005-11-07 01:01:07 -0800 | [diff] [blame] | 282 | * or %NULL |
Randy Dunlap | 1e5d533 | 2005-11-07 01:01:06 -0800 | [diff] [blame] | 283 | * |
| 284 | * Shutdown everything and perform a clean reboot. |
| 285 | * This is not safe to call in interrupt context. |
| 286 | */ |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 287 | void kernel_restart(char *cmd) |
| 288 | { |
| 289 | kernel_restart_prepare(cmd); |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 290 | if (!cmd) |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 291 | printk(KERN_EMERG "Restarting system.\n"); |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 292 | else |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 293 | printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd); |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 294 | machine_restart(cmd); |
| 295 | } |
| 296 | EXPORT_SYMBOL_GPL(kernel_restart); |
| 297 | |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 298 | /** |
| 299 | * kernel_kexec - reboot the system |
| 300 | * |
| 301 | * Move into place and start executing a preloaded standalone |
| 302 | * executable. If nothing was preloaded return an error. |
| 303 | */ |
Adrian Bunk | 83cc5ed | 2006-06-25 05:47:41 -0700 | [diff] [blame] | 304 | static void kernel_kexec(void) |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 305 | { |
| 306 | #ifdef CONFIG_KEXEC |
| 307 | struct kimage *image; |
Al Viro | 4bb8089 | 2006-02-01 05:57:32 -0500 | [diff] [blame] | 308 | image = xchg(&kexec_image, NULL); |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 309 | if (!image) |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 310 | return; |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 311 | kernel_restart_prepare(NULL); |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 312 | printk(KERN_EMERG "Starting new kernel\n"); |
| 313 | machine_shutdown(); |
| 314 | machine_kexec(image); |
| 315 | #endif |
| 316 | } |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 317 | |
Alexey Starikovskiy | 729b4d4 | 2005-12-01 04:29:00 -0500 | [diff] [blame] | 318 | void kernel_shutdown_prepare(enum system_states state) |
| 319 | { |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 320 | blocking_notifier_call_chain(&reboot_notifier_list, |
Alexey Starikovskiy | 729b4d4 | 2005-12-01 04:29:00 -0500 | [diff] [blame] | 321 | (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL); |
| 322 | system_state = state; |
| 323 | device_shutdown(); |
| 324 | } |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 325 | /** |
| 326 | * kernel_halt - halt the system |
| 327 | * |
| 328 | * Shutdown everything and perform a clean system halt. |
| 329 | */ |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 330 | void kernel_halt(void) |
| 331 | { |
Alexey Starikovskiy | 729b4d4 | 2005-12-01 04:29:00 -0500 | [diff] [blame] | 332 | kernel_shutdown_prepare(SYSTEM_HALT); |
Rafael J. Wysocki | 58b3b71 | 2007-07-26 16:29:55 +0200 | [diff] [blame] | 333 | sysdev_shutdown(); |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 334 | printk(KERN_EMERG "System halted.\n"); |
| 335 | machine_halt(); |
| 336 | } |
Alexey Starikovskiy | 729b4d4 | 2005-12-01 04:29:00 -0500 | [diff] [blame] | 337 | |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 338 | EXPORT_SYMBOL_GPL(kernel_halt); |
| 339 | |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 340 | /** |
| 341 | * kernel_power_off - power_off the system |
| 342 | * |
| 343 | * Shutdown everything and perform a clean system power_off. |
| 344 | */ |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 345 | void kernel_power_off(void) |
| 346 | { |
Alexey Starikovskiy | 729b4d4 | 2005-12-01 04:29:00 -0500 | [diff] [blame] | 347 | kernel_shutdown_prepare(SYSTEM_POWER_OFF); |
Rafael J. Wysocki | bd804eb | 2007-07-19 01:47:40 -0700 | [diff] [blame] | 348 | if (pm_power_off_prepare) |
| 349 | pm_power_off_prepare(); |
Mark Lord | 4047727 | 2007-10-01 01:20:10 -0700 | [diff] [blame] | 350 | disable_nonboot_cpus(); |
Rafael J. Wysocki | 58b3b71 | 2007-07-26 16:29:55 +0200 | [diff] [blame] | 351 | sysdev_shutdown(); |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 352 | printk(KERN_EMERG "Power down.\n"); |
| 353 | machine_power_off(); |
| 354 | } |
| 355 | EXPORT_SYMBOL_GPL(kernel_power_off); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 356 | /* |
| 357 | * Reboot system call: for obvious reasons only root may call it, |
| 358 | * and even root needs to set up some magic numbers in the registers |
| 359 | * so that some mistake won't make this reboot the whole machine. |
| 360 | * You can also set the meaning of the ctrl-alt-del-key here. |
| 361 | * |
| 362 | * reboot doesn't sync: do that yourself before calling this. |
| 363 | */ |
| 364 | asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg) |
| 365 | { |
| 366 | char buffer[256]; |
| 367 | |
| 368 | /* We only trust the superuser with rebooting the system. */ |
| 369 | if (!capable(CAP_SYS_BOOT)) |
| 370 | return -EPERM; |
| 371 | |
| 372 | /* For safety, we require "magic" arguments. */ |
| 373 | if (magic1 != LINUX_REBOOT_MAGIC1 || |
| 374 | (magic2 != LINUX_REBOOT_MAGIC2 && |
| 375 | magic2 != LINUX_REBOOT_MAGIC2A && |
| 376 | magic2 != LINUX_REBOOT_MAGIC2B && |
| 377 | magic2 != LINUX_REBOOT_MAGIC2C)) |
| 378 | return -EINVAL; |
| 379 | |
Eric W. Biederman | 5e38291 | 2006-01-08 01:03:46 -0800 | [diff] [blame] | 380 | /* Instead of trying to make the power_off code look like |
| 381 | * halt when pm_power_off is not set do it the easy way. |
| 382 | */ |
| 383 | if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off) |
| 384 | cmd = LINUX_REBOOT_CMD_HALT; |
| 385 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 386 | lock_kernel(); |
| 387 | switch (cmd) { |
| 388 | case LINUX_REBOOT_CMD_RESTART: |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 389 | kernel_restart(NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 390 | break; |
| 391 | |
| 392 | case LINUX_REBOOT_CMD_CAD_ON: |
| 393 | C_A_D = 1; |
| 394 | break; |
| 395 | |
| 396 | case LINUX_REBOOT_CMD_CAD_OFF: |
| 397 | C_A_D = 0; |
| 398 | break; |
| 399 | |
| 400 | case LINUX_REBOOT_CMD_HALT: |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 401 | kernel_halt(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 402 | unlock_kernel(); |
| 403 | do_exit(0); |
| 404 | break; |
| 405 | |
| 406 | case LINUX_REBOOT_CMD_POWER_OFF: |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 407 | kernel_power_off(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 408 | unlock_kernel(); |
| 409 | do_exit(0); |
| 410 | break; |
| 411 | |
| 412 | case LINUX_REBOOT_CMD_RESTART2: |
| 413 | if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) { |
| 414 | unlock_kernel(); |
| 415 | return -EFAULT; |
| 416 | } |
| 417 | buffer[sizeof(buffer) - 1] = '\0'; |
| 418 | |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 419 | kernel_restart(buffer); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 420 | break; |
| 421 | |
Eric W. Biederman | dc009d9 | 2005-06-25 14:57:52 -0700 | [diff] [blame] | 422 | case LINUX_REBOOT_CMD_KEXEC: |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 423 | kernel_kexec(); |
| 424 | unlock_kernel(); |
| 425 | return -EINVAL; |
| 426 | |
Rafael J. Wysocki | b0cb1a1 | 2007-07-29 23:24:36 +0200 | [diff] [blame] | 427 | #ifdef CONFIG_HIBERNATION |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 428 | case LINUX_REBOOT_CMD_SW_SUSPEND: |
| 429 | { |
Rafael J. Wysocki | a3d25c2 | 2007-05-09 02:33:18 -0700 | [diff] [blame] | 430 | int ret = hibernate(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 431 | unlock_kernel(); |
| 432 | return ret; |
| 433 | } |
| 434 | #endif |
| 435 | |
| 436 | default: |
| 437 | unlock_kernel(); |
| 438 | return -EINVAL; |
| 439 | } |
| 440 | unlock_kernel(); |
| 441 | return 0; |
| 442 | } |
| 443 | |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 444 | static void deferred_cad(struct work_struct *dummy) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 445 | { |
Eric W. Biederman | abcd9e5 | 2005-07-26 11:27:34 -0600 | [diff] [blame] | 446 | kernel_restart(NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 447 | } |
| 448 | |
| 449 | /* |
| 450 | * This function gets called by ctrl-alt-del - ie the keyboard interrupt. |
| 451 | * As it's called within an interrupt, it may NOT sync: the only choice |
| 452 | * is whether to reboot at once, or just ignore the ctrl-alt-del. |
