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> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 35 | |
| 36 | #include <linux/compat.h> |
| 37 | #include <linux/syscalls.h> |
Keshavamurthy Anil S | 00d7c05 | 2005-12-12 00:37:33 -0800 | [diff] [blame] | 38 | #include <linux/kprobes.h> |
Cedric Le Goater | acce292 | 2007-07-15 23:40:59 -0700 | [diff] [blame] | 39 | #include <linux/user_namespace.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 40 | |
| 41 | #include <asm/uaccess.h> |
| 42 | #include <asm/io.h> |
| 43 | #include <asm/unistd.h> |
| 44 | |
| 45 | #ifndef SET_UNALIGN_CTL |
| 46 | # define SET_UNALIGN_CTL(a,b) (-EINVAL) |
| 47 | #endif |
| 48 | #ifndef GET_UNALIGN_CTL |
| 49 | # define GET_UNALIGN_CTL(a,b) (-EINVAL) |
| 50 | #endif |
| 51 | #ifndef SET_FPEMU_CTL |
| 52 | # define SET_FPEMU_CTL(a,b) (-EINVAL) |
| 53 | #endif |
| 54 | #ifndef GET_FPEMU_CTL |
| 55 | # define GET_FPEMU_CTL(a,b) (-EINVAL) |
| 56 | #endif |
| 57 | #ifndef SET_FPEXC_CTL |
| 58 | # define SET_FPEXC_CTL(a,b) (-EINVAL) |
| 59 | #endif |
| 60 | #ifndef GET_FPEXC_CTL |
| 61 | # define GET_FPEXC_CTL(a,b) (-EINVAL) |
| 62 | #endif |
Anton Blanchard | 651d765 | 2006-06-07 16:10:19 +1000 | [diff] [blame] | 63 | #ifndef GET_ENDIAN |
| 64 | # define GET_ENDIAN(a,b) (-EINVAL) |
| 65 | #endif |
| 66 | #ifndef SET_ENDIAN |
| 67 | # define SET_ENDIAN(a,b) (-EINVAL) |
| 68 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 69 | |
| 70 | /* |
| 71 | * this is where the system-wide overflow UID and GID are defined, for |
| 72 | * architectures that now have 32-bit UID/GID but didn't in the past |
| 73 | */ |
| 74 | |
| 75 | int overflowuid = DEFAULT_OVERFLOWUID; |
| 76 | int overflowgid = DEFAULT_OVERFLOWGID; |
| 77 | |
| 78 | #ifdef CONFIG_UID16 |
| 79 | EXPORT_SYMBOL(overflowuid); |
| 80 | EXPORT_SYMBOL(overflowgid); |
| 81 | #endif |
| 82 | |
| 83 | /* |
| 84 | * the same as above, but for filesystems which can only store a 16-bit |
| 85 | * UID and GID. as such, this is needed on all architectures |
| 86 | */ |
| 87 | |
| 88 | int fs_overflowuid = DEFAULT_FS_OVERFLOWUID; |
| 89 | int fs_overflowgid = DEFAULT_FS_OVERFLOWUID; |
| 90 | |
| 91 | EXPORT_SYMBOL(fs_overflowuid); |
| 92 | EXPORT_SYMBOL(fs_overflowgid); |
| 93 | |
| 94 | /* |
| 95 | * this indicates whether you can reboot with ctrl-alt-del: the default is yes |
| 96 | */ |
| 97 | |
| 98 | int C_A_D = 1; |
Cedric Le Goater | 9ec5209 | 2006-10-02 02:19:00 -0700 | [diff] [blame] | 99 | struct pid *cad_pid; |
| 100 | EXPORT_SYMBOL(cad_pid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 101 | |
| 102 | /* |
| 103 | * Notifier list for kernel code which wants to be called |
| 104 | * at shutdown. This is used to stop any idling DMA operations |
| 105 | * and the like. |
| 106 | */ |
| 107 | |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 108 | static BLOCKING_NOTIFIER_HEAD(reboot_notifier_list); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 109 | |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 110 | /* |
| 111 | * Notifier chain core routines. The exported routines below |
| 112 | * are layered on top of these, with appropriate locking added. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 113 | */ |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 114 | |
| 115 | static int notifier_chain_register(struct notifier_block **nl, |
| 116 | struct notifier_block *n) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 117 | { |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 118 | while ((*nl) != NULL) { |
| 119 | if (n->priority > (*nl)->priority) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 120 | break; |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 121 | nl = &((*nl)->next); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 122 | } |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 123 | n->next = *nl; |
| 124 | rcu_assign_pointer(*nl, n); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 125 | return 0; |
| 126 | } |
| 127 | |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 128 | static int notifier_chain_unregister(struct notifier_block **nl, |
| 129 | struct notifier_block *n) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 130 | { |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 131 | while ((*nl) != NULL) { |
| 132 | if ((*nl) == n) { |
| 133 | rcu_assign_pointer(*nl, n->next); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 134 | return 0; |
| 135 | } |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 136 | nl = &((*nl)->next); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 137 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 138 | return -ENOENT; |
| 139 | } |
| 140 | |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 141 | /** |
| 142 | * notifier_call_chain - Informs the registered notifiers about an event. |
| 143 | * @nl: Pointer to head of the blocking notifier chain |
| 144 | * @val: Value passed unmodified to notifier function |
| 145 | * @v: Pointer passed unmodified to notifier function |
| 146 | * @nr_to_call: Number of notifier functions to be called. Don't care |
| 147 | * value of this parameter is -1. |
| 148 | * @nr_calls: Records the number of notifications sent. Don't care |
| 149 | * value of this field is NULL. |
| 150 | * @returns: notifier_call_chain returns the value returned by the |
| 151 | * last notifier function called. |
| 152 | */ |
| 153 | |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 154 | static int __kprobes notifier_call_chain(struct notifier_block **nl, |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 155 | unsigned long val, void *v, |
| 156 | int nr_to_call, int *nr_calls) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 157 | { |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 158 | int ret = NOTIFY_DONE; |
Alan Stern | bbb1747 | 2006-06-25 05:47:15 -0700 | [diff] [blame] | 159 | struct notifier_block *nb, *next_nb; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 160 | |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 161 | nb = rcu_dereference(*nl); |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 162 | |
| 163 | while (nb && nr_to_call) { |
Alan Stern | bbb1747 | 2006-06-25 05:47:15 -0700 | [diff] [blame] | 164 | next_nb = rcu_dereference(nb->next); |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 165 | ret = nb->notifier_call(nb, val, v); |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 166 | |
| 167 | if (nr_calls) |
| 168 | (*nr_calls)++; |
| 169 | |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 170 | if ((ret & NOTIFY_STOP_MASK) == NOTIFY_STOP_MASK) |
| 171 | break; |
Alan Stern | bbb1747 | 2006-06-25 05:47:15 -0700 | [diff] [blame] | 172 | nb = next_nb; |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 173 | nr_to_call--; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 174 | } |
| 175 | return ret; |
| 176 | } |
| 177 | |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 178 | /* |
| 179 | * Atomic notifier chain routines. Registration and unregistration |
Alan Stern | eabc069 | 2006-10-04 02:17:04 -0700 | [diff] [blame] | 180 | * use a spinlock, and call_chain is synchronized by RCU (no locks). |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 181 | */ |
| 182 | |
| 183 | /** |
| 184 | * atomic_notifier_chain_register - Add notifier to an atomic notifier chain |
| 185 | * @nh: Pointer to head of the atomic notifier chain |
| 186 | * @n: New entry in notifier chain |
| 187 | * |
| 188 | * Adds a notifier to an atomic notifier chain. |
| 189 | * |
| 190 | * Currently always returns zero. |
| 191 | */ |
| 192 | |
| 193 | int atomic_notifier_chain_register(struct atomic_notifier_head *nh, |
| 194 | struct notifier_block *n) |
| 195 | { |
| 196 | unsigned long flags; |
| 197 | int ret; |
| 198 | |
| 199 | spin_lock_irqsave(&nh->lock, flags); |
| 200 | ret = notifier_chain_register(&nh->head, n); |
| 201 | spin_unlock_irqrestore(&nh->lock, flags); |
| 202 | return ret; |
| 203 | } |
| 204 | |
| 205 | EXPORT_SYMBOL_GPL(atomic_notifier_chain_register); |
| 206 | |
| 207 | /** |
| 208 | * atomic_notifier_chain_unregister - Remove notifier from an atomic notifier chain |
| 209 | * @nh: Pointer to head of the atomic notifier chain |
| 210 | * @n: Entry to remove from notifier chain |
| 211 | * |
| 212 | * Removes a notifier from an atomic notifier chain. |
| 213 | * |
| 214 | * Returns zero on success or %-ENOENT on failure. |
| 215 | */ |
| 216 | int atomic_notifier_chain_unregister(struct atomic_notifier_head *nh, |
| 217 | struct notifier_block *n) |
| 218 | { |
| 219 | unsigned long flags; |
| 220 | int ret; |
| 221 | |
| 222 | spin_lock_irqsave(&nh->lock, flags); |
| 223 | ret = notifier_chain_unregister(&nh->head, n); |
| 224 | spin_unlock_irqrestore(&nh->lock, flags); |
| 225 | synchronize_rcu(); |
| 226 | return ret; |
| 227 | } |
| 228 | |
| 229 | EXPORT_SYMBOL_GPL(atomic_notifier_chain_unregister); |
| 230 | |
| 231 | /** |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 232 | * __atomic_notifier_call_chain - Call functions in an atomic notifier chain |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 233 | * @nh: Pointer to head of the atomic notifier chain |
| 234 | * @val: Value passed unmodified to notifier function |
| 235 | * @v: Pointer passed unmodified to notifier function |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 236 | * @nr_to_call: See the comment for notifier_call_chain. |
| 237 | * @nr_calls: See the comment for notifier_call_chain. |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 238 | * |
| 239 | * Calls each function in a notifier chain in turn. The functions |
| 240 | * run in an atomic context, so they must not block. |
| 241 | * This routine uses RCU to synchronize with changes to the chain. |
| 242 | * |
| 243 | * If the return value of the notifier can be and'ed |
Robert P. J. Day | 72fd4a3 | 2007-02-10 01:45:59 -0800 | [diff] [blame] | 244 | * with %NOTIFY_STOP_MASK then atomic_notifier_call_chain() |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 245 | * will return immediately, with the return value of |