| 453 | */ |
| 454 | void ctrl_alt_del(void) |
| 455 | { |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 456 | static DECLARE_WORK(cad_work, deferred_cad); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 457 | |
| 458 | if (C_A_D) |
| 459 | schedule_work(&cad_work); |
| 460 | else |
Cedric Le Goater | 9ec5209 | 2006-10-02 02:19:00 -0700 | [diff] [blame] | 461 | kill_cad_pid(SIGINT, 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 462 | } |
| 463 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 464 | /* |
| 465 | * Unprivileged users may change the real gid to the effective gid |
| 466 | * or vice versa. (BSD-style) |
| 467 | * |
| 468 | * If you set the real gid at all, or set the effective gid to a value not |
| 469 | * equal to the real gid, then the saved gid is set to the new effective gid. |
| 470 | * |
| 471 | * This makes it possible for a setgid program to completely drop its |
| 472 | * privileges, which is often a useful assertion to make when you are doing |
| 473 | * a security audit over a program. |
| 474 | * |
| 475 | * The general idea is that a program which uses just setregid() will be |
| 476 | * 100% compatible with BSD. A program which uses just setgid() will be |
| 477 | * 100% compatible with POSIX with saved IDs. |
| 478 | * |
| 479 | * SMP: There are not races, the GIDs are checked only by filesystem |
| 480 | * operations (as far as semantic preservation is concerned). |
| 481 | */ |
| 482 | asmlinkage long sys_setregid(gid_t rgid, gid_t egid) |
| 483 | { |
| 484 | int old_rgid = current->gid; |
| 485 | int old_egid = current->egid; |
| 486 | int new_rgid = old_rgid; |
| 487 | int new_egid = old_egid; |
| 488 | int retval; |
| 489 | |
| 490 | retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE); |
| 491 | if (retval) |
| 492 | return retval; |
| 493 | |
| 494 | if (rgid != (gid_t) -1) { |
| 495 | if ((old_rgid == rgid) || |
| 496 | (current->egid==rgid) || |
| 497 | capable(CAP_SETGID)) |
| 498 | new_rgid = rgid; |
| 499 | else |
| 500 | return -EPERM; |
| 501 | } |
| 502 | if (egid != (gid_t) -1) { |
| 503 | if ((old_rgid == egid) || |
| 504 | (current->egid == egid) || |
| 505 | (current->sgid == egid) || |
| 506 | capable(CAP_SETGID)) |
| 507 | new_egid = egid; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 508 | else |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 509 | return -EPERM; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 510 | } |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 511 | if (new_egid != old_egid) { |
Kawai, Hidehiro | 6c5d523 | 2007-07-19 01:48:27 -0700 | [diff] [blame] | 512 | set_dumpable(current->mm, suid_dumpable); |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 513 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 514 | } |
| 515 | if (rgid != (gid_t) -1 || |
| 516 | (egid != (gid_t) -1 && egid != old_rgid)) |
| 517 | current->sgid = new_egid; |
| 518 | current->fsgid = new_egid; |
| 519 | current->egid = new_egid; |
| 520 | current->gid = new_rgid; |
| 521 | key_fsgid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 522 | proc_id_connector(current, PROC_EVENT_GID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 523 | return 0; |
| 524 | } |
| 525 | |
| 526 | /* |
| 527 | * setgid() is implemented like SysV w/ SAVED_IDS |
| 528 | * |
| 529 | * SMP: Same implicit races as above. |
| 530 | */ |
| 531 | asmlinkage long sys_setgid(gid_t gid) |
| 532 | { |
| 533 | int old_egid = current->egid; |
| 534 | int retval; |
| 535 | |
| 536 | retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID); |
| 537 | if (retval) |
| 538 | return retval; |
| 539 | |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 540 | if (capable(CAP_SETGID)) { |
| 541 | if (old_egid != gid) { |
Kawai, Hidehiro | 6c5d523 | 2007-07-19 01:48:27 -0700 | [diff] [blame] | 542 | set_dumpable(current->mm, suid_dumpable); |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 543 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 544 | } |
| 545 | current->gid = current->egid = current->sgid = current->fsgid = gid; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 546 | } else if ((gid == current->gid) || (gid == current->sgid)) { |
| 547 | if (old_egid != gid) { |
Kawai, Hidehiro | 6c5d523 | 2007-07-19 01:48:27 -0700 | [diff] [blame] | 548 | set_dumpable(current->mm, suid_dumpable); |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 549 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 550 | } |
| 551 | current->egid = current->fsgid = gid; |
| 552 | } |
| 553 | else |
| 554 | return -EPERM; |
| 555 | |
| 556 | key_fsgid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 557 | proc_id_connector(current, PROC_EVENT_GID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 558 | return 0; |
| 559 | } |
| 560 | |
| 561 | static int set_user(uid_t new_ruid, int dumpclear) |
| 562 | { |
| 563 | struct user_struct *new_user; |
| 564 | |
Cedric Le Goater | acce292 | 2007-07-15 23:40:59 -0700 | [diff] [blame] | 565 | new_user = alloc_uid(current->nsproxy->user_ns, new_ruid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 566 | if (!new_user) |
| 567 | return -EAGAIN; |
| 568 | |
| 569 | if (atomic_read(&new_user->processes) >= |
| 570 | current->signal->rlim[RLIMIT_NPROC].rlim_cur && |
Cedric Le Goater | acce292 | 2007-07-15 23:40:59 -0700 | [diff] [blame] | 571 | new_user != current->nsproxy->user_ns->root_user) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 572 | free_uid(new_user); |
| 573 | return -EAGAIN; |
| 574 | } |
| 575 | |
| 576 | switch_uid(new_user); |
| 577 | |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 578 | if (dumpclear) { |
Kawai, Hidehiro | 6c5d523 | 2007-07-19 01:48:27 -0700 | [diff] [blame] | 579 | set_dumpable(current->mm, suid_dumpable); |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 580 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 581 | } |
| 582 | current->uid = new_ruid; |
| 583 | return 0; |
| 584 | } |
| 585 | |
| 586 | /* |
| 587 | * Unprivileged users may change the real uid to the effective uid |
| 588 | * or vice versa. (BSD-style) |
| 589 | * |
| 590 | * If you set the real uid at all, or set the effective uid to a value not |
| 591 | * equal to the real uid, then the saved uid is set to the new effective uid. |
| 592 | * |
| 593 | * This makes it possible for a setuid program to completely drop its |
| 594 | * privileges, which is often a useful assertion to make when you are doing |
| 595 | * a security audit over a program. |
| 596 | * |
| 597 | * The general idea is that a program which uses just setreuid() will be |
| 598 | * 100% compatible with BSD. A program which uses just setuid() will be |
| 599 | * 100% compatible with POSIX with saved IDs. |
| 600 | */ |
| 601 | asmlinkage long sys_setreuid(uid_t ruid, uid_t euid) |
| 602 | { |
| 603 | int old_ruid, old_euid, old_suid, new_ruid, new_euid; |
| 604 | int retval; |
| 605 | |
| 606 | retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE); |
| 607 | if (retval) |
| 608 | return retval; |
| 609 | |
| 610 | new_ruid = old_ruid = current->uid; |
| 611 | new_euid = old_euid = current->euid; |
| 612 | old_suid = current->suid; |
| 613 | |
| 614 | if (ruid != (uid_t) -1) { |
| 615 | new_ruid = ruid; |
| 616 | if ((old_ruid != ruid) && |
| 617 | (current->euid != ruid) && |
| 618 | !capable(CAP_SETUID)) |
| 619 | return -EPERM; |
| 620 | } |
| 621 | |
| 622 | if (euid != (uid_t) -1) { |
| 623 | new_euid = euid; |
| 624 | if ((old_ruid != euid) && |
| 625 | (current->euid != euid) && |
| 626 | (current->suid != euid) && |
| 627 | !capable(CAP_SETUID)) |
| 628 | return -EPERM; |
| 629 | } |
| 630 | |
| 631 | if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0) |
| 632 | return -EAGAIN; |
| 633 | |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 634 | if (new_euid != old_euid) { |
Kawai, Hidehiro | 6c5d523 | 2007-07-19 01:48:27 -0700 | [diff] [blame] | 635 | set_dumpable(current->mm, suid_dumpable); |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 636 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 637 | } |
| 638 | current->fsuid = current->euid = new_euid; |
| 639 | if (ruid != (uid_t) -1 || |
| 640 | (euid != (uid_t) -1 && euid != old_ruid)) |
| 641 | current->suid = current->euid; |
| 642 | current->fsuid = current->euid; |
| 643 | |
| 644 | key_fsuid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 645 | proc_id_connector(current, PROC_EVENT_UID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 646 | |