| 246 | * the notifier function which halted execution. |
| 247 | * Otherwise the return value is the return value |
| 248 | * of the last notifier function called. |
| 249 | */ |
| 250 | |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 251 | int __kprobes __atomic_notifier_call_chain(struct atomic_notifier_head *nh, |
| 252 | unsigned long val, void *v, |
| 253 | int nr_to_call, int *nr_calls) |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 254 | { |
| 255 | int ret; |
| 256 | |
| 257 | rcu_read_lock(); |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 258 | ret = notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls); |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 259 | rcu_read_unlock(); |
| 260 | return ret; |
| 261 | } |
| 262 | |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 263 | EXPORT_SYMBOL_GPL(__atomic_notifier_call_chain); |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 264 | |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 265 | int __kprobes atomic_notifier_call_chain(struct atomic_notifier_head *nh, |
| 266 | unsigned long val, void *v) |
| 267 | { |
| 268 | return __atomic_notifier_call_chain(nh, val, v, -1, NULL); |
| 269 | } |
| 270 | |
| 271 | EXPORT_SYMBOL_GPL(atomic_notifier_call_chain); |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 272 | /* |
| 273 | * Blocking notifier chain routines. All access to the chain is |
| 274 | * synchronized by an rwsem. |
| 275 | */ |
| 276 | |
| 277 | /** |
| 278 | * blocking_notifier_chain_register - Add notifier to a blocking notifier chain |
| 279 | * @nh: Pointer to head of the blocking notifier chain |
| 280 | * @n: New entry in notifier chain |
| 281 | * |
| 282 | * Adds a notifier to a blocking notifier chain. |
| 283 | * Must be called in process context. |
| 284 | * |
| 285 | * Currently always returns zero. |
| 286 | */ |
| 287 | |
| 288 | int blocking_notifier_chain_register(struct blocking_notifier_head *nh, |
| 289 | struct notifier_block *n) |
| 290 | { |
| 291 | int ret; |
| 292 | |
| 293 | /* |
| 294 | * This code gets used during boot-up, when task switching is |
| 295 | * not yet working and interrupts must remain disabled. At |
| 296 | * such times we must not call down_write(). |
| 297 | */ |
| 298 | if (unlikely(system_state == SYSTEM_BOOTING)) |
| 299 | return notifier_chain_register(&nh->head, n); |
| 300 | |
| 301 | down_write(&nh->rwsem); |
| 302 | ret = notifier_chain_register(&nh->head, n); |
| 303 | up_write(&nh->rwsem); |
| 304 | return ret; |
| 305 | } |
| 306 | |
| 307 | EXPORT_SYMBOL_GPL(blocking_notifier_chain_register); |
| 308 | |
| 309 | /** |
| 310 | * blocking_notifier_chain_unregister - Remove notifier from a blocking notifier chain |
| 311 | * @nh: Pointer to head of the blocking notifier chain |
| 312 | * @n: Entry to remove from notifier chain |
| 313 | * |
| 314 | * Removes a notifier from a blocking notifier chain. |
| 315 | * Must be called from process context. |
| 316 | * |
| 317 | * Returns zero on success or %-ENOENT on failure. |
| 318 | */ |
| 319 | int blocking_notifier_chain_unregister(struct blocking_notifier_head *nh, |
| 320 | struct notifier_block *n) |
| 321 | { |
| 322 | int ret; |
| 323 | |
| 324 | /* |
| 325 | * This code gets used during boot-up, when task switching is |
| 326 | * not yet working and interrupts must remain disabled. At |
| 327 | * such times we must not call down_write(). |
| 328 | */ |
| 329 | if (unlikely(system_state == SYSTEM_BOOTING)) |
| 330 | return notifier_chain_unregister(&nh->head, n); |
| 331 | |
| 332 | down_write(&nh->rwsem); |
| 333 | ret = notifier_chain_unregister(&nh->head, n); |
| 334 | up_write(&nh->rwsem); |
| 335 | return ret; |
| 336 | } |
| 337 | |
| 338 | EXPORT_SYMBOL_GPL(blocking_notifier_chain_unregister); |
| 339 | |
| 340 | /** |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 341 | * __blocking_notifier_call_chain - Call functions in a blocking notifier chain |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 342 | * @nh: Pointer to head of the blocking notifier chain |
| 343 | * @val: Value passed unmodified to notifier function |
| 344 | * @v: Pointer passed unmodified to notifier function |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 345 | * @nr_to_call: See comment for notifier_call_chain. |
| 346 | * @nr_calls: See comment for notifier_call_chain. |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 347 | * |
| 348 | * Calls each function in a notifier chain in turn. The functions |
| 349 | * run in a process context, so they are allowed to block. |
| 350 | * |
| 351 | * If the return value of the notifier can be and'ed |
Robert P. J. Day | 72fd4a3 | 2007-02-10 01:45:59 -0800 | [diff] [blame] | 352 | * with %NOTIFY_STOP_MASK then blocking_notifier_call_chain() |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 353 | * will return immediately, with the return value of |
| 354 | * the notifier function which halted execution. |
| 355 | * Otherwise the return value is the return value |
| 356 | * of the last notifier function called. |
| 357 | */ |
| 358 | |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 359 | int __blocking_notifier_call_chain(struct blocking_notifier_head *nh, |
| 360 | unsigned long val, void *v, |
| 361 | int nr_to_call, int *nr_calls) |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 362 | { |
Ingo Molnar | 1b5180b | 2007-01-23 10:45:50 +0100 | [diff] [blame] | 363 | int ret = NOTIFY_DONE; |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 364 | |
Ingo Molnar | 1b5180b | 2007-01-23 10:45:50 +0100 | [diff] [blame] | 365 | /* |
| 366 | * We check the head outside the lock, but if this access is |
| 367 | * racy then it does not matter what the result of the test |
| 368 | * is, we re-check the list after having taken the lock anyway: |
| 369 | */ |
| 370 | if (rcu_dereference(nh->head)) { |
| 371 | down_read(&nh->rwsem); |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 372 | ret = notifier_call_chain(&nh->head, val, v, nr_to_call, |
| 373 | nr_calls); |
Ingo Molnar | 1b5180b | 2007-01-23 10:45:50 +0100 | [diff] [blame] | 374 | up_read(&nh->rwsem); |
| 375 | } |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 376 | return ret; |
| 377 | } |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 378 | EXPORT_SYMBOL_GPL(__blocking_notifier_call_chain); |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 379 | |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 380 | int blocking_notifier_call_chain(struct blocking_notifier_head *nh, |
| 381 | unsigned long val, void *v) |
| 382 | { |
| 383 | return __blocking_notifier_call_chain(nh, val, v, -1, NULL); |
| 384 | } |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 385 | EXPORT_SYMBOL_GPL(blocking_notifier_call_chain); |
| 386 | |
| 387 | /* |
| 388 | * Raw notifier chain routines. There is no protection; |
| 389 | * the caller must provide it. Use at your own risk! |
| 390 | */ |
| 391 | |
| 392 | /** |
| 393 | * raw_notifier_chain_register - Add notifier to a raw notifier chain |
| 394 | * @nh: Pointer to head of the raw notifier chain |
| 395 | * @n: New entry in notifier chain |
| 396 | * |
| 397 | * Adds a notifier to a raw notifier chain. |
| 398 | * All locking must be provided by the caller. |
| 399 | * |
| 400 | * Currently always returns zero. |
| 401 | */ |
| 402 | |
| 403 | int raw_notifier_chain_register(struct raw_notifier_head *nh, |
| 404 | struct notifier_block *n) |
| 405 | { |
| 406 | return notifier_chain_register(&nh->head, n); |
| 407 | } |
| 408 | |
| 409 | EXPORT_SYMBOL_GPL(raw_notifier_chain_register); |
| 410 | |
| 411 | /** |
| 412 | * raw_notifier_chain_unregister - Remove notifier from a raw notifier chain |
| 413 | * @nh: Pointer to head of the raw notifier chain |
| 414 | * @n: Entry to remove from notifier chain |
| 415 | * |
| 416 | * Removes a notifier from a raw notifier chain. |
| 417 | * All locking must be provided by the caller. |
| 418 | * |
| 419 | * Returns zero on success or %-ENOENT on failure. |
| 420 | */ |
| 421 | int raw_notifier_chain_unregister(struct raw_notifier_head *nh, |
| 422 | struct notifier_block *n) |
| 423 | { |
| 424 | return notifier_chain_unregister(&nh->head, n); |
| 425 | } |
| 426 | |
| 427 | EXPORT_SYMBOL_GPL(raw_notifier_chain_unregister); |
| 428 | |
| 429 | /** |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 430 | * __raw_notifier_call_chain - Call functions in a raw notifier chain |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 431 | * @nh: Pointer to head of the raw notifier chain |
| 432 | * @val: Value passed unmodified to notifier function |
| 433 | * @v: Pointer passed unmodified to notifier function |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 434 | * @nr_to_call: See comment for notifier_call_chain. |
| 435 | * @nr_calls: See comment for notifier_call_chain |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 436 | * |
| 437 | * Calls each function in a notifier chain in turn. The functions |
| 438 | * run in an undefined context. |
| 439 | * All locking must be provided by the caller. |
| 440 | * |
| 441 | * If the return value of the notifier can be and'ed |
Robert P. J. Day | 72fd4a3 | 2007-02-10 01:45:59 -0800 | [diff] [blame] | 442 | * with %NOTIFY_STOP_MASK then raw_notifier_call_chain() |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 443 | * will return immediately, with the return value of |
| 444 | * the notifier function which halted execution. |
| 445 | * Otherwise the return value is the return value |
| 446 | * of the last notifier function called. |
| 447 | */ |
| 448 | |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 449 | int __raw_notifier_call_chain(struct raw_notifier_head *nh, |
| 450 | unsigned long val, void *v, |
| 451 | int nr_to_call, int *nr_calls) |
| 452 | { |
| 453 | return notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls); |
| 454 | } |
| 455 | |
| 456 | EXPORT_SYMBOL_GPL(__raw_notifier_call_chain); |
| 457 | |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 458 | int raw_notifier_call_chain(struct raw_notifier_head *nh, |
| 459 | unsigned long val, void *v) |
| 460 | { |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 461 | return __raw_notifier_call_chain(nh, val, v, -1, NULL); |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 462 | } |
| 463 | |
| 464 | EXPORT_SYMBOL_GPL(raw_notifier_call_chain); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 465 | |
Alan Stern | eabc069 | 2006-10-04 02:17:04 -0700 | [diff] [blame] | 466 | /* |
| 467 | * SRCU notifier chain routines. Registration and unregistration |
| 468 | * use a mutex, and call_chain is synchronized by SRCU (no locks). |
| 469 | */ |
| 470 | |
| 471 | /** |
| 472 | * srcu_notifier_chain_register - Add notifier to an SRCU notifier chain |
| 473 | * @nh: Pointer to head of the SRCU notifier chain |
| 474 | * @n: New entry in notifier chain |
| 475 | * |
| 476 | * Adds a notifier to an SRCU notifier chain. |
| 477 | * Must be called in process context. |
| 478 | * |
| 479 | * Currently always returns zero. |
| 480 | */ |
| 481 | |
| 482 | int srcu_notifier_chain_register(struct srcu_notifier_head *nh, |
| 483 | struct notifier_block *n) |
| 484 | { |
| 485 | int ret; |
| 486 | |
| 487 | /* |
| 488 | * This code gets used during boot-up, when task switching is |
| 489 | * not yet working and interrupts must remain disabled. At |
| 490 | * such times we must not call mutex_lock(). |
| 491 | */ |
| 492 | if (unlikely(system_state == SYSTEM_BOOTING)) |
| 493 | return notifier_chain_register(&nh->head, n); |
| 494 | |
| 495 | mutex_lock(&nh->mutex); |
| 496 | ret = notifier_chain_register(&nh->head, n); |
| 497 | mutex_unlock(&nh->mutex); |
| 498 | return ret; |
| 499 | } |
| 500 | |
| 501 | EXPORT_SYMBOL_GPL(srcu_notifier_chain_register); |
| 502 | |
| 503 | /** |
| 504 | * srcu_notifier_chain_unregister - Remove notifier from an SRCU notifier chain |
| 505 | * @nh: Pointer to head of the SRCU notifier chain |
| 506 | * @n: Entry to remove from notifier chain |
| 507 | * |
| 508 | * Removes a notifier from an SRCU notifier chain. |
| 509 | * Must be called from process context. |
| 510 | * |
| 511 | * Returns zero on success or %-ENOENT on failure. |
| 512 | */ |
| 513 | int srcu_notifier_chain_unregister(struct srcu_notifier_head *nh, |
| 514 | struct notifier_block *n) |
| 515 | { |
| 516 | int ret; |
| 517 | |
| 518 | /* |
| 519 | * This code gets used during boot-up, when task switching is |
| 520 | * not yet working and interrupts must remain disabled. At |
| 521 | * such times we must not call mutex_lock(). |
| 522 | */ |
| 523 | if (unlikely(system_state == SYSTEM_BOOTING)) |
| 524 | return notifier_chain_unregister(&nh->head, n); |
| 525 | |
| 526 | mutex_lock(&nh->mutex); |
| 527 | ret = notifier_chain_unregister(&nh->head, n); |
| 528 | mutex_unlock(&nh->mutex); |
| 529 | synchronize_srcu(&nh->srcu); |
| 530 | return ret; |
| 531 | } |
| 532 | |
| 533 | EXPORT_SYMBOL_GPL(srcu_notifier_chain_unregister); |
| 534 | |
| 535 | /** |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 536 | * __srcu_notifier_call_chain - Call functions in an SRCU notifier chain |
Alan Stern | eabc069 | 2006-10-04 02:17:04 -0700 | [diff] [blame] | 537 | * @nh: Pointer to head of the SRCU notifier chain |
| 538 | * @val: Value passed unmodified to notifier function |
| 539 | * @v: Pointer passed unmodified to notifier function |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 540 | * @nr_to_call: See comment for notifier_call_chain. |
| 541 | * @nr_calls: See comment for notifier_call_chain |
Alan Stern | eabc069 | 2006-10-04 02:17:04 -0700 | [diff] [blame] | 542 | * |
| 543 | * Calls each function in a notifier chain in turn. The functions |
| 544 | * run in a process context, so they are allowed to block. |
| 545 | * |
| 546 | * If the return value of the notifier can be and'ed |
Robert P. J. Day | 72fd4a3 | 2007-02-10 01:45:59 -0800 | [diff] [blame] | 547 | * with %NOTIFY_STOP_MASK then srcu_notifier_call_chain() |
Alan Stern | eabc069 | 2006-10-04 02:17:04 -0700 | [diff] [blame] | 548 | * will return immediately, with the return value of |
| 549 | * the notifier function which halted execution. |
| 550 | * Otherwise the return value is the return value |
| 551 | * of the last notifier function called. |
| 552 | */ |
| 553 | |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 554 | int __srcu_notifier_call_chain(struct srcu_notifier_head *nh, |
| 555 | unsigned long val, void *v, |
| 556 | int nr_to_call, int *nr_calls) |
Alan Stern | eabc069 | 2006-10-04 02:17:04 -0700 | [diff] [blame] | 557 | { |
| 558 | int ret; |
| 559 | int idx; |
| 560 | |
| 561 | idx = srcu_read_lock(&nh->srcu); |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 562 | ret = notifier_call_chain(&nh->head, val, v, nr_to_call, nr_calls); |
Alan Stern | eabc069 | 2006-10-04 02:17:04 -0700 | [diff] [blame] | 563 | srcu_read_unlock(&nh->srcu, idx); |
| 564 | return ret; |
| 565 | } |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 566 | EXPORT_SYMBOL_GPL(__srcu_notifier_call_chain); |
Alan Stern | eabc069 | 2006-10-04 02:17:04 -0700 | [diff] [blame] | 567 | |
Gautham R Shenoy | 6f7cc11 | 2007-05-09 02:34:02 -0700 | [diff] [blame] | 568 | int srcu_notifier_call_chain(struct srcu_notifier_head *nh, |
| 569 | unsigned long val, void *v) |
| 570 | { |
| 571 | return __srcu_notifier_call_chain(nh, val, v, -1, NULL); |
| 572 | } |
Alan Stern | eabc069 | 2006-10-04 02:17:04 -0700 | [diff] [blame] | 573 | EXPORT_SYMBOL_GPL(srcu_notifier_call_chain); |
| 574 | |
| 575 | /** |
| 576 | * srcu_init_notifier_head - Initialize an SRCU notifier head |
| 577 | * @nh: Pointer to head of the srcu notifier chain |
| 578 | * |
| 579 | * Unlike other sorts of notifier heads, SRCU notifier heads require |
| 580 | * dynamic initialization. Be sure to call this routine before |
| 581 | * calling any of the other SRCU notifier routines for this head. |
| 582 | * |
| 583 | * If an SRCU notifier head is deallocated, it must first be cleaned |
| 584 | * up by calling srcu_cleanup_notifier_head(). Otherwise the head's |
| 585 | * per-cpu data (used by the SRCU mechanism) will leak. |
| 586 | */ |
| 587 | |
| 588 | void srcu_init_notifier_head(struct srcu_notifier_head *nh) |
| 589 | { |
| 590 | mutex_init(&nh->mutex); |
Alan Stern | e6a9201 | 2006-10-04 02:17:05 -0700 | [diff] [blame] | 591 | if (init_srcu_struct(&nh->srcu) < 0) |
| 592 | BUG(); |
Alan Stern | eabc069 | 2006-10-04 02:17:04 -0700 | [diff] [blame] | 593 | nh->head = NULL; |
| 594 | } |
| 595 | |
| 596 | EXPORT_SYMBOL_GPL(srcu_init_notifier_head); |
| 597 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 598 | /** |
| 599 | * register_reboot_notifier - Register function to be called at reboot time |
| 600 | * @nb: Info about notifier function to be called |
| 601 | * |
| 602 | * Registers a function with the list of functions |
| 603 | * to be called at reboot time. |
| 604 | * |
Robert P. J. Day | 72fd4a3 | 2007-02-10 01:45:59 -0800 | [diff] [blame] | 605 | * Currently always returns zero, as blocking_notifier_chain_register() |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 606 | * always returns zero. |
| 607 | */ |
| 608 | |
| 609 | int register_reboot_notifier(struct notifier_block * nb) |
| 610 | { |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 611 | return blocking_notifier_chain_register(&reboot_notifier_list, nb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 612 | } |
| 613 | |
| 614 | EXPORT_SYMBOL(register_reboot_notifier); |
| 615 | |
| 616 | /** |
| 617 | * unregister_reboot_notifier - Unregister previously registered reboot notifier |
| 618 | * @nb: Hook to be unregistered |
| 619 | * |
| 620 | * Unregisters a previously registered reboot |
| 621 | * notifier function. |
| 622 | * |
| 623 | * Returns zero on success, or %-ENOENT on failure. |
| 624 | */ |
| 625 | |
| 626 | int unregister_reboot_notifier(struct notifier_block * nb) |
| 627 | { |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 628 | return blocking_notifier_chain_unregister(&reboot_notifier_list, nb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 629 | } |
| 630 | |
| 631 | EXPORT_SYMBOL(unregister_reboot_notifier); |
| 632 | |
| 633 | static int set_one_prio(struct task_struct *p, int niceval, int error) |
| 634 | { |
| 635 | int no_nice; |
| 636 | |
| 637 | if (p->uid != current->euid && |
| 638 | p->euid != current->euid && !capable(CAP_SYS_NICE)) { |
| 639 | error = -EPERM; |
| 640 | goto out; |
| 641 | } |
Matt Mackall | e43379f | 2005-05-01 08:59:00 -0700 | [diff] [blame] | 642 | if (niceval < task_nice(p) && !can_nice(p, niceval)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 643 | error = -EACCES; |
| 644 | goto out; |
| 645 | } |
| 646 | no_nice = security_task_setnice(p, niceval); |
| 647 | if (no_nice) { |
| 648 | error = no_nice; |
| 649 | goto out; |
| 650 | } |
| 651 | if (error == -ESRCH) |
| 652 | error = 0; |
| 653 | set_user_nice(p, niceval); |
| 654 | out: |
| 655 | return error; |
| 656 | } |
| 657 | |
| 658 | asmlinkage long sys_setpriority(int which, int who, int niceval) |
| 659 | { |
| 660 | struct task_struct *g, *p; |
| 661 | struct user_struct *user; |
| 662 | int error = -EINVAL; |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 663 | struct pid *pgrp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 664 | |
Daniel Walker | 3e88c55 | 2007-05-10 22:22:53 -0700 | [diff] [blame] | 665 | if (which > PRIO_USER || which < PRIO_PROCESS) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 666 | goto out; |
| 667 | |
| 668 | /* normalize: avoid signed division (rounding problems) */ |
| 669 | error = -ESRCH; |
| 670 | if (niceval < -20) |
| 671 | niceval = -20; |
| 672 | if (niceval > 19) |
| 673 | niceval = 19; |
| 674 | |
| 675 | read_lock(&tasklist_lock); |
| 676 | switch (which) { |
| 677 | case PRIO_PROCESS: |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 678 | if (who) |
| 679 | p = find_task_by_pid(who); |
| 680 | else |
| 681 | p = current; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 682 | if (p) |