| 647 | return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE); |
| 648 | } |
| 649 | |
| 650 | |
| 651 | |
| 652 | /* |
| 653 | * setuid() is implemented like SysV with SAVED_IDS |
| 654 | * |
| 655 | * Note that SAVED_ID's is deficient in that a setuid root program |
| 656 | * like sendmail, for example, cannot set its uid to be a normal |
| 657 | * user and then switch back, because if you're root, setuid() sets |
| 658 | * the saved uid too. If you don't like this, blame the bright people |
| 659 | * in the POSIX committee and/or USG. Note that the BSD-style setreuid() |
| 660 | * will allow a root program to temporarily drop privileges and be able to |
| 661 | * regain them by swapping the real and effective uid. |
| 662 | */ |
| 663 | asmlinkage long sys_setuid(uid_t uid) |
| 664 | { |
| 665 | int old_euid = current->euid; |
David Rientjes | a09c17a | 2006-12-06 20:40:18 -0800 | [diff] [blame] | 666 | int old_ruid, old_suid, new_suid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 667 | int retval; |
| 668 | |
| 669 | retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID); |
| 670 | if (retval) |
| 671 | return retval; |
| 672 | |
David Rientjes | a09c17a | 2006-12-06 20:40:18 -0800 | [diff] [blame] | 673 | old_ruid = current->uid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 674 | old_suid = current->suid; |
| 675 | new_suid = old_suid; |
| 676 | |
| 677 | if (capable(CAP_SETUID)) { |
| 678 | if (uid != old_ruid && set_user(uid, old_euid != uid) < 0) |
| 679 | return -EAGAIN; |
| 680 | new_suid = uid; |
| 681 | } else if ((uid != current->uid) && (uid != new_suid)) |
| 682 | return -EPERM; |
| 683 | |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 684 | if (old_euid != uid) { |
Kawai, Hidehiro | 6c5d523 | 2007-07-19 01:48:27 -0700 | [diff] [blame] | 685 | set_dumpable(current->mm, suid_dumpable); |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 686 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 687 | } |
| 688 | current->fsuid = current->euid = uid; |
| 689 | current->suid = new_suid; |
| 690 | |
| 691 | key_fsuid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 692 | proc_id_connector(current, PROC_EVENT_UID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 693 | |
| 694 | return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID); |
| 695 | } |
| 696 | |
| 697 | |
| 698 | /* |
| 699 | * This function implements a generic ability to update ruid, euid, |
| 700 | * and suid. This allows you to implement the 4.4 compatible seteuid(). |
| 701 | */ |
| 702 | asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid) |
| 703 | { |
| 704 | int old_ruid = current->uid; |
| 705 | int old_euid = current->euid; |
| 706 | int old_suid = current->suid; |
| 707 | int retval; |
| 708 | |
| 709 | retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES); |
| 710 | if (retval) |
| 711 | return retval; |
| 712 | |
| 713 | if (!capable(CAP_SETUID)) { |
| 714 | if ((ruid != (uid_t) -1) && (ruid != current->uid) && |
| 715 | (ruid != current->euid) && (ruid != current->suid)) |
| 716 | return -EPERM; |
| 717 | if ((euid != (uid_t) -1) && (euid != current->uid) && |
| 718 | (euid != current->euid) && (euid != current->suid)) |
| 719 | return -EPERM; |
| 720 | if ((suid != (uid_t) -1) && (suid != current->uid) && |
| 721 | (suid != current->euid) && (suid != current->suid)) |
| 722 | return -EPERM; |
| 723 | } |
| 724 | if (ruid != (uid_t) -1) { |
| 725 | if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0) |
| 726 | return -EAGAIN; |
| 727 | } |
| 728 | if (euid != (uid_t) -1) { |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 729 | if (euid != current->euid) { |
Kawai, Hidehiro | 6c5d523 | 2007-07-19 01:48:27 -0700 | [diff] [blame] | 730 | set_dumpable(current->mm, suid_dumpable); |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 731 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 732 | } |
| 733 | current->euid = euid; |
| 734 | } |
| 735 | current->fsuid = current->euid; |
| 736 | if (suid != (uid_t) -1) |
| 737 | current->suid = suid; |
| 738 | |
| 739 | key_fsuid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 740 | proc_id_connector(current, PROC_EVENT_UID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 741 | |
| 742 | return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES); |
| 743 | } |
| 744 | |
| 745 | asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid) |
| 746 | { |
| 747 | int retval; |
| 748 | |
| 749 | if (!(retval = put_user(current->uid, ruid)) && |
| 750 | !(retval = put_user(current->euid, euid))) |
| 751 | retval = put_user(current->suid, suid); |
| 752 | |
| 753 | return retval; |
| 754 | } |
| 755 | |
| 756 | /* |
| 757 | * Same as above, but for rgid, egid, sgid. |
| 758 | */ |
| 759 | asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid) |
| 760 | { |
| 761 | int retval; |
| 762 | |
| 763 | retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES); |
| 764 | if (retval) |
| 765 | return retval; |
| 766 | |
| 767 | if (!capable(CAP_SETGID)) { |
| 768 | if ((rgid != (gid_t) -1) && (rgid != current->gid) && |
| 769 | (rgid != current->egid) && (rgid != current->sgid)) |
| 770 | return -EPERM; |
| 771 | if ((egid != (gid_t) -1) && (egid != current->gid) && |
| 772 | (egid != current->egid) && (egid != current->sgid)) |
| 773 | return -EPERM; |
| 774 | if ((sgid != (gid_t) -1) && (sgid != current->gid) && |
| 775 | (sgid != current->egid) && (sgid != current->sgid)) |
| 776 | return -EPERM; |
| 777 | } |
| 778 | if (egid != (gid_t) -1) { |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 779 | if (egid != current->egid) { |
Kawai, Hidehiro | 6c5d523 | 2007-07-19 01:48:27 -0700 | [diff] [blame] | 780 | set_dumpable(current->mm, suid_dumpable); |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 781 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 782 | } |
| 783 | current->egid = egid; |
| 784 | } |
| 785 | current->fsgid = current->egid; |
| 786 | if (rgid != (gid_t) -1) |
| 787 | current->gid = rgid; |
| 788 | if (sgid != (gid_t) -1) |
| 789 | current->sgid = sgid; |
| 790 | |
| 791 | key_fsgid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 792 | proc_id_connector(current, PROC_EVENT_GID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 793 | return 0; |
| 794 | } |
| 795 | |
| 796 | asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid) |
| 797 | { |
| 798 | int retval; |
| 799 | |
| 800 | if (!(retval = put_user(current->gid, rgid)) && |
| 801 | !(retval = put_user(current->egid, egid))) |
| 802 | retval = put_user(current->sgid, sgid); |
| 803 | |
| 804 | return retval; |
| 805 | } |
| 806 | |
| 807 | |
| 808 | /* |
| 809 | * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This |
| 810 | * is used for "access()" and for the NFS daemon (letting nfsd stay at |
| 811 | * whatever uid it wants to). It normally shadows "euid", except when |
| 812 | * explicitly set by setfsuid() or for access.. |
| 813 | */ |
| 814 | asmlinkage long sys_setfsuid(uid_t uid) |
| 815 | { |
| 816 | int old_fsuid; |
| 817 | |
| 818 | old_fsuid = current->fsuid; |
| 819 | if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS)) |
| 820 | return old_fsuid; |
| 821 | |
| 822 | if (uid == current->uid || uid == current->euid || |
| 823 | uid == current->suid || uid == current->fsuid || |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 824 | capable(CAP_SETUID)) { |
| 825 | if (uid != old_fsuid) { |
Kawai, Hidehiro | 6c5d523 | 2007-07-19 01:48:27 -0700 | [diff] [blame] | 826 | set_dumpable(current->mm, suid_dumpable); |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 827 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 828 | } |
| 829 | current->fsuid = uid; |
| 830 | } |
| 831 | |
| 832 | key_fsuid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 833 | proc_id_connector(current, PROC_EVENT_UID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 834 | |
| 835 | security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS); |
| 836 | |
| 837 | return old_fsuid; |
| 838 | } |
| 839 | |
| 840 | /* |
John Anthony Kazos Jr | f42df9e | 2007-05-09 08:23:08 +0200 | [diff] [blame] | 841 | * Samma på svenska.. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 842 | */ |
| 843 | asmlinkage long sys_setfsgid(gid_t gid) |
| 844 | { |
| 845 | int old_fsgid; |
| 846 | |
| 847 | old_fsgid = current->fsgid; |
| 848 | if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS)) |