| 683 | error = set_one_prio(p, niceval, error); |
| 684 | break; |
| 685 | case PRIO_PGRP: |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 686 | if (who) |
| 687 | pgrp = find_pid(who); |
| 688 | else |
| 689 | pgrp = task_pgrp(current); |
| 690 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 691 | error = set_one_prio(p, niceval, error); |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 692 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 693 | break; |
| 694 | case PRIO_USER: |
| 695 | user = current->user; |
| 696 | if (!who) |
| 697 | who = current->uid; |
| 698 | else |
| 699 | if ((who != current->uid) && !(user = find_user(who))) |
| 700 | goto out_unlock; /* No processes for this user */ |
| 701 | |
| 702 | do_each_thread(g, p) |
| 703 | if (p->uid == who) |
| 704 | error = set_one_prio(p, niceval, error); |
| 705 | while_each_thread(g, p); |
| 706 | if (who != current->uid) |
| 707 | free_uid(user); /* For find_user() */ |
| 708 | break; |
| 709 | } |
| 710 | out_unlock: |
| 711 | read_unlock(&tasklist_lock); |
| 712 | out: |
| 713 | return error; |
| 714 | } |
| 715 | |
| 716 | /* |
| 717 | * Ugh. To avoid negative return values, "getpriority()" will |
| 718 | * not return the normal nice-value, but a negated value that |
| 719 | * has been offset by 20 (ie it returns 40..1 instead of -20..19) |
| 720 | * to stay compatible. |
| 721 | */ |
| 722 | asmlinkage long sys_getpriority(int which, int who) |
| 723 | { |
| 724 | struct task_struct *g, *p; |
| 725 | struct user_struct *user; |
| 726 | long niceval, retval = -ESRCH; |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 727 | struct pid *pgrp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 728 | |
Daniel Walker | 3e88c55 | 2007-05-10 22:22:53 -0700 | [diff] [blame] | 729 | if (which > PRIO_USER || which < PRIO_PROCESS) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 730 | return -EINVAL; |
| 731 | |
| 732 | read_lock(&tasklist_lock); |
| 733 | switch (which) { |
| 734 | case PRIO_PROCESS: |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 735 | if (who) |
| 736 | p = find_task_by_pid(who); |
| 737 | else |
| 738 | p = current; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 739 | if (p) { |
| 740 | niceval = 20 - task_nice(p); |
| 741 | if (niceval > retval) |
| 742 | retval = niceval; |
| 743 | } |
| 744 | break; |
| 745 | case PRIO_PGRP: |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 746 | if (who) |
| 747 | pgrp = find_pid(who); |
| 748 | else |
| 749 | pgrp = task_pgrp(current); |
| 750 | do_each_pid_task(pgrp, PIDTYPE_PGID, p) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 751 | niceval = 20 - task_nice(p); |
| 752 | if (niceval > retval) |
| 753 | retval = niceval; |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 754 | } while_each_pid_task(pgrp, PIDTYPE_PGID, p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 755 | break; |
| 756 | case PRIO_USER: |
| 757 | user = current->user; |
| 758 | if (!who) |
| 759 | who = current->uid; |
| 760 | else |
| 761 | if ((who != current->uid) && !(user = find_user(who))) |
| 762 | goto out_unlock; /* No processes for this user */ |
| 763 | |
| 764 | do_each_thread(g, p) |
| 765 | if (p->uid == who) { |
| 766 | niceval = 20 - task_nice(p); |
| 767 | if (niceval > retval) |
| 768 | retval = niceval; |
| 769 | } |
| 770 | while_each_thread(g, p); |
| 771 | if (who != current->uid) |
| 772 | free_uid(user); /* for find_user() */ |
| 773 | break; |
| 774 | } |
| 775 | out_unlock: |
| 776 | read_unlock(&tasklist_lock); |
| 777 | |
| 778 | return retval; |
| 779 | } |
| 780 | |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 781 | /** |
| 782 | * emergency_restart - reboot the system |
| 783 | * |
| 784 | * Without shutting down any hardware or taking any locks |
| 785 | * reboot the system. This is called when we know we are in |
| 786 | * trouble so this is our best effort to reboot. This is |
| 787 | * safe to call in interrupt context. |
| 788 | */ |
Eric W. Biederman | 7c90347 | 2005-07-26 11:29:55 -0600 | [diff] [blame] | 789 | void emergency_restart(void) |
| 790 | { |
| 791 | machine_emergency_restart(); |
| 792 | } |
| 793 | EXPORT_SYMBOL_GPL(emergency_restart); |
| 794 | |
Adrian Bunk | 83cc5ed | 2006-06-25 05:47:41 -0700 | [diff] [blame] | 795 | static void kernel_restart_prepare(char *cmd) |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 796 | { |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 797 | blocking_notifier_call_chain(&reboot_notifier_list, SYS_RESTART, cmd); |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 798 | system_state = SYSTEM_RESTART; |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 799 | device_shutdown(); |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 800 | } |
Randy Dunlap | 1e5d533 | 2005-11-07 01:01:06 -0800 | [diff] [blame] | 801 | |
| 802 | /** |
| 803 | * kernel_restart - reboot the system |
| 804 | * @cmd: pointer to buffer containing command to execute for restart |
Randy Dunlap | b8887e6 | 2005-11-07 01:01:07 -0800 | [diff] [blame] | 805 | * or %NULL |
Randy Dunlap | 1e5d533 | 2005-11-07 01:01:06 -0800 | [diff] [blame] | 806 | * |
| 807 | * Shutdown everything and perform a clean reboot. |
| 808 | * This is not safe to call in interrupt context. |
| 809 | */ |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 810 | void kernel_restart(char *cmd) |
| 811 | { |
| 812 | kernel_restart_prepare(cmd); |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 813 | if (!cmd) |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 814 | printk(KERN_EMERG "Restarting system.\n"); |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 815 | else |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 816 | printk(KERN_EMERG "Restarting system with command '%s'.\n", cmd); |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 817 | machine_restart(cmd); |
| 818 | } |
| 819 | EXPORT_SYMBOL_GPL(kernel_restart); |
| 820 | |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 821 | /** |
| 822 | * kernel_kexec - reboot the system |
| 823 | * |
| 824 | * Move into place and start executing a preloaded standalone |
| 825 | * executable. If nothing was preloaded return an error. |
| 826 | */ |
Adrian Bunk | 83cc5ed | 2006-06-25 05:47:41 -0700 | [diff] [blame] | 827 | static void kernel_kexec(void) |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 828 | { |
| 829 | #ifdef CONFIG_KEXEC |
| 830 | struct kimage *image; |
Al Viro | 4bb8089 | 2006-02-01 05:57:32 -0500 | [diff] [blame] | 831 | image = xchg(&kexec_image, NULL); |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 832 | if (!image) |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 833 | return; |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 834 | kernel_restart_prepare(NULL); |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 835 | printk(KERN_EMERG "Starting new kernel\n"); |
| 836 | machine_shutdown(); |
| 837 | machine_kexec(image); |
| 838 | #endif |
| 839 | } |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 840 | |
Alexey Starikovskiy | 729b4d4 | 2005-12-01 04:29:00 -0500 | [diff] [blame] | 841 | void kernel_shutdown_prepare(enum system_states state) |
| 842 | { |
Alan Stern | e041c68 | 2006-03-27 01:16:30 -0800 | [diff] [blame] | 843 | blocking_notifier_call_chain(&reboot_notifier_list, |
Alexey Starikovskiy | 729b4d4 | 2005-12-01 04:29:00 -0500 | [diff] [blame] | 844 | (state == SYSTEM_HALT)?SYS_HALT:SYS_POWER_OFF, NULL); |
| 845 | system_state = state; |
| 846 | device_shutdown(); |
| 847 | } |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 848 | /** |
| 849 | * kernel_halt - halt the system |
| 850 | * |
| 851 | * Shutdown everything and perform a clean system halt. |
| 852 | */ |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 853 | void kernel_halt(void) |
| 854 | { |
Alexey Starikovskiy | 729b4d4 | 2005-12-01 04:29:00 -0500 | [diff] [blame] | 855 | kernel_shutdown_prepare(SYSTEM_HALT); |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 856 | printk(KERN_EMERG "System halted.\n"); |
| 857 | machine_halt(); |
| 858 | } |
Alexey Starikovskiy | 729b4d4 | 2005-12-01 04:29:00 -0500 | [diff] [blame] | 859 | |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 860 | EXPORT_SYMBOL_GPL(kernel_halt); |
| 861 | |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 862 | /** |
| 863 | * kernel_power_off - power_off the system |
| 864 | * |
| 865 | * Shutdown everything and perform a clean system power_off. |
| 866 | */ |
Eric W. Biederman | e4c9433 | 2005-09-22 21:43:45 -0700 | [diff] [blame] | 867 | void kernel_power_off(void) |
| 868 | { |
Alexey Starikovskiy | 729b4d4 | 2005-12-01 04:29:00 -0500 | [diff] [blame] | 869 | kernel_shutdown_prepare(SYSTEM_POWER_OFF); |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 870 | printk(KERN_EMERG "Power down.\n"); |
| 871 | machine_power_off(); |
| 872 | } |
| 873 | EXPORT_SYMBOL_GPL(kernel_power_off); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 874 | /* |
| 875 | * Reboot system call: for obvious reasons only root may call it, |
| 876 | * and even root needs to set up some magic numbers in the registers |
| 877 | * so that some mistake won't make this reboot the whole machine. |
| 878 | * You can also set the meaning of the ctrl-alt-del-key here. |
| 879 | * |
| 880 | * reboot doesn't sync: do that yourself before calling this. |
| 881 | */ |
| 882 | asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void __user * arg) |
| 883 | { |
| 884 | char buffer[256]; |
| 885 | |
| 886 | /* We only trust the superuser with rebooting the system. */ |
| 887 | if (!capable(CAP_SYS_BOOT)) |
| 888 | return -EPERM; |
| 889 | |
| 890 | /* For safety, we require "magic" arguments. */ |
| 891 | if (magic1 != LINUX_REBOOT_MAGIC1 || |
| 892 | (magic2 != LINUX_REBOOT_MAGIC2 && |
| 893 | magic2 != LINUX_REBOOT_MAGIC2A && |
| 894 | magic2 != LINUX_REBOOT_MAGIC2B && |
| 895 | magic2 != LINUX_REBOOT_MAGIC2C)) |
| 896 | return -EINVAL; |
| 897 | |