| 849 | return old_fsgid; |
| 850 | |
| 851 | if (gid == current->gid || gid == current->egid || |
| 852 | gid == current->sgid || gid == current->fsgid || |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 853 | capable(CAP_SETGID)) { |
| 854 | if (gid != old_fsgid) { |
Kawai, Hidehiro | 6c5d523 | 2007-07-19 01:48:27 -0700 | [diff] [blame] | 855 | set_dumpable(current->mm, suid_dumpable); |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 856 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 857 | } |
| 858 | current->fsgid = gid; |
| 859 | key_fsgid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 860 | proc_id_connector(current, PROC_EVENT_GID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 861 | } |
| 862 | return old_fsgid; |
| 863 | } |
| 864 | |
| 865 | asmlinkage long sys_times(struct tms __user * tbuf) |
| 866 | { |
| 867 | /* |
| 868 | * In the SMP world we might just be unlucky and have one of |
| 869 | * the times increment as we use it. Since the value is an |
| 870 | * atomically safe type this is just fine. Conceptually its |
| 871 | * as if the syscall took an instant longer to occur. |
| 872 | */ |
| 873 | if (tbuf) { |
| 874 | struct tms tmp; |
Oleg Nesterov | 35f5cad | 2006-03-28 16:11:19 -0800 | [diff] [blame] | 875 | struct task_struct *tsk = current; |
| 876 | struct task_struct *t; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 877 | cputime_t utime, stime, cutime, cstime; |
| 878 | |
Oleg Nesterov | 7d7185c | 2006-03-28 16:11:21 -0800 | [diff] [blame] | 879 | spin_lock_irq(&tsk->sighand->siglock); |
Oleg Nesterov | 35f5cad | 2006-03-28 16:11:19 -0800 | [diff] [blame] | 880 | utime = tsk->signal->utime; |
| 881 | stime = tsk->signal->stime; |
| 882 | t = tsk; |
| 883 | do { |
| 884 | utime = cputime_add(utime, t->utime); |
| 885 | stime = cputime_add(stime, t->stime); |
| 886 | t = next_thread(t); |
| 887 | } while (t != tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 888 | |
Oleg Nesterov | 35f5cad | 2006-03-28 16:11:19 -0800 | [diff] [blame] | 889 | cutime = tsk->signal->cutime; |
| 890 | cstime = tsk->signal->cstime; |
| 891 | spin_unlock_irq(&tsk->sighand->siglock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 892 | |
| 893 | tmp.tms_utime = cputime_to_clock_t(utime); |
| 894 | tmp.tms_stime = cputime_to_clock_t(stime); |
| 895 | tmp.tms_cutime = cputime_to_clock_t(cutime); |
| 896 | tmp.tms_cstime = cputime_to_clock_t(cstime); |
| 897 | if (copy_to_user(tbuf, &tmp, sizeof(struct tms))) |
| 898 | return -EFAULT; |
| 899 | } |
| 900 | return (long) jiffies_64_to_clock_t(get_jiffies_64()); |
| 901 | } |
| 902 | |
| 903 | /* |
| 904 | * This needs some heavy checking ... |
| 905 | * I just haven't the stomach for it. I also don't fully |
| 906 | * understand sessions/pgrp etc. Let somebody who does explain it. |
| 907 | * |
| 908 | * OK, I think I have the protection semantics right.... this is really |
| 909 | * only important on a multi-user system anyway, to make sure one user |
| 910 | * can't send a signal to a process owned by another. -TYT, 12/12/91 |
| 911 | * |
| 912 | * Auch. Had to add the 'did_exec' flag to conform completely to POSIX. |
| 913 | * LBT 04.03.94 |
| 914 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 915 | asmlinkage long sys_setpgid(pid_t pid, pid_t pgid) |
| 916 | { |
| 917 | struct task_struct *p; |
Oleg Nesterov | ee0acf9 | 2006-01-08 01:03:53 -0800 | [diff] [blame] | 918 | struct task_struct *group_leader = current->group_leader; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 919 | int err = -EINVAL; |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 920 | struct pid_namespace *ns; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 921 | |
| 922 | if (!pid) |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 923 | pid = task_pid_vnr(group_leader); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 924 | if (!pgid) |
| 925 | pgid = pid; |
| 926 | if (pgid < 0) |
| 927 | return -EINVAL; |
| 928 | |
| 929 | /* From this point forward we keep holding onto the tasklist lock |
| 930 | * so that our parent does not change from under us. -DaveM |
| 931 | */ |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 932 | ns = current->nsproxy->pid_ns; |
| 933 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 934 | write_lock_irq(&tasklist_lock); |
| 935 | |
| 936 | err = -ESRCH; |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 937 | p = find_task_by_pid_ns(pid, ns); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 938 | if (!p) |
| 939 | goto out; |
| 940 | |
| 941 | err = -EINVAL; |
| 942 | if (!thread_group_leader(p)) |
| 943 | goto out; |
| 944 | |
Oleg Nesterov | b07e35f | 2007-08-30 23:56:27 -0700 | [diff] [blame] | 945 | if (p->real_parent->tgid == group_leader->tgid) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 946 | err = -EPERM; |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 947 | if (task_session(p) != task_session(group_leader)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 948 | goto out; |
| 949 | err = -EACCES; |
| 950 | if (p->did_exec) |
| 951 | goto out; |
| 952 | } else { |
| 953 | err = -ESRCH; |
Oleg Nesterov | ee0acf9 | 2006-01-08 01:03:53 -0800 | [diff] [blame] | 954 | if (p != group_leader) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 955 | goto out; |
| 956 | } |
| 957 | |
| 958 | err = -EPERM; |
| 959 | if (p->signal->leader) |
| 960 | goto out; |
| 961 | |
| 962 | if (pgid != pid) { |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 963 | struct task_struct *g; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 964 | |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 965 | g = find_task_by_pid_type_ns(PIDTYPE_PGID, pgid, ns); |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 966 | if (!g || task_session(g) != task_session(group_leader)) |
Oleg Nesterov | f020bc4 | 2006-12-08 02:38:02 -0800 | [diff] [blame] | 967 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 968 | } |
| 969 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 970 | err = security_task_setpgid(p, pgid); |
| 971 | if (err) |
| 972 | goto out; |
| 973 | |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 974 | if (task_pgrp_nr_ns(p, ns) != pgid) { |
| 975 | struct pid *pid; |
| 976 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 977 | detach_pid(p, PIDTYPE_PGID); |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 978 | pid = find_vpid(pgid); |
| 979 | attach_pid(p, PIDTYPE_PGID, pid); |
Pavel Emelyanov | 9a2e705 | 2007-10-18 23:40:39 -0700 | [diff] [blame] | 980 | set_task_pgrp(p, pid_nr(pid)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 981 | } |
| 982 | |
| 983 | err = 0; |
| 984 | out: |
| 985 | /* All paths lead to here, thus we are safe. -DaveM */ |
| 986 | write_unlock_irq(&tasklist_lock); |
| 987 | return err; |
| 988 | } |
| 989 | |
| 990 | asmlinkage long sys_getpgid(pid_t pid) |
| 991 | { |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 992 | if (!pid) |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 993 | return task_pgrp_vnr(current); |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 994 | else { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 995 | int retval; |
| 996 | struct task_struct *p; |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 997 | struct pid_namespace *ns; |
| 998 | |
| 999 | ns = current->nsproxy->pid_ns; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1000 | |
| 1001 | read_lock(&tasklist_lock); |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 1002 | p = find_task_by_pid_ns(pid, ns); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1003 | retval = -ESRCH; |
| 1004 | if (p) { |
| 1005 | retval = security_task_getpgid(p); |
| 1006 | if (!retval) |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 1007 | retval = task_pgrp_nr_ns(p, ns); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1008 | } |
| 1009 | read_unlock(&tasklist_lock); |
| 1010 | return retval; |
| 1011 | } |
| 1012 | } |
| 1013 | |
| 1014 | #ifdef __ARCH_WANT_SYS_GETPGRP |
| 1015 | |
| 1016 | asmlinkage long sys_getpgrp(void) |
| 1017 | { |
| 1018 | /* SMP - assuming writes are word atomic this is fine */ |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 1019 | return task_pgrp_vnr(current); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1020 | } |
| 1021 | |
| 1022 | #endif |
| 1023 | |
| 1024 | asmlinkage long sys_getsid(pid_t pid) |
| 1025 | { |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1026 | if (!pid) |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 1027 | return task_session_vnr(current); |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1028 | else { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1029 | int retval; |
| 1030 | struct task_struct *p; |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 1031 | struct pid_namespace *ns; |
| 1032 | |
| 1033 | ns = current->nsproxy->pid_ns; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1034 | |
| 1035 | read_lock(&tasklist_lock); |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 1036 | p = find_task_by_pid_ns(pid, ns); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1037 | retval = -ESRCH; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1038 | if (p) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1039 | retval = security_task_getsid(p); |
| 1040 | if (!retval) |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 1041 | retval = task_session_nr_ns(p, ns); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1042 | } |
| 1043 | read_unlock(&tasklist_lock); |
| 1044 | return retval; |
| 1045 | } |
| 1046 | } |
| 1047 | |
| 1048 | asmlinkage long sys_setsid(void) |
| 1049 | { |
Oren Laadan | e19f247 | 2006-01-08 01:03:58 -0800 | [diff] [blame] | 1050 | struct task_struct *group_leader = current->group_leader; |
Eric W. Biederman | 390e2ff | 2006-03-31 02:31:33 -0800 | [diff] [blame] | 1051 | pid_t session; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1052 | int err = -EPERM; |
| 1053 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1054 | write_lock_irq(&tasklist_lock); |
| 1055 | |
Eric W. Biederman | 390e2ff | 2006-03-31 02:31:33 -0800 | [diff] [blame] | 1056 | /* Fail if I am already a session leader */ |
| 1057 | if (group_leader->signal->leader) |
| 1058 | goto out; |
| 1059 | |
| 1060 | session = group_leader->pid; |
| 1061 | /* Fail if a process group id already exists that equals the |
| 1062 | * proposed session id. |
| 1063 | * |
| 1064 | * Don't check if session id == 1 because kernel threads use this |
| 1065 | * session id and so the check will always fail and make it so |
| 1066 | * init cannot successfully call setsid. |
| 1067 | */ |
Pavel Emelyanov | 228ebcb | 2007-10-18 23:40:16 -0700 | [diff] [blame] | 1068 | if (session > 1 && find_task_by_pid_type_ns(PIDTYPE_PGID, |
| 1069 | session, &init_pid_ns)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1070 | goto out; |
| 1071 | |
Oren Laadan | e19f247 | 2006-01-08 01:03:58 -0800 | [diff] [blame] | 1072 | group_leader->signal->leader = 1; |
Eric W. Biederman | 390e2ff | 2006-03-31 02:31:33 -0800 | [diff] [blame] | 1073 | __set_special_pids(session, session); |
Peter Zijlstra | 24ec839 | 2006-12-08 02:36:04 -0800 | [diff] [blame] | 1074 | |
| 1075 | spin_lock(&group_leader->sighand->siglock); |
Oren Laadan | e19f247 | 2006-01-08 01:03:58 -0800 | [diff] [blame] | 1076 | group_leader->signal->tty = NULL; |
Peter Zijlstra | 24ec839 | 2006-12-08 02:36:04 -0800 | [diff] [blame] | 1077 | spin_unlock(&group_leader->sighand->siglock); |
| 1078 | |
Pavel Emelyanov | b488893 | 2007-10-18 23:40:14 -0700 | [diff] [blame] | 1079 | err = task_pgrp_vnr(group_leader); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1080 | out: |
| 1081 | write_unlock_irq(&tasklist_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1082 | return err; |
| 1083 | } |
| 1084 | |
| 1085 | /* |
| 1086 | * Supplementary group IDs |
| 1087 | */ |
| 1088 | |
| 1089 | /* init to 2 - one for init_task, one to ensure it is never freed */ |
| 1090 | struct group_info init_groups = { .usage = ATOMIC_INIT(2) }; |
| 1091 | |
| 1092 | struct group_info *groups_alloc(int gidsetsize) |
| 1093 | { |
| 1094 | struct group_info *group_info; |
| 1095 | int nblocks; |
| 1096 | int i; |
| 1097 | |
| 1098 | nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK; |
| 1099 | /* Make sure we always allocate at least one indirect block pointer */ |
| 1100 | nblocks = nblocks ? : 1; |
| 1101 | group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER); |
| 1102 | if (!group_info) |
| 1103 | return NULL; |
| 1104 | group_info->ngroups = gidsetsize; |
| 1105 | group_info->nblocks = nblocks; |
| 1106 | atomic_set(&group_info->usage, 1); |
| 1107 | |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1108 | if (gidsetsize <= NGROUPS_SMALL) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1109 | group_info->blocks[0] = group_info->small_block; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1110 | else { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1111 | for (i = 0; i < nblocks; i++) { |
| 1112 | gid_t *b; |
| 1113 | b = (void *)__get_free_page(GFP_USER); |
| 1114 | if (!b) |
| 1115 | goto out_undo_partial_alloc; |
| 1116 | group_info->blocks[i] = b; |
| 1117 | } |
| 1118 | } |
| 1119 | return group_info; |
| 1120 | |
| 1121 | out_undo_partial_alloc: |
| 1122 | while (--i >= 0) { |
| 1123 | free_page((unsigned long)group_info->blocks[i]); |
| 1124 | } |
| 1125 | kfree(group_info); |
| 1126 | return NULL; |
| 1127 | } |
| 1128 | |
| 1129 | EXPORT_SYMBOL(groups_alloc); |
| 1130 | |
| 1131 | void groups_free(struct group_info *group_info) |
| 1132 | { |
| 1133 | if (group_info->blocks[0] != group_info->small_block) { |
| 1134 | int i; |
| 1135 | for (i = 0; i < group_info->nblocks; i++) |
| 1136 | free_page((unsigned long)group_info->blocks[i]); |
| 1137 | } |
| 1138 | kfree(group_info); |
| 1139 | } |
| 1140 | |
| 1141 | EXPORT_SYMBOL(groups_free); |
| 1142 | |
| 1143 | /* export the group_info to a user-space array */ |
| 1144 | static int groups_to_user(gid_t __user *grouplist, |
| 1145 | struct group_info *group_info) |
| 1146 | { |
| 1147 | int i; |
| 1148 | int count = group_info->ngroups; |
| 1149 | |
| 1150 | for (i = 0; i < group_info->nblocks; i++) { |
| 1151 | int cp_count = min(NGROUPS_PER_BLOCK, count); |
| 1152 | int off = i * NGROUPS_PER_BLOCK; |
| 1153 | int len = cp_count * sizeof(*grouplist); |
| 1154 | |
| 1155 | if (copy_to_user(grouplist+off, group_info->blocks[i], len)) |
| 1156 | return -EFAULT; |
| 1157 | |
| 1158 | count -= cp_count; |
| 1159 | } |
| 1160 | return 0; |
| 1161 | } |
| 1162 | |
| 1163 | /* fill a group_info from a user-space array - it must be allocated already */ |
| 1164 | static int groups_from_user(struct group_info *group_info, |
| 1165 | gid_t __user *grouplist) |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1166 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1167 | int i; |
| 1168 | int count = group_info->ngroups; |
| 1169 | |
| 1170 | for (i = 0; i < group_info->nblocks; i++) { |
| 1171 | int cp_count = min(NGROUPS_PER_BLOCK, count); |
| 1172 | int off = i * NGROUPS_PER_BLOCK; |
| 1173 | int len = cp_count * sizeof(*grouplist); |
| 1174 | |
| 1175 | if (copy_from_user(group_info->blocks[i], grouplist+off, len)) |
| 1176 | return -EFAULT; |
| 1177 | |
| 1178 | count -= cp_count; |
| 1179 | } |
| 1180 | return 0; |
| 1181 | } |
| 1182 | |
Domen Puncer | ebe8b54 | 2005-05-05 16:16:19 -0700 | [diff] [blame] | 1183 | /* a simple Shell sort */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1184 | static void groups_sort(struct group_info *group_info) |
| 1185 | { |
| 1186 | int base, max, stride; |
| 1187 | int gidsetsize = group_info->ngroups; |
| 1188 | |
| 1189 | for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1) |
| 1190 | ; /* nothing */ |
| 1191 | stride /= 3; |
| 1192 | |
| 1193 | while (stride) { |
| 1194 | max = gidsetsize - stride; |
| 1195 | for (base = 0; base < max; base++) { |
| 1196 | int left = base; |
| 1197 | int right = left + stride; |
| 1198 | gid_t tmp = GROUP_AT(group_info, right); |
| 1199 | |
| 1200 | while (left >= 0 && GROUP_AT(group_info, left) > tmp) { |
| 1201 | GROUP_AT(group_info, right) = |
| 1202 | GROUP_AT(group_info, left); |
| 1203 | right = left; |
| 1204 | left -= stride; |
| 1205 | } |
| 1206 | GROUP_AT(group_info, right) = tmp; |
| 1207 | } |
| 1208 | stride /= 3; |
| 1209 | } |
| 1210 | } |
| 1211 | |
| 1212 | /* a simple bsearch */ |
David Howells | 3e30148 | 2005-06-23 22:00:56 -0700 | [diff] [blame] | 1213 | int groups_search(struct group_info *group_info, gid_t grp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1214 | { |
Eric Dumazet | d74beb9f | 2006-03-25 03:08:19 -0800 | [diff] [blame] | 1215 | unsigned int left, right; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1216 | |
| 1217 | if (!group_info) |
| 1218 | return 0; |
| 1219 | |
| 1220 | left = 0; |
| 1221 | right = group_info->ngroups; |