Eric W. Biederman | 5e38291 | 2006-01-08 01:03:46 -0800 | [diff] [blame] | 898 | /* Instead of trying to make the power_off code look like |
| 899 | * halt when pm_power_off is not set do it the easy way. |
| 900 | */ |
| 901 | if ((cmd == LINUX_REBOOT_CMD_POWER_OFF) && !pm_power_off) |
| 902 | cmd = LINUX_REBOOT_CMD_HALT; |
| 903 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 904 | lock_kernel(); |
| 905 | switch (cmd) { |
| 906 | case LINUX_REBOOT_CMD_RESTART: |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 907 | kernel_restart(NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 908 | break; |
| 909 | |
| 910 | case LINUX_REBOOT_CMD_CAD_ON: |
| 911 | C_A_D = 1; |
| 912 | break; |
| 913 | |
| 914 | case LINUX_REBOOT_CMD_CAD_OFF: |
| 915 | C_A_D = 0; |
| 916 | break; |
| 917 | |
| 918 | case LINUX_REBOOT_CMD_HALT: |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 919 | kernel_halt(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 920 | unlock_kernel(); |
| 921 | do_exit(0); |
| 922 | break; |
| 923 | |
| 924 | case LINUX_REBOOT_CMD_POWER_OFF: |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 925 | kernel_power_off(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 926 | unlock_kernel(); |
| 927 | do_exit(0); |
| 928 | break; |
| 929 | |
| 930 | case LINUX_REBOOT_CMD_RESTART2: |
| 931 | if (strncpy_from_user(&buffer[0], arg, sizeof(buffer) - 1) < 0) { |
| 932 | unlock_kernel(); |
| 933 | return -EFAULT; |
| 934 | } |
| 935 | buffer[sizeof(buffer) - 1] = '\0'; |
| 936 | |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 937 | kernel_restart(buffer); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 938 | break; |
| 939 | |
Eric W. Biederman | dc009d9 | 2005-06-25 14:57:52 -0700 | [diff] [blame] | 940 | case LINUX_REBOOT_CMD_KEXEC: |
Eric W. Biederman | 4a00ea1 | 2005-07-26 11:24:14 -0600 | [diff] [blame] | 941 | kernel_kexec(); |
| 942 | unlock_kernel(); |
| 943 | return -EINVAL; |
| 944 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 945 | #ifdef CONFIG_SOFTWARE_SUSPEND |
| 946 | case LINUX_REBOOT_CMD_SW_SUSPEND: |
| 947 | { |
Rafael J. Wysocki | a3d25c2 | 2007-05-09 02:33:18 -0700 | [diff] [blame] | 948 | int ret = hibernate(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 949 | unlock_kernel(); |
| 950 | return ret; |
| 951 | } |
| 952 | #endif |
| 953 | |
| 954 | default: |
| 955 | unlock_kernel(); |
| 956 | return -EINVAL; |
| 957 | } |
| 958 | unlock_kernel(); |
| 959 | return 0; |
| 960 | } |
| 961 | |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 962 | static void deferred_cad(struct work_struct *dummy) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 963 | { |
Eric W. Biederman | abcd9e5 | 2005-07-26 11:27:34 -0600 | [diff] [blame] | 964 | kernel_restart(NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 965 | } |
| 966 | |
| 967 | /* |
| 968 | * This function gets called by ctrl-alt-del - ie the keyboard interrupt. |
| 969 | * As it's called within an interrupt, it may NOT sync: the only choice |
| 970 | * is whether to reboot at once, or just ignore the ctrl-alt-del. |
| 971 | */ |
| 972 | void ctrl_alt_del(void) |
| 973 | { |
David Howells | 65f27f3 | 2006-11-22 14:55:48 +0000 | [diff] [blame] | 974 | static DECLARE_WORK(cad_work, deferred_cad); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 975 | |
| 976 | if (C_A_D) |
| 977 | schedule_work(&cad_work); |
| 978 | else |
Cedric Le Goater | 9ec5209 | 2006-10-02 02:19:00 -0700 | [diff] [blame] | 979 | kill_cad_pid(SIGINT, 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 980 | } |
| 981 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 982 | /* |
| 983 | * Unprivileged users may change the real gid to the effective gid |
| 984 | * or vice versa. (BSD-style) |
| 985 | * |
| 986 | * If you set the real gid at all, or set the effective gid to a value not |
| 987 | * equal to the real gid, then the saved gid is set to the new effective gid. |
| 988 | * |
| 989 | * This makes it possible for a setgid program to completely drop its |
| 990 | * privileges, which is often a useful assertion to make when you are doing |
| 991 | * a security audit over a program. |
| 992 | * |
| 993 | * The general idea is that a program which uses just setregid() will be |
| 994 | * 100% compatible with BSD. A program which uses just setgid() will be |
| 995 | * 100% compatible with POSIX with saved IDs. |
| 996 | * |
| 997 | * SMP: There are not races, the GIDs are checked only by filesystem |
| 998 | * operations (as far as semantic preservation is concerned). |
| 999 | */ |
| 1000 | asmlinkage long sys_setregid(gid_t rgid, gid_t egid) |
| 1001 | { |
| 1002 | int old_rgid = current->gid; |
| 1003 | int old_egid = current->egid; |
| 1004 | int new_rgid = old_rgid; |
| 1005 | int new_egid = old_egid; |
| 1006 | int retval; |
| 1007 | |
| 1008 | retval = security_task_setgid(rgid, egid, (gid_t)-1, LSM_SETID_RE); |
| 1009 | if (retval) |
| 1010 | return retval; |
| 1011 | |
| 1012 | if (rgid != (gid_t) -1) { |
| 1013 | if ((old_rgid == rgid) || |
| 1014 | (current->egid==rgid) || |
| 1015 | capable(CAP_SETGID)) |
| 1016 | new_rgid = rgid; |
| 1017 | else |
| 1018 | return -EPERM; |
| 1019 | } |
| 1020 | if (egid != (gid_t) -1) { |
| 1021 | if ((old_rgid == egid) || |
| 1022 | (current->egid == egid) || |
| 1023 | (current->sgid == egid) || |
| 1024 | capable(CAP_SETGID)) |
| 1025 | new_egid = egid; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1026 | else |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1027 | return -EPERM; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1028 | } |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1029 | if (new_egid != old_egid) { |
Alan Cox | d6e7114 | 2005-06-23 00:09:43 -0700 | [diff] [blame] | 1030 | current->mm->dumpable = suid_dumpable; |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 1031 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1032 | } |
| 1033 | if (rgid != (gid_t) -1 || |
| 1034 | (egid != (gid_t) -1 && egid != old_rgid)) |
| 1035 | current->sgid = new_egid; |
| 1036 | current->fsgid = new_egid; |
| 1037 | current->egid = new_egid; |
| 1038 | current->gid = new_rgid; |
| 1039 | key_fsgid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 1040 | proc_id_connector(current, PROC_EVENT_GID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1041 | return 0; |
| 1042 | } |
| 1043 | |
| 1044 | /* |
| 1045 | * setgid() is implemented like SysV w/ SAVED_IDS |
| 1046 | * |
| 1047 | * SMP: Same implicit races as above. |
| 1048 | */ |
| 1049 | asmlinkage long sys_setgid(gid_t gid) |
| 1050 | { |
| 1051 | int old_egid = current->egid; |
| 1052 | int retval; |
| 1053 | |
| 1054 | retval = security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_ID); |
| 1055 | if (retval) |
| 1056 | return retval; |
| 1057 | |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1058 | if (capable(CAP_SETGID)) { |
| 1059 | if (old_egid != gid) { |
Alan Cox | d6e7114 | 2005-06-23 00:09:43 -0700 | [diff] [blame] | 1060 | current->mm->dumpable = suid_dumpable; |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 1061 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1062 | } |
| 1063 | current->gid = current->egid = current->sgid = current->fsgid = gid; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1064 | } else if ((gid == current->gid) || (gid == current->sgid)) { |
| 1065 | if (old_egid != gid) { |
Alan Cox | d6e7114 | 2005-06-23 00:09:43 -0700 | [diff] [blame] | 1066 | current->mm->dumpable = suid_dumpable; |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 1067 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1068 | } |
| 1069 | current->egid = current->fsgid = gid; |
| 1070 | } |
| 1071 | else |
| 1072 | return -EPERM; |
| 1073 | |
| 1074 | key_fsgid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 1075 | proc_id_connector(current, PROC_EVENT_GID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1076 | return 0; |
| 1077 | } |
| 1078 | |
| 1079 | static int set_user(uid_t new_ruid, int dumpclear) |
| 1080 | { |
| 1081 | struct user_struct *new_user; |
| 1082 | |
Cedric Le Goater | acce292 | 2007-07-15 23:40:59 -0700 | [diff] [blame] | 1083 | new_user = alloc_uid(current->nsproxy->user_ns, new_ruid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1084 | if (!new_user) |
| 1085 | return -EAGAIN; |
| 1086 | |
| 1087 | if (atomic_read(&new_user->processes) >= |
| 1088 | current->signal->rlim[RLIMIT_NPROC].rlim_cur && |
Cedric Le Goater | acce292 | 2007-07-15 23:40:59 -0700 | [diff] [blame] | 1089 | new_user != current->nsproxy->user_ns->root_user) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1090 | free_uid(new_user); |
| 1091 | return -EAGAIN; |
| 1092 | } |
| 1093 | |
| 1094 | switch_uid(new_user); |
| 1095 | |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1096 | if (dumpclear) { |
Alan Cox | d6e7114 | 2005-06-23 00:09:43 -0700 | [diff] [blame] | 1097 | current->mm->dumpable = suid_dumpable; |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 1098 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1099 | } |
| 1100 | current->uid = new_ruid; |
| 1101 | return 0; |
| 1102 | } |
| 1103 | |
| 1104 | /* |
| 1105 | * Unprivileged users may change the real uid to the effective uid |
| 1106 | * or vice versa. (BSD-style) |
| 1107 | * |
| 1108 | * If you set the real uid at all, or set the effective uid to a value not |
| 1109 | * equal to the real uid, then the saved uid is set to the new effective uid. |
| 1110 | * |
| 1111 | * This makes it possible for a setuid program to completely drop its |
| 1112 | * privileges, which is often a useful assertion to make when you are doing |
| 1113 | * a security audit over a program. |
| 1114 | * |
| 1115 | * The general idea is that a program which uses just setreuid() will be |
| 1116 | * 100% compatible with BSD. A program which uses just setuid() will be |
| 1117 | * 100% compatible with POSIX with saved IDs. |
| 1118 | */ |
| 1119 | asmlinkage long sys_setreuid(uid_t ruid, uid_t euid) |
| 1120 | { |
| 1121 | int old_ruid, old_euid, old_suid, new_ruid, new_euid; |
| 1122 | int retval; |
| 1123 | |