| 1222 | while (left < right) { |
Eric Dumazet | d74beb9f | 2006-03-25 03:08:19 -0800 | [diff] [blame] | 1223 | unsigned int mid = (left+right)/2; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1224 | int cmp = grp - GROUP_AT(group_info, mid); |
| 1225 | if (cmp > 0) |
| 1226 | left = mid + 1; |
| 1227 | else if (cmp < 0) |
| 1228 | right = mid; |
| 1229 | else |
| 1230 | return 1; |
| 1231 | } |
| 1232 | return 0; |
| 1233 | } |
| 1234 | |
| 1235 | /* validate and set current->group_info */ |
| 1236 | int set_current_groups(struct group_info *group_info) |
| 1237 | { |
| 1238 | int retval; |
| 1239 | struct group_info *old_info; |
| 1240 | |
| 1241 | retval = security_task_setgroups(group_info); |
| 1242 | if (retval) |
| 1243 | return retval; |
| 1244 | |
| 1245 | groups_sort(group_info); |
| 1246 | get_group_info(group_info); |
| 1247 | |
| 1248 | task_lock(current); |
| 1249 | old_info = current->group_info; |
| 1250 | current->group_info = group_info; |
| 1251 | task_unlock(current); |
| 1252 | |
| 1253 | put_group_info(old_info); |
| 1254 | |
| 1255 | return 0; |
| 1256 | } |
| 1257 | |
| 1258 | EXPORT_SYMBOL(set_current_groups); |
| 1259 | |
| 1260 | asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist) |
| 1261 | { |
| 1262 | int i = 0; |
| 1263 | |
| 1264 | /* |
| 1265 | * SMP: Nobody else can change our grouplist. Thus we are |
| 1266 | * safe. |
| 1267 | */ |
| 1268 | |
| 1269 | if (gidsetsize < 0) |
| 1270 | return -EINVAL; |
| 1271 | |
| 1272 | /* no need to grab task_lock here; it cannot change */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1273 | i = current->group_info->ngroups; |
| 1274 | if (gidsetsize) { |
| 1275 | if (i > gidsetsize) { |
| 1276 | i = -EINVAL; |
| 1277 | goto out; |
| 1278 | } |
| 1279 | if (groups_to_user(grouplist, current->group_info)) { |
| 1280 | i = -EFAULT; |
| 1281 | goto out; |
| 1282 | } |
| 1283 | } |
| 1284 | out: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1285 | return i; |
| 1286 | } |
| 1287 | |
| 1288 | /* |
| 1289 | * SMP: Our groups are copy-on-write. We can set them safely |
| 1290 | * without another task interfering. |
| 1291 | */ |
| 1292 | |
| 1293 | asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist) |
| 1294 | { |
| 1295 | struct group_info *group_info; |
| 1296 | int retval; |
| 1297 | |
| 1298 | if (!capable(CAP_SETGID)) |
| 1299 | return -EPERM; |
| 1300 | if ((unsigned)gidsetsize > NGROUPS_MAX) |
| 1301 | return -EINVAL; |
| 1302 | |
| 1303 | group_info = groups_alloc(gidsetsize); |
| 1304 | if (!group_info) |
| 1305 | return -ENOMEM; |
| 1306 | retval = groups_from_user(group_info, grouplist); |
| 1307 | if (retval) { |
| 1308 | put_group_info(group_info); |
| 1309 | return retval; |
| 1310 | } |
| 1311 | |
| 1312 | retval = set_current_groups(group_info); |
| 1313 | put_group_info(group_info); |
| 1314 | |
| 1315 | return retval; |
| 1316 | } |
| 1317 | |
| 1318 | /* |
| 1319 | * Check whether we're fsgid/egid or in the supplemental group.. |
| 1320 | */ |
| 1321 | int in_group_p(gid_t grp) |
| 1322 | { |
| 1323 | int retval = 1; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1324 | if (grp != current->fsgid) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1325 | retval = groups_search(current->group_info, grp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1326 | return retval; |
| 1327 | } |
| 1328 | |
| 1329 | EXPORT_SYMBOL(in_group_p); |
| 1330 | |
| 1331 | int in_egroup_p(gid_t grp) |
| 1332 | { |
| 1333 | int retval = 1; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1334 | if (grp != current->egid) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1335 | retval = groups_search(current->group_info, grp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1336 | return retval; |
| 1337 | } |
| 1338 | |
| 1339 | EXPORT_SYMBOL(in_egroup_p); |
| 1340 | |
| 1341 | DECLARE_RWSEM(uts_sem); |
| 1342 | |
David S. Miller | 393b072 | 2005-11-10 12:47:50 -0800 | [diff] [blame] | 1343 | EXPORT_SYMBOL(uts_sem); |
| 1344 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1345 | asmlinkage long sys_newuname(struct new_utsname __user * name) |
| 1346 | { |
| 1347 | int errno = 0; |
| 1348 | |
| 1349 | down_read(&uts_sem); |
Serge E. Hallyn | e9ff399 | 2006-10-02 02:18:11 -0700 | [diff] [blame] | 1350 | if (copy_to_user(name, utsname(), sizeof *name)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1351 | errno = -EFAULT; |
| 1352 | up_read(&uts_sem); |
| 1353 | return errno; |
| 1354 | } |
| 1355 | |
| 1356 | asmlinkage long sys_sethostname(char __user *name, int len) |
| 1357 | { |
| 1358 | int errno; |
| 1359 | char tmp[__NEW_UTS_LEN]; |
| 1360 | |
| 1361 | if (!capable(CAP_SYS_ADMIN)) |
| 1362 | return -EPERM; |
| 1363 | if (len < 0 || len > __NEW_UTS_LEN) |
| 1364 | return -EINVAL; |
| 1365 | down_write(&uts_sem); |
| 1366 | errno = -EFAULT; |
| 1367 | if (!copy_from_user(tmp, name, len)) { |
Serge E. Hallyn | e9ff399 | 2006-10-02 02:18:11 -0700 | [diff] [blame] | 1368 | memcpy(utsname()->nodename, tmp, len); |
| 1369 | utsname()->nodename[len] = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1370 | errno = 0; |
| 1371 | } |
| 1372 | up_write(&uts_sem); |
| 1373 | return errno; |
| 1374 | } |
| 1375 | |
| 1376 | #ifdef __ARCH_WANT_SYS_GETHOSTNAME |
| 1377 | |
| 1378 | asmlinkage long sys_gethostname(char __user *name, int len) |
| 1379 | { |
| 1380 | int i, errno; |
| 1381 | |
| 1382 | if (len < 0) |
| 1383 | return -EINVAL; |
| 1384 | down_read(&uts_sem); |
Serge E. Hallyn | e9ff399 | 2006-10-02 02:18:11 -0700 | [diff] [blame] | 1385 | i = 1 + strlen(utsname()->nodename); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1386 | if (i > len) |
| 1387 | i = len; |
| 1388 | errno = 0; |
Serge E. Hallyn | e9ff399 | 2006-10-02 02:18:11 -0700 | [diff] [blame] | 1389 | if (copy_to_user(name, utsname()->nodename, i)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1390 | errno = -EFAULT; |
| 1391 | up_read(&uts_sem); |
| 1392 | return errno; |
| 1393 | } |
| 1394 | |
| 1395 | #endif |
| 1396 | |
| 1397 | /* |
| 1398 | * Only setdomainname; getdomainname can be implemented by calling |
| 1399 | * uname() |
| 1400 | */ |
| 1401 | asmlinkage long sys_setdomainname(char __user *name, int len) |
| 1402 | { |
| 1403 | int errno; |
| 1404 | char tmp[__NEW_UTS_LEN]; |
| 1405 | |
| 1406 | if (!capable(CAP_SYS_ADMIN)) |
| 1407 | return -EPERM; |
| 1408 | if (len < 0 || len > __NEW_UTS_LEN) |
| 1409 | return -EINVAL; |
| 1410 | |
| 1411 | down_write(&uts_sem); |
| 1412 | errno = -EFAULT; |
| 1413 | if (!copy_from_user(tmp, name, len)) { |
Serge E. Hallyn | e9ff399 | 2006-10-02 02:18:11 -0700 | [diff] [blame] | 1414 | memcpy(utsname()->domainname, tmp, len); |
| 1415 | utsname()->domainname[len] = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1416 | errno = 0; |
| 1417 | } |
| 1418 | up_write(&uts_sem); |
| 1419 | return errno; |
| 1420 | } |
| 1421 | |
| 1422 | asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim) |
| 1423 | { |
| 1424 | if (resource >= RLIM_NLIMITS) |
| 1425 | return -EINVAL; |
| 1426 | else { |
| 1427 | struct rlimit value; |
| 1428 | task_lock(current->group_leader); |
| 1429 | value = current->signal->rlim[resource]; |
| 1430 | task_unlock(current->group_leader); |
| 1431 | return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0; |
| 1432 | } |
| 1433 | } |
| 1434 | |
| 1435 | #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT |
| 1436 | |
| 1437 | /* |
| 1438 | * Back compatibility for getrlimit. Needed for some apps. |
| 1439 | */ |
| 1440 | |
| 1441 | asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim) |
| 1442 | { |
| 1443 | struct rlimit x; |
| 1444 | if (resource >= RLIM_NLIMITS) |
| 1445 | return -EINVAL; |
| 1446 | |
| 1447 | task_lock(current->group_leader); |
| 1448 | x = current->signal->rlim[resource]; |
| 1449 | task_unlock(current->group_leader); |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1450 | if (x.rlim_cur > 0x7FFFFFFF) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1451 | x.rlim_cur = 0x7FFFFFFF; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1452 | if (x.rlim_max > 0x7FFFFFFF) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1453 | x.rlim_max = 0x7FFFFFFF; |
| 1454 | return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0; |
| 1455 | } |
| 1456 | |
| 1457 | #endif |
| 1458 | |
| 1459 | asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim) |
| 1460 | { |
| 1461 | struct rlimit new_rlim, *old_rlim; |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 1462 | unsigned long it_prof_secs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1463 | int retval; |
| 1464 | |
| 1465 | if (resource >= RLIM_NLIMITS) |
| 1466 | return -EINVAL; |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 1467 | if (copy_from_user(&new_rlim, rlim, sizeof(*rlim))) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1468 | return -EFAULT; |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 1469 | if (new_rlim.rlim_cur > new_rlim.rlim_max) |
| 1470 | return -EINVAL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1471 | old_rlim = current->signal->rlim + resource; |
| 1472 | if ((new_rlim.rlim_max > old_rlim->rlim_max) && |
| 1473 | !capable(CAP_SYS_RESOURCE)) |
| 1474 | return -EPERM; |
| 1475 | if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN) |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 1476 | return -EPERM; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1477 | |
| 1478 | retval = security_task_setrlimit(resource, &new_rlim); |
| 1479 | if (retval) |
| 1480 | return retval; |
| 1481 | |
Tom Alsberg | 9926e4c | 2007-05-08 00:30:31 -0700 | [diff] [blame] | 1482 | if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) { |
| 1483 | /* |
| 1484 | * The caller is asking for an immediate RLIMIT_CPU |
| 1485 | * expiry. But we use the zero value to mean "it was |
| 1486 | * never set". So let's cheat and make it one second |
| 1487 | * instead |
| 1488 | */ |
| 1489 | new_rlim.rlim_cur = 1; |
| 1490 | } |
| 1491 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1492 | task_lock(current->group_leader); |
| 1493 | *old_rlim = new_rlim; |
| 1494 | task_unlock(current->group_leader); |
| 1495 | |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 1496 | if (resource != RLIMIT_CPU) |
| 1497 | goto out; |
Andrew Morton | d3561f7 | 2006-03-24 03:18:36 -0800 | [diff] [blame] | 1498 | |
| 1499 | /* |
| 1500 | * RLIMIT_CPU handling. Note that the kernel fails to return an error |
| 1501 | * code if it rejected the user's attempt to set RLIMIT_CPU. This is a |
| 1502 | * very long-standing error, and fixing it now risks breakage of |
| 1503 | * applications, so we live with it |
| 1504 | */ |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 1505 | if (new_rlim.rlim_cur == RLIM_INFINITY) |
| 1506 | goto out; |
| 1507 | |
| 1508 | it_prof_secs = cputime_to_secs(current->signal->it_prof_expires); |
| 1509 | if (it_prof_secs == 0 || new_rlim.rlim_cur <= it_prof_secs) { |
Andrew Morton | e066111 | 2006-03-24 03:18:35 -0800 | [diff] [blame] | 1510 | unsigned long rlim_cur = new_rlim.rlim_cur; |
| 1511 | cputime_t cputime; |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 1512 | |
Andrew Morton | e066111 | 2006-03-24 03:18:35 -0800 | [diff] [blame] | 1513 | cputime = secs_to_cputime(rlim_cur); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1514 | read_lock(&tasklist_lock); |
| 1515 | spin_lock_irq(¤t->sighand->siglock); |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 1516 | set_process_cpu_timer(current, CPUCLOCK_PROF, &cputime, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1517 | spin_unlock_irq(¤t->sighand->siglock); |
| 1518 | read_unlock(&tasklist_lock); |
| 1519 | } |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 1520 | out: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1521 | return 0; |
| 1522 | } |
| 1523 | |
| 1524 | /* |
| 1525 | * It would make sense to put struct rusage in the task_struct, |
| 1526 | * except that would make the task_struct be *really big*. After |
| 1527 | * task_struct gets moved into malloc'ed memory, it would |
| 1528 | * make sense to do this. It will make moving the rest of the information |
| 1529 | * a lot simpler! (Which we're not doing right now because we're not |
| 1530 | * measuring them yet). |
| 1531 | * |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1532 | * When sampling multiple threads for RUSAGE_SELF, under SMP we might have |
| 1533 | * races with threads incrementing their own counters. But since word |
| 1534 | * reads are atomic, we either get new values or old values and we don't |
| 1535 | * care which for the sums. We always take the siglock to protect reading |
| 1536 | * the c* fields from p->signal from races with exit.c updating those |
| 1537 | * fields when reaping, so a sample either gets all the additions of a |
| 1538 | * given child after it's reaped, or none so this sample is before reaping. |
Ravikiran G Thirumalai | 2dd0ebc | 2006-03-23 03:00:13 -0800 | [diff] [blame] | 1539 | * |
Ravikiran G Thirumalai | de047c1b | 2006-06-22 14:47:26 -0700 | [diff] [blame] | 1540 | * Locking: |
| 1541 | * We need to take the siglock for CHILDEREN, SELF and BOTH |
| 1542 | * for the cases current multithreaded, non-current single threaded |
| 1543 | * non-current multithreaded. Thread traversal is now safe with |
| 1544 | * the siglock held. |
| 1545 | * Strictly speaking, we donot need to take the siglock if we are current and |
| 1546 | * single threaded, as no one else can take our signal_struct away, no one |
| 1547 | * else can reap the children to update signal->c* counters, and no one else |
| 1548 | * can race with the signal-> fields. If we do not take any lock, the |
| 1549 | * signal-> fields could be read out of order while another thread was just |
| 1550 | * exiting. So we should place a read memory barrier when we avoid the lock. |
| 1551 | * On the writer side, write memory barrier is implied in __exit_signal |
| 1552 | * as __exit_signal releases the siglock spinlock after updating the signal-> |
| 1553 | * fields. But we don't do this yet to keep things simple. |
Ravikiran G Thirumalai | 2dd0ebc | 2006-03-23 03:00:13 -0800 | [diff] [blame] | 1554 | * |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1555 | */ |
| 1556 | |
| 1557 | static void k_getrusage(struct task_struct *p, int who, struct rusage *r) |
| 1558 | { |
| 1559 | struct task_struct *t; |
| 1560 | unsigned long flags; |
| 1561 | cputime_t utime, stime; |
| 1562 | |
| 1563 | memset((char *) r, 0, sizeof *r); |
Oleg Nesterov | 0f59cc4 | 2006-01-08 01:05:15 -0800 | [diff] [blame] | 1564 | utime = stime = cputime_zero; |
| 1565 | |
Ravikiran G Thirumalai | de047c1b | 2006-06-22 14:47:26 -0700 | [diff] [blame] | 1566 | rcu_read_lock(); |
| 1567 | if (!lock_task_sighand(p, &flags)) { |
| 1568 | rcu_read_unlock(); |
| 1569 | return; |
| 1570 | } |
Ravikiran G Thirumalai | 2dd0ebc | 2006-03-23 03:00:13 -0800 | [diff] [blame] | 1571 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1572 | switch (who) { |
Oleg Nesterov | 0f59cc4 | 2006-01-08 01:05:15 -0800 | [diff] [blame] | 1573 | case RUSAGE_BOTH: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1574 | case RUSAGE_CHILDREN: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1575 | utime = p->signal->cutime; |
| 1576 | stime = p->signal->cstime; |
| 1577 | r->ru_nvcsw = p->signal->cnvcsw; |
| 1578 | r->ru_nivcsw = p->signal->cnivcsw; |
| 1579 | r->ru_minflt = p->signal->cmin_flt; |
| 1580 | r->ru_majflt = p->signal->cmaj_flt; |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 1581 | r->ru_inblock = p->signal->cinblock; |
| 1582 | r->ru_oublock = p->signal->coublock; |
Oleg Nesterov | 0f59cc4 | 2006-01-08 01:05:15 -0800 | [diff] [blame] | 1583 | |
| 1584 | if (who == RUSAGE_CHILDREN) |
| 1585 | break; |
| 1586 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1587 | case RUSAGE_SELF: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1588 | utime = cputime_add(utime, p->signal->utime); |
| 1589 | stime = cputime_add(stime, p->signal->stime); |
| 1590 | r->ru_nvcsw += p->signal->nvcsw; |
| 1591 | r->ru_nivcsw += p->signal->nivcsw; |
| 1592 | r->ru_minflt += p->signal->min_flt; |
| 1593 | r->ru_majflt += p->signal->maj_flt; |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 1594 | r->ru_inblock += p->signal->inblock; |
| 1595 | r->ru_oublock += p->signal->oublock; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1596 | t = p; |
| 1597 | do { |
| 1598 | utime = cputime_add(utime, t->utime); |
| 1599 | stime = cputime_add(stime, t->stime); |
| 1600 | r->ru_nvcsw += t->nvcsw; |
| 1601 | r->ru_nivcsw += t->nivcsw; |
| 1602 | r->ru_minflt += t->min_flt; |
| 1603 | r->ru_majflt += t->maj_flt; |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 1604 | r->ru_inblock += task_io_get_inblock(t); |
| 1605 | r->ru_oublock += task_io_get_oublock(t); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1606 | t = next_thread(t); |