| 1124 | retval = security_task_setuid(ruid, euid, (uid_t)-1, LSM_SETID_RE); |
| 1125 | if (retval) |
| 1126 | return retval; |
| 1127 | |
| 1128 | new_ruid = old_ruid = current->uid; |
| 1129 | new_euid = old_euid = current->euid; |
| 1130 | old_suid = current->suid; |
| 1131 | |
| 1132 | if (ruid != (uid_t) -1) { |
| 1133 | new_ruid = ruid; |
| 1134 | if ((old_ruid != ruid) && |
| 1135 | (current->euid != ruid) && |
| 1136 | !capable(CAP_SETUID)) |
| 1137 | return -EPERM; |
| 1138 | } |
| 1139 | |
| 1140 | if (euid != (uid_t) -1) { |
| 1141 | new_euid = euid; |
| 1142 | if ((old_ruid != euid) && |
| 1143 | (current->euid != euid) && |
| 1144 | (current->suid != euid) && |
| 1145 | !capable(CAP_SETUID)) |
| 1146 | return -EPERM; |
| 1147 | } |
| 1148 | |
| 1149 | if (new_ruid != old_ruid && set_user(new_ruid, new_euid != old_euid) < 0) |
| 1150 | return -EAGAIN; |
| 1151 | |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1152 | if (new_euid != old_euid) { |
Alan Cox | d6e7114 | 2005-06-23 00:09:43 -0700 | [diff] [blame] | 1153 | current->mm->dumpable = suid_dumpable; |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 1154 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1155 | } |
| 1156 | current->fsuid = current->euid = new_euid; |
| 1157 | if (ruid != (uid_t) -1 || |
| 1158 | (euid != (uid_t) -1 && euid != old_ruid)) |
| 1159 | current->suid = current->euid; |
| 1160 | current->fsuid = current->euid; |
| 1161 | |
| 1162 | key_fsuid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 1163 | proc_id_connector(current, PROC_EVENT_UID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1164 | |
| 1165 | return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RE); |
| 1166 | } |
| 1167 | |
| 1168 | |
| 1169 | |
| 1170 | /* |
| 1171 | * setuid() is implemented like SysV with SAVED_IDS |
| 1172 | * |
| 1173 | * Note that SAVED_ID's is deficient in that a setuid root program |
| 1174 | * like sendmail, for example, cannot set its uid to be a normal |
| 1175 | * user and then switch back, because if you're root, setuid() sets |
| 1176 | * the saved uid too. If you don't like this, blame the bright people |
| 1177 | * in the POSIX committee and/or USG. Note that the BSD-style setreuid() |
| 1178 | * will allow a root program to temporarily drop privileges and be able to |
| 1179 | * regain them by swapping the real and effective uid. |
| 1180 | */ |
| 1181 | asmlinkage long sys_setuid(uid_t uid) |
| 1182 | { |
| 1183 | int old_euid = current->euid; |
David Rientjes | a09c17a | 2006-12-06 20:40:18 -0800 | [diff] [blame] | 1184 | int old_ruid, old_suid, new_suid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1185 | int retval; |
| 1186 | |
| 1187 | retval = security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_ID); |
| 1188 | if (retval) |
| 1189 | return retval; |
| 1190 | |
David Rientjes | a09c17a | 2006-12-06 20:40:18 -0800 | [diff] [blame] | 1191 | old_ruid = current->uid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1192 | old_suid = current->suid; |
| 1193 | new_suid = old_suid; |
| 1194 | |
| 1195 | if (capable(CAP_SETUID)) { |
| 1196 | if (uid != old_ruid && set_user(uid, old_euid != uid) < 0) |
| 1197 | return -EAGAIN; |
| 1198 | new_suid = uid; |
| 1199 | } else if ((uid != current->uid) && (uid != new_suid)) |
| 1200 | return -EPERM; |
| 1201 | |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1202 | if (old_euid != uid) { |
Alan Cox | d6e7114 | 2005-06-23 00:09:43 -0700 | [diff] [blame] | 1203 | current->mm->dumpable = suid_dumpable; |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 1204 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1205 | } |
| 1206 | current->fsuid = current->euid = uid; |
| 1207 | current->suid = new_suid; |
| 1208 | |
| 1209 | key_fsuid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 1210 | proc_id_connector(current, PROC_EVENT_UID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1211 | |
| 1212 | return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_ID); |
| 1213 | } |
| 1214 | |
| 1215 | |
| 1216 | /* |
| 1217 | * This function implements a generic ability to update ruid, euid, |
| 1218 | * and suid. This allows you to implement the 4.4 compatible seteuid(). |
| 1219 | */ |
| 1220 | asmlinkage long sys_setresuid(uid_t ruid, uid_t euid, uid_t suid) |
| 1221 | { |
| 1222 | int old_ruid = current->uid; |
| 1223 | int old_euid = current->euid; |
| 1224 | int old_suid = current->suid; |
| 1225 | int retval; |
| 1226 | |
| 1227 | retval = security_task_setuid(ruid, euid, suid, LSM_SETID_RES); |
| 1228 | if (retval) |
| 1229 | return retval; |
| 1230 | |
| 1231 | if (!capable(CAP_SETUID)) { |
| 1232 | if ((ruid != (uid_t) -1) && (ruid != current->uid) && |
| 1233 | (ruid != current->euid) && (ruid != current->suid)) |
| 1234 | return -EPERM; |
| 1235 | if ((euid != (uid_t) -1) && (euid != current->uid) && |
| 1236 | (euid != current->euid) && (euid != current->suid)) |
| 1237 | return -EPERM; |
| 1238 | if ((suid != (uid_t) -1) && (suid != current->uid) && |
| 1239 | (suid != current->euid) && (suid != current->suid)) |
| 1240 | return -EPERM; |
| 1241 | } |
| 1242 | if (ruid != (uid_t) -1) { |
| 1243 | if (ruid != current->uid && set_user(ruid, euid != current->euid) < 0) |
| 1244 | return -EAGAIN; |
| 1245 | } |
| 1246 | if (euid != (uid_t) -1) { |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1247 | if (euid != current->euid) { |
Alan Cox | d6e7114 | 2005-06-23 00:09:43 -0700 | [diff] [blame] | 1248 | current->mm->dumpable = suid_dumpable; |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 1249 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1250 | } |
| 1251 | current->euid = euid; |
| 1252 | } |
| 1253 | current->fsuid = current->euid; |
| 1254 | if (suid != (uid_t) -1) |
| 1255 | current->suid = suid; |
| 1256 | |
| 1257 | key_fsuid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 1258 | proc_id_connector(current, PROC_EVENT_UID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1259 | |
| 1260 | return security_task_post_setuid(old_ruid, old_euid, old_suid, LSM_SETID_RES); |
| 1261 | } |
| 1262 | |
| 1263 | asmlinkage long sys_getresuid(uid_t __user *ruid, uid_t __user *euid, uid_t __user *suid) |
| 1264 | { |
| 1265 | int retval; |
| 1266 | |
| 1267 | if (!(retval = put_user(current->uid, ruid)) && |
| 1268 | !(retval = put_user(current->euid, euid))) |
| 1269 | retval = put_user(current->suid, suid); |
| 1270 | |
| 1271 | return retval; |
| 1272 | } |
| 1273 | |
| 1274 | /* |
| 1275 | * Same as above, but for rgid, egid, sgid. |
| 1276 | */ |
| 1277 | asmlinkage long sys_setresgid(gid_t rgid, gid_t egid, gid_t sgid) |
| 1278 | { |
| 1279 | int retval; |
| 1280 | |
| 1281 | retval = security_task_setgid(rgid, egid, sgid, LSM_SETID_RES); |
| 1282 | if (retval) |
| 1283 | return retval; |
| 1284 | |
| 1285 | if (!capable(CAP_SETGID)) { |
| 1286 | if ((rgid != (gid_t) -1) && (rgid != current->gid) && |
| 1287 | (rgid != current->egid) && (rgid != current->sgid)) |
| 1288 | return -EPERM; |
| 1289 | if ((egid != (gid_t) -1) && (egid != current->gid) && |
| 1290 | (egid != current->egid) && (egid != current->sgid)) |
| 1291 | return -EPERM; |
| 1292 | if ((sgid != (gid_t) -1) && (sgid != current->gid) && |
| 1293 | (sgid != current->egid) && (sgid != current->sgid)) |
| 1294 | return -EPERM; |
| 1295 | } |
| 1296 | if (egid != (gid_t) -1) { |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1297 | if (egid != current->egid) { |
Alan Cox | d6e7114 | 2005-06-23 00:09:43 -0700 | [diff] [blame] | 1298 | current->mm->dumpable = suid_dumpable; |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 1299 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1300 | } |
| 1301 | current->egid = egid; |
| 1302 | } |
| 1303 | current->fsgid = current->egid; |
| 1304 | if (rgid != (gid_t) -1) |
| 1305 | current->gid = rgid; |
| 1306 | if (sgid != (gid_t) -1) |
| 1307 | current->sgid = sgid; |
| 1308 | |
| 1309 | key_fsgid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 1310 | proc_id_connector(current, PROC_EVENT_GID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1311 | return 0; |
| 1312 | } |
| 1313 | |
| 1314 | asmlinkage long sys_getresgid(gid_t __user *rgid, gid_t __user *egid, gid_t __user *sgid) |
| 1315 | { |
| 1316 | int retval; |
| 1317 | |
| 1318 | if (!(retval = put_user(current->gid, rgid)) && |
| 1319 | !(retval = put_user(current->egid, egid))) |
| 1320 | retval = put_user(current->sgid, sgid); |
| 1321 | |
| 1322 | return retval; |
| 1323 | } |
| 1324 | |
| 1325 | |
| 1326 | /* |
| 1327 | * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This |
| 1328 | * is used for "access()" and for the NFS daemon (letting nfsd stay at |
| 1329 | * whatever uid it wants to). It normally shadows "euid", except when |
| 1330 | * explicitly set by setfsuid() or for access.. |
| 1331 | */ |
| 1332 | asmlinkage long sys_setfsuid(uid_t uid) |
| 1333 | { |
| 1334 | int old_fsuid; |
| 1335 | |
| 1336 | old_fsuid = current->fsuid; |
| 1337 | if (security_task_setuid(uid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS)) |
| 1338 | return old_fsuid; |
| 1339 | |
| 1340 | if (uid == current->uid || uid == current->euid || |
| 1341 | uid == current->suid || uid == current->fsuid || |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1342 | capable(CAP_SETUID)) { |
| 1343 | if (uid != old_fsuid) { |
Alan Cox | d6e7114 | 2005-06-23 00:09:43 -0700 | [diff] [blame] | 1344 | current->mm->dumpable = suid_dumpable; |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 1345 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1346 | } |
| 1347 | current->fsuid = uid; |
| 1348 | } |
| 1349 | |
| 1350 | key_fsuid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 1351 | proc_id_connector(current, PROC_EVENT_UID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1352 | |
| 1353 | security_task_post_setuid(old_fsuid, (uid_t)-1, (uid_t)-1, LSM_SETID_FS); |
| 1354 | |
| 1355 | return old_fsuid; |
| 1356 | } |
| 1357 | |
| 1358 | /* |
John Anthony Kazos Jr | f42df9e | 2007-05-09 08:23:08 +0200 | [diff] [blame] | 1359 | * Samma på svenska.. |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1360 | */ |
| 1361 | asmlinkage long sys_setfsgid(gid_t gid) |
| 1362 | { |
| 1363 | int old_fsgid; |
| 1364 | |
| 1365 | old_fsgid = current->fsgid; |
| 1366 | if (security_task_setgid(gid, (gid_t)-1, (gid_t)-1, LSM_SETID_FS)) |
| 1367 | return old_fsgid; |
| 1368 | |
| 1369 | if (gid == current->gid || gid == current->egid || |
| 1370 | gid == current->sgid || gid == current->fsgid || |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1371 | capable(CAP_SETGID)) { |
| 1372 | if (gid != old_fsgid) { |
Alan Cox | d6e7114 | 2005-06-23 00:09:43 -0700 | [diff] [blame] | 1373 | current->mm->dumpable = suid_dumpable; |
akpm@osdl.org | d59dd46 | 2005-05-01 08:58:47 -0700 | [diff] [blame] | 1374 | smp_wmb(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1375 | } |
| 1376 | current->fsgid = gid; |
| 1377 | key_fsgid_changed(current); |
Matt Helsley | 9f46080 | 2005-11-07 00:59:16 -0800 | [diff] [blame] | 1378 | proc_id_connector(current, PROC_EVENT_GID); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1379 | } |
| 1380 | return old_fsgid; |
| 1381 | } |
| 1382 | |
| 1383 | asmlinkage long sys_times(struct tms __user * tbuf) |
| 1384 | { |
| 1385 | /* |
| 1386 | * In the SMP world we might just be unlucky and have one of |
| 1387 | * the times increment as we use it. Since the value is an |
| 1388 | * atomically safe type this is just fine. Conceptually its |
| 1389 | * as if the syscall took an instant longer to occur. |
| 1390 | */ |
| 1391 | if (tbuf) { |
| 1392 | struct tms tmp; |
Oleg Nesterov | 35f5cad | 2006-03-28 16:11:19 -0800 | [diff] [blame] | 1393 | struct task_struct *tsk = current; |
| 1394 | struct task_struct *t; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1395 | cputime_t utime, stime, cutime, cstime; |
| 1396 | |
Oleg Nesterov | 7d7185c | 2006-03-28 16:11:21 -0800 | [diff] [blame] | 1397 | spin_lock_irq(&tsk->sighand->siglock); |
Oleg Nesterov | 35f5cad | 2006-03-28 16:11:19 -0800 | [diff] [blame] | 1398 | utime = tsk->signal->utime; |
| 1399 | stime = tsk->signal->stime; |
| 1400 | t = tsk; |
| 1401 | do { |
| 1402 | utime = cputime_add(utime, t->utime); |
| 1403 | stime = cputime_add(stime, t->stime); |
| 1404 | t = next_thread(t); |
| 1405 | } while (t != tsk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1406 | |
Oleg Nesterov | 35f5cad | 2006-03-28 16:11:19 -0800 | [diff] [blame] | 1407 | cutime = tsk->signal->cutime; |
| 1408 | cstime = tsk->signal->cstime; |
| 1409 | spin_unlock_irq(&tsk->sighand->siglock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1410 | |
| 1411 | tmp.tms_utime = cputime_to_clock_t(utime); |
| 1412 | tmp.tms_stime = cputime_to_clock_t(stime); |
| 1413 | tmp.tms_cutime = cputime_to_clock_t(cutime); |
| 1414 | tmp.tms_cstime = cputime_to_clock_t(cstime); |
| 1415 | if (copy_to_user(tbuf, &tmp, sizeof(struct tms))) |
| 1416 | return -EFAULT; |
| 1417 | } |
| 1418 | return (long) jiffies_64_to_clock_t(get_jiffies_64()); |
| 1419 | } |
| 1420 | |
| 1421 | /* |
| 1422 | * This needs some heavy checking ... |
| 1423 | * I just haven't the stomach for it. I also don't fully |
| 1424 | * understand sessions/pgrp etc. Let somebody who does explain it. |
| 1425 | * |
| 1426 | * OK, I think I have the protection semantics right.... this is really |
| 1427 | * only important on a multi-user system anyway, to make sure one user |
| 1428 | * can't send a signal to a process owned by another. -TYT, 12/12/91 |
| 1429 | * |
| 1430 | * Auch. Had to add the 'did_exec' flag to conform completely to POSIX. |
| 1431 | * LBT 04.03.94 |
| 1432 | */ |
| 1433 | |
| 1434 | asmlinkage long sys_setpgid(pid_t pid, pid_t pgid) |
| 1435 | { |
| 1436 | struct task_struct *p; |
Oleg Nesterov | ee0acf9 | 2006-01-08 01:03:53 -0800 | [diff] [blame] | 1437 | struct task_struct *group_leader = current->group_leader; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1438 | int err = -EINVAL; |
| 1439 | |
| 1440 | if (!pid) |
Oleg Nesterov | ee0acf9 | 2006-01-08 01:03:53 -0800 | [diff] [blame] | 1441 | pid = group_leader->pid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1442 | if (!pgid) |
| 1443 | pgid = pid; |
| 1444 | if (pgid < 0) |
| 1445 | return -EINVAL; |
| 1446 | |
| 1447 | /* From this point forward we keep holding onto the tasklist lock |
| 1448 | * so that our parent does not change from under us. -DaveM |
| 1449 | */ |
| 1450 | write_lock_irq(&tasklist_lock); |
| 1451 | |
| 1452 | err = -ESRCH; |
| 1453 | p = find_task_by_pid(pid); |
| 1454 | if (!p) |
| 1455 | goto out; |
| 1456 | |
| 1457 | err = -EINVAL; |
| 1458 | if (!thread_group_leader(p)) |
| 1459 | goto out; |
| 1460 | |
Oleg Nesterov | f7dd795 | 2006-01-08 01:03:59 -0800 | [diff] [blame] | 1461 | if (p->real_parent == group_leader) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1462 | err = -EPERM; |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 1463 | if (task_session(p) != task_session(group_leader)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1464 | goto out; |
| 1465 | err = -EACCES; |
| 1466 | if (p->did_exec) |
| 1467 | goto out; |
| 1468 | } else { |
| 1469 | err = -ESRCH; |
Oleg Nesterov | ee0acf9 | 2006-01-08 01:03:53 -0800 | [diff] [blame] | 1470 | if (p != group_leader) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1471 | goto out; |
| 1472 | } |
| 1473 | |
| 1474 | err = -EPERM; |
| 1475 | if (p->signal->leader) |
| 1476 | goto out; |
| 1477 | |
| 1478 | if (pgid != pid) { |
Oleg Nesterov | f020bc4 | 2006-12-08 02:38:02 -0800 | [diff] [blame] | 1479 | struct task_struct *g = |
| 1480 | find_task_by_pid_type(PIDTYPE_PGID, pgid); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1481 | |
Eric W. Biederman | 41487c6 | 2007-02-12 00:53:01 -0800 | [diff] [blame] | 1482 | if (!g || task_session(g) != task_session(group_leader)) |
Oleg Nesterov | f020bc4 | 2006-12-08 02:38:02 -0800 | [diff] [blame] | 1483 | goto out; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1484 | } |
| 1485 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1486 | err = security_task_setpgid(p, pgid); |
| 1487 | if (err) |
| 1488 | goto out; |
| 1489 | |
| 1490 | if (process_group(p) != pgid) { |
| 1491 | detach_pid(p, PIDTYPE_PGID); |
| 1492 | p->signal->pgrp = pgid; |
Sukadev Bhattiprolu | e713d0d | 2007-05-10 22:22:58 -0700 | [diff] [blame] | 1493 | attach_pid(p, PIDTYPE_PGID, find_pid(pgid)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1494 | } |
| 1495 | |
| 1496 | err = 0; |
| 1497 | out: |
| 1498 | /* All paths lead to here, thus we are safe. -DaveM */ |
| 1499 | write_unlock_irq(&tasklist_lock); |
| 1500 | return err; |
| 1501 | } |
| 1502 | |
| 1503 | asmlinkage long sys_getpgid(pid_t pid) |
| 1504 | { |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1505 | if (!pid) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1506 | return process_group(current); |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1507 | else { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1508 | int retval; |
| 1509 | struct task_struct *p; |
| 1510 | |
| 1511 | read_lock(&tasklist_lock); |
| 1512 | p = find_task_by_pid(pid); |
| 1513 | |
| 1514 | retval = -ESRCH; |
| 1515 | if (p) { |
| 1516 | retval = security_task_getpgid(p); |
| 1517 | if (!retval) |
| 1518 | retval = process_group(p); |
| 1519 | } |
| 1520 | read_unlock(&tasklist_lock); |
| 1521 | return retval; |
| 1522 | } |
| 1523 | } |
| 1524 | |
| 1525 | #ifdef __ARCH_WANT_SYS_GETPGRP |
| 1526 | |
| 1527 | asmlinkage long sys_getpgrp(void) |
| 1528 | { |
| 1529 | /* SMP - assuming writes are word atomic this is fine */ |
| 1530 | return process_group(current); |
| 1531 | } |
| 1532 | |
| 1533 | #endif |
| 1534 | |
| 1535 | asmlinkage long sys_getsid(pid_t pid) |
| 1536 | { |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1537 | if (!pid) |
Cedric Le Goater | 937949d | 2006-12-08 02:37:54 -0800 | [diff] [blame] | 1538 | return process_session(current); |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1539 | else { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1540 | int retval; |
| 1541 | struct task_struct *p; |
| 1542 | |
| 1543 | read_lock(&tasklist_lock); |
| 1544 | p = find_task_by_pid(pid); |
| 1545 | |
| 1546 | retval = -ESRCH; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1547 | if (p) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1548 | retval = security_task_getsid(p); |
| 1549 | if (!retval) |
Cedric Le Goater | 937949d | 2006-12-08 02:37:54 -0800 | [diff] [blame] | 1550 | retval = process_session(p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1551 | } |
| 1552 | read_unlock(&tasklist_lock); |
| 1553 | return retval; |
| 1554 | } |
| 1555 | } |
| 1556 | |
| 1557 | asmlinkage long sys_setsid(void) |
| 1558 | { |
Oren Laadan | e19f247 | 2006-01-08 01:03:58 -0800 | [diff] [blame] | 1559 | struct task_struct *group_leader = current->group_leader; |
Eric W. Biederman | 390e2ff | 2006-03-31 02:31:33 -0800 | [diff] [blame] | 1560 | pid_t session; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1561 | int err = -EPERM; |
| 1562 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1563 | write_lock_irq(&tasklist_lock); |
| 1564 | |
Eric W. Biederman | 390e2ff | 2006-03-31 02:31:33 -0800 | [diff] [blame] | 1565 | /* Fail if I am already a session leader */ |
| 1566 | if (group_leader->signal->leader) |
| 1567 | goto out; |
| 1568 | |
| 1569 | session = group_leader->pid; |
| 1570 | /* Fail if a process group id already exists that equals the |
| 1571 | * proposed session id. |
| 1572 | * |
| 1573 | * Don't check if session id == 1 because kernel threads use this |
| 1574 | * session id and so the check will always fail and make it so |
| 1575 | * init cannot successfully call setsid. |
| 1576 | */ |
| 1577 | if (session > 1 && find_task_by_pid_type(PIDTYPE_PGID, session)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1578 | goto out; |
| 1579 | |
Oren Laadan | e19f247 | 2006-01-08 01:03:58 -0800 | [diff] [blame] | 1580 | group_leader->signal->leader = 1; |