| 1607 | } while (t != p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1608 | break; |
Oleg Nesterov | 0f59cc4 | 2006-01-08 01:05:15 -0800 | [diff] [blame] | 1609 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1610 | default: |
| 1611 | BUG(); |
| 1612 | } |
Oleg Nesterov | 0f59cc4 | 2006-01-08 01:05:15 -0800 | [diff] [blame] | 1613 | |
Ravikiran G Thirumalai | de047c1b | 2006-06-22 14:47:26 -0700 | [diff] [blame] | 1614 | unlock_task_sighand(p, &flags); |
| 1615 | rcu_read_unlock(); |
| 1616 | |
Oleg Nesterov | 0f59cc4 | 2006-01-08 01:05:15 -0800 | [diff] [blame] | 1617 | cputime_to_timeval(utime, &r->ru_utime); |
| 1618 | cputime_to_timeval(stime, &r->ru_stime); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1619 | } |
| 1620 | |
| 1621 | int getrusage(struct task_struct *p, int who, struct rusage __user *ru) |
| 1622 | { |
| 1623 | struct rusage r; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1624 | k_getrusage(p, who, &r); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1625 | return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0; |
| 1626 | } |
| 1627 | |
| 1628 | asmlinkage long sys_getrusage(int who, struct rusage __user *ru) |
| 1629 | { |
| 1630 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN) |
| 1631 | return -EINVAL; |
| 1632 | return getrusage(current, who, ru); |
| 1633 | } |
| 1634 | |
| 1635 | asmlinkage long sys_umask(int mask) |
| 1636 | { |
| 1637 | mask = xchg(¤t->fs->umask, mask & S_IRWXUGO); |
| 1638 | return mask; |
| 1639 | } |
| 1640 | |
| 1641 | asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3, |
| 1642 | unsigned long arg4, unsigned long arg5) |
| 1643 | { |
| 1644 | long error; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1645 | |
| 1646 | error = security_task_prctl(option, arg2, arg3, arg4, arg5); |
| 1647 | if (error) |
| 1648 | return error; |
| 1649 | |
| 1650 | switch (option) { |
| 1651 | case PR_SET_PDEATHSIG: |
Jesper Juhl | 0730ded | 2005-09-06 15:17:37 -0700 | [diff] [blame] | 1652 | if (!valid_signal(arg2)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1653 | error = -EINVAL; |
| 1654 | break; |
| 1655 | } |
Jesper Juhl | 0730ded | 2005-09-06 15:17:37 -0700 | [diff] [blame] | 1656 | current->pdeath_signal = arg2; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1657 | break; |
| 1658 | case PR_GET_PDEATHSIG: |
| 1659 | error = put_user(current->pdeath_signal, (int __user *)arg2); |
| 1660 | break; |
| 1661 | case PR_GET_DUMPABLE: |
Kawai, Hidehiro | 6c5d523 | 2007-07-19 01:48:27 -0700 | [diff] [blame] | 1662 | error = get_dumpable(current->mm); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1663 | break; |
| 1664 | case PR_SET_DUMPABLE: |
Marcel Holtmann | abf75a5 | 2006-07-12 13:12:00 +0200 | [diff] [blame] | 1665 | if (arg2 < 0 || arg2 > 1) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1666 | error = -EINVAL; |
| 1667 | break; |
| 1668 | } |
Kawai, Hidehiro | 6c5d523 | 2007-07-19 01:48:27 -0700 | [diff] [blame] | 1669 | set_dumpable(current->mm, arg2); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1670 | break; |
| 1671 | |
| 1672 | case PR_SET_UNALIGN: |
| 1673 | error = SET_UNALIGN_CTL(current, arg2); |
| 1674 | break; |
| 1675 | case PR_GET_UNALIGN: |
| 1676 | error = GET_UNALIGN_CTL(current, arg2); |
| 1677 | break; |
| 1678 | case PR_SET_FPEMU: |
| 1679 | error = SET_FPEMU_CTL(current, arg2); |
| 1680 | break; |
| 1681 | case PR_GET_FPEMU: |
| 1682 | error = GET_FPEMU_CTL(current, arg2); |
| 1683 | break; |
| 1684 | case PR_SET_FPEXC: |
| 1685 | error = SET_FPEXC_CTL(current, arg2); |
| 1686 | break; |
| 1687 | case PR_GET_FPEXC: |
| 1688 | error = GET_FPEXC_CTL(current, arg2); |
| 1689 | break; |
| 1690 | case PR_GET_TIMING: |
| 1691 | error = PR_TIMING_STATISTICAL; |
| 1692 | break; |
| 1693 | case PR_SET_TIMING: |
| 1694 | if (arg2 == PR_TIMING_STATISTICAL) |
| 1695 | error = 0; |
| 1696 | else |
| 1697 | error = -EINVAL; |
| 1698 | break; |
| 1699 | |
| 1700 | case PR_GET_KEEPCAPS: |
| 1701 | if (current->keep_capabilities) |
| 1702 | error = 1; |
| 1703 | break; |
| 1704 | case PR_SET_KEEPCAPS: |
| 1705 | if (arg2 != 0 && arg2 != 1) { |
| 1706 | error = -EINVAL; |
| 1707 | break; |
| 1708 | } |
| 1709 | current->keep_capabilities = arg2; |
| 1710 | break; |
| 1711 | case PR_SET_NAME: { |
| 1712 | struct task_struct *me = current; |
| 1713 | unsigned char ncomm[sizeof(me->comm)]; |
| 1714 | |
| 1715 | ncomm[sizeof(me->comm)-1] = 0; |
| 1716 | if (strncpy_from_user(ncomm, (char __user *)arg2, |
| 1717 | sizeof(me->comm)-1) < 0) |
| 1718 | return -EFAULT; |
| 1719 | set_task_comm(me, ncomm); |
| 1720 | return 0; |
| 1721 | } |
| 1722 | case PR_GET_NAME: { |
| 1723 | struct task_struct *me = current; |
| 1724 | unsigned char tcomm[sizeof(me->comm)]; |
| 1725 | |
| 1726 | get_task_comm(tcomm, me); |
| 1727 | if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm))) |
| 1728 | return -EFAULT; |
| 1729 | return 0; |
| 1730 | } |
Anton Blanchard | 651d765 | 2006-06-07 16:10:19 +1000 | [diff] [blame] | 1731 | case PR_GET_ENDIAN: |
| 1732 | error = GET_ENDIAN(current, arg2); |
| 1733 | break; |
| 1734 | case PR_SET_ENDIAN: |
| 1735 | error = SET_ENDIAN(current, arg2); |
| 1736 | break; |
| 1737 | |
Andrea Arcangeli | 1d9d02f | 2007-07-15 23:41:32 -0700 | [diff] [blame] | 1738 | case PR_GET_SECCOMP: |
| 1739 | error = prctl_get_seccomp(); |
| 1740 | break; |
| 1741 | case PR_SET_SECCOMP: |
| 1742 | error = prctl_set_seccomp(arg2); |
| 1743 | break; |
| 1744 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1745 | default: |
| 1746 | error = -EINVAL; |
| 1747 | break; |
| 1748 | } |
| 1749 | return error; |
| 1750 | } |
Andi Kleen | 3cfc348 | 2006-09-26 10:52:28 +0200 | [diff] [blame] | 1751 | |
| 1752 | asmlinkage long sys_getcpu(unsigned __user *cpup, unsigned __user *nodep, |
Ingo Molnar | 4307d1e | 2007-11-07 18:37:48 +0100 | [diff] [blame] | 1753 | struct getcpu_cache __user *unused) |
Andi Kleen | 3cfc348 | 2006-09-26 10:52:28 +0200 | [diff] [blame] | 1754 | { |
| 1755 | int err = 0; |
| 1756 | int cpu = raw_smp_processor_id(); |
| 1757 | if (cpup) |
| 1758 | err |= put_user(cpu, cpup); |
| 1759 | if (nodep) |
| 1760 | err |= put_user(cpu_to_node(cpu), nodep); |
Andi Kleen | 3cfc348 | 2006-09-26 10:52:28 +0200 | [diff] [blame] | 1761 | return err ? -EFAULT : 0; |
| 1762 | } |
Jeremy Fitzhardinge | 10a0a8d | 2007-07-17 18:37:02 -0700 | [diff] [blame] | 1763 | |
| 1764 | char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff"; |
| 1765 | |
| 1766 | static void argv_cleanup(char **argv, char **envp) |
| 1767 | { |
| 1768 | argv_free(argv); |
| 1769 | } |
| 1770 | |
| 1771 | /** |
| 1772 | * orderly_poweroff - Trigger an orderly system poweroff |
| 1773 | * @force: force poweroff if command execution fails |
| 1774 | * |
| 1775 | * This may be called from any context to trigger a system shutdown. |
| 1776 | * If the orderly shutdown fails, it will force an immediate shutdown. |
| 1777 | */ |
| 1778 | int orderly_poweroff(bool force) |
| 1779 | { |
| 1780 | int argc; |
| 1781 | char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc); |
| 1782 | static char *envp[] = { |
| 1783 | "HOME=/", |
| 1784 | "PATH=/sbin:/bin:/usr/sbin:/usr/bin", |
| 1785 | NULL |
| 1786 | }; |
| 1787 | int ret = -ENOMEM; |
| 1788 | struct subprocess_info *info; |
| 1789 | |
| 1790 | if (argv == NULL) { |
| 1791 | printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n", |
| 1792 | __func__, poweroff_cmd); |
| 1793 | goto out; |
| 1794 | } |
| 1795 | |
| 1796 | info = call_usermodehelper_setup(argv[0], argv, envp); |
| 1797 | if (info == NULL) { |
| 1798 | argv_free(argv); |
| 1799 | goto out; |
| 1800 | } |
| 1801 | |
| 1802 | call_usermodehelper_setcleanup(info, argv_cleanup); |
| 1803 | |
Jeremy Fitzhardinge | 86313c4 | 2007-07-17 18:37:03 -0700 | [diff] [blame] | 1804 | ret = call_usermodehelper_exec(info, UMH_NO_WAIT); |
Jeremy Fitzhardinge | 10a0a8d | 2007-07-17 18:37:02 -0700 | [diff] [blame] | 1805 | |
| 1806 | out: |
| 1807 | if (ret && force) { |
| 1808 | printk(KERN_WARNING "Failed to start orderly shutdown: " |
| 1809 | "forcing the issue\n"); |
| 1810 | |
| 1811 | /* I guess this should try to kick off some daemon to |
| 1812 | sync and poweroff asap. Or not even bother syncing |
| 1813 | if we're doing an emergency shutdown? */ |
| 1814 | emergency_sync(); |
| 1815 | kernel_power_off(); |
| 1816 | } |
| 1817 | |
| 1818 | return ret; |
| 1819 | } |
| 1820 | EXPORT_SYMBOL_GPL(orderly_poweroff); |