Eric W. Biederman | 390e2ff | 2006-03-31 02:31:33 -0800 | [diff] [blame] | 1581 | __set_special_pids(session, session); |
Peter Zijlstra | 24ec839 | 2006-12-08 02:36:04 -0800 | [diff] [blame] | 1582 | |
| 1583 | spin_lock(&group_leader->sighand->siglock); |
Oren Laadan | e19f247 | 2006-01-08 01:03:58 -0800 | [diff] [blame] | 1584 | group_leader->signal->tty = NULL; |
Peter Zijlstra | 24ec839 | 2006-12-08 02:36:04 -0800 | [diff] [blame] | 1585 | spin_unlock(&group_leader->sighand->siglock); |
| 1586 | |
Oren Laadan | e19f247 | 2006-01-08 01:03:58 -0800 | [diff] [blame] | 1587 | err = process_group(group_leader); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1588 | out: |
| 1589 | write_unlock_irq(&tasklist_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1590 | return err; |
| 1591 | } |
| 1592 | |
| 1593 | /* |
| 1594 | * Supplementary group IDs |
| 1595 | */ |
| 1596 | |
| 1597 | /* init to 2 - one for init_task, one to ensure it is never freed */ |
| 1598 | struct group_info init_groups = { .usage = ATOMIC_INIT(2) }; |
| 1599 | |
| 1600 | struct group_info *groups_alloc(int gidsetsize) |
| 1601 | { |
| 1602 | struct group_info *group_info; |
| 1603 | int nblocks; |
| 1604 | int i; |
| 1605 | |
| 1606 | nblocks = (gidsetsize + NGROUPS_PER_BLOCK - 1) / NGROUPS_PER_BLOCK; |
| 1607 | /* Make sure we always allocate at least one indirect block pointer */ |
| 1608 | nblocks = nblocks ? : 1; |
| 1609 | group_info = kmalloc(sizeof(*group_info) + nblocks*sizeof(gid_t *), GFP_USER); |
| 1610 | if (!group_info) |
| 1611 | return NULL; |
| 1612 | group_info->ngroups = gidsetsize; |
| 1613 | group_info->nblocks = nblocks; |
| 1614 | atomic_set(&group_info->usage, 1); |
| 1615 | |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1616 | if (gidsetsize <= NGROUPS_SMALL) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1617 | group_info->blocks[0] = group_info->small_block; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1618 | else { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1619 | for (i = 0; i < nblocks; i++) { |
| 1620 | gid_t *b; |
| 1621 | b = (void *)__get_free_page(GFP_USER); |
| 1622 | if (!b) |
| 1623 | goto out_undo_partial_alloc; |
| 1624 | group_info->blocks[i] = b; |
| 1625 | } |
| 1626 | } |
| 1627 | return group_info; |
| 1628 | |
| 1629 | out_undo_partial_alloc: |
| 1630 | while (--i >= 0) { |
| 1631 | free_page((unsigned long)group_info->blocks[i]); |
| 1632 | } |
| 1633 | kfree(group_info); |
| 1634 | return NULL; |
| 1635 | } |
| 1636 | |
| 1637 | EXPORT_SYMBOL(groups_alloc); |
| 1638 | |
| 1639 | void groups_free(struct group_info *group_info) |
| 1640 | { |
| 1641 | if (group_info->blocks[0] != group_info->small_block) { |
| 1642 | int i; |
| 1643 | for (i = 0; i < group_info->nblocks; i++) |
| 1644 | free_page((unsigned long)group_info->blocks[i]); |
| 1645 | } |
| 1646 | kfree(group_info); |
| 1647 | } |
| 1648 | |
| 1649 | EXPORT_SYMBOL(groups_free); |
| 1650 | |
| 1651 | /* export the group_info to a user-space array */ |
| 1652 | static int groups_to_user(gid_t __user *grouplist, |
| 1653 | struct group_info *group_info) |
| 1654 | { |
| 1655 | int i; |
| 1656 | int count = group_info->ngroups; |
| 1657 | |
| 1658 | for (i = 0; i < group_info->nblocks; i++) { |
| 1659 | int cp_count = min(NGROUPS_PER_BLOCK, count); |
| 1660 | int off = i * NGROUPS_PER_BLOCK; |
| 1661 | int len = cp_count * sizeof(*grouplist); |
| 1662 | |
| 1663 | if (copy_to_user(grouplist+off, group_info->blocks[i], len)) |
| 1664 | return -EFAULT; |
| 1665 | |
| 1666 | count -= cp_count; |
| 1667 | } |
| 1668 | return 0; |
| 1669 | } |
| 1670 | |
| 1671 | /* fill a group_info from a user-space array - it must be allocated already */ |
| 1672 | static int groups_from_user(struct group_info *group_info, |
| 1673 | gid_t __user *grouplist) |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1674 | { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1675 | int i; |
| 1676 | int count = group_info->ngroups; |
| 1677 | |
| 1678 | for (i = 0; i < group_info->nblocks; i++) { |
| 1679 | int cp_count = min(NGROUPS_PER_BLOCK, count); |
| 1680 | int off = i * NGROUPS_PER_BLOCK; |
| 1681 | int len = cp_count * sizeof(*grouplist); |
| 1682 | |
| 1683 | if (copy_from_user(group_info->blocks[i], grouplist+off, len)) |
| 1684 | return -EFAULT; |
| 1685 | |
| 1686 | count -= cp_count; |
| 1687 | } |
| 1688 | return 0; |
| 1689 | } |
| 1690 | |
Domen Puncer | ebe8b54 | 2005-05-05 16:16:19 -0700 | [diff] [blame] | 1691 | /* a simple Shell sort */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1692 | static void groups_sort(struct group_info *group_info) |
| 1693 | { |
| 1694 | int base, max, stride; |
| 1695 | int gidsetsize = group_info->ngroups; |
| 1696 | |
| 1697 | for (stride = 1; stride < gidsetsize; stride = 3 * stride + 1) |
| 1698 | ; /* nothing */ |
| 1699 | stride /= 3; |
| 1700 | |
| 1701 | while (stride) { |
| 1702 | max = gidsetsize - stride; |
| 1703 | for (base = 0; base < max; base++) { |
| 1704 | int left = base; |
| 1705 | int right = left + stride; |
| 1706 | gid_t tmp = GROUP_AT(group_info, right); |
| 1707 | |
| 1708 | while (left >= 0 && GROUP_AT(group_info, left) > tmp) { |
| 1709 | GROUP_AT(group_info, right) = |
| 1710 | GROUP_AT(group_info, left); |
| 1711 | right = left; |
| 1712 | left -= stride; |
| 1713 | } |
| 1714 | GROUP_AT(group_info, right) = tmp; |
| 1715 | } |
| 1716 | stride /= 3; |
| 1717 | } |
| 1718 | } |
| 1719 | |
| 1720 | /* a simple bsearch */ |
David Howells | 3e30148 | 2005-06-23 22:00:56 -0700 | [diff] [blame] | 1721 | int groups_search(struct group_info *group_info, gid_t grp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1722 | { |
Eric Dumazet | d74beb9f | 2006-03-25 03:08:19 -0800 | [diff] [blame] | 1723 | unsigned int left, right; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1724 | |
| 1725 | if (!group_info) |
| 1726 | return 0; |
| 1727 | |
| 1728 | left = 0; |
| 1729 | right = group_info->ngroups; |
| 1730 | while (left < right) { |
Eric Dumazet | d74beb9f | 2006-03-25 03:08:19 -0800 | [diff] [blame] | 1731 | unsigned int mid = (left+right)/2; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1732 | int cmp = grp - GROUP_AT(group_info, mid); |
| 1733 | if (cmp > 0) |
| 1734 | left = mid + 1; |
| 1735 | else if (cmp < 0) |
| 1736 | right = mid; |
| 1737 | else |
| 1738 | return 1; |
| 1739 | } |
| 1740 | return 0; |
| 1741 | } |
| 1742 | |
| 1743 | /* validate and set current->group_info */ |
| 1744 | int set_current_groups(struct group_info *group_info) |
| 1745 | { |
| 1746 | int retval; |
| 1747 | struct group_info *old_info; |
| 1748 | |
| 1749 | retval = security_task_setgroups(group_info); |
| 1750 | if (retval) |
| 1751 | return retval; |
| 1752 | |
| 1753 | groups_sort(group_info); |
| 1754 | get_group_info(group_info); |
| 1755 | |
| 1756 | task_lock(current); |
| 1757 | old_info = current->group_info; |
| 1758 | current->group_info = group_info; |
| 1759 | task_unlock(current); |
| 1760 | |
| 1761 | put_group_info(old_info); |
| 1762 | |
| 1763 | return 0; |
| 1764 | } |
| 1765 | |
| 1766 | EXPORT_SYMBOL(set_current_groups); |
| 1767 | |
| 1768 | asmlinkage long sys_getgroups(int gidsetsize, gid_t __user *grouplist) |
| 1769 | { |
| 1770 | int i = 0; |
| 1771 | |
| 1772 | /* |
| 1773 | * SMP: Nobody else can change our grouplist. Thus we are |
| 1774 | * safe. |
| 1775 | */ |
| 1776 | |
| 1777 | if (gidsetsize < 0) |
| 1778 | return -EINVAL; |
| 1779 | |
| 1780 | /* no need to grab task_lock here; it cannot change */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1781 | i = current->group_info->ngroups; |
| 1782 | if (gidsetsize) { |
| 1783 | if (i > gidsetsize) { |
| 1784 | i = -EINVAL; |
| 1785 | goto out; |
| 1786 | } |
| 1787 | if (groups_to_user(grouplist, current->group_info)) { |
| 1788 | i = -EFAULT; |
| 1789 | goto out; |
| 1790 | } |
| 1791 | } |
| 1792 | out: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1793 | return i; |
| 1794 | } |
| 1795 | |
| 1796 | /* |
| 1797 | * SMP: Our groups are copy-on-write. We can set them safely |
| 1798 | * without another task interfering. |
| 1799 | */ |
| 1800 | |
| 1801 | asmlinkage long sys_setgroups(int gidsetsize, gid_t __user *grouplist) |
| 1802 | { |
| 1803 | struct group_info *group_info; |
| 1804 | int retval; |
| 1805 | |
| 1806 | if (!capable(CAP_SETGID)) |
| 1807 | return -EPERM; |
| 1808 | if ((unsigned)gidsetsize > NGROUPS_MAX) |
| 1809 | return -EINVAL; |
| 1810 | |
| 1811 | group_info = groups_alloc(gidsetsize); |
| 1812 | if (!group_info) |
| 1813 | return -ENOMEM; |
| 1814 | retval = groups_from_user(group_info, grouplist); |
| 1815 | if (retval) { |
| 1816 | put_group_info(group_info); |
| 1817 | return retval; |
| 1818 | } |
| 1819 | |
| 1820 | retval = set_current_groups(group_info); |
| 1821 | put_group_info(group_info); |
| 1822 | |
| 1823 | return retval; |
| 1824 | } |
| 1825 | |
| 1826 | /* |
| 1827 | * Check whether we're fsgid/egid or in the supplemental group.. |
| 1828 | */ |
| 1829 | int in_group_p(gid_t grp) |
| 1830 | { |
| 1831 | int retval = 1; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1832 | if (grp != current->fsgid) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1833 | retval = groups_search(current->group_info, grp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1834 | return retval; |
| 1835 | } |
| 1836 | |
| 1837 | EXPORT_SYMBOL(in_group_p); |
| 1838 | |
| 1839 | int in_egroup_p(gid_t grp) |
| 1840 | { |
| 1841 | int retval = 1; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1842 | if (grp != current->egid) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1843 | retval = groups_search(current->group_info, grp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1844 | return retval; |
| 1845 | } |
| 1846 | |
| 1847 | EXPORT_SYMBOL(in_egroup_p); |
| 1848 | |
| 1849 | DECLARE_RWSEM(uts_sem); |
| 1850 | |
David S. Miller | 393b072 | 2005-11-10 12:47:50 -0800 | [diff] [blame] | 1851 | EXPORT_SYMBOL(uts_sem); |
| 1852 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1853 | asmlinkage long sys_newuname(struct new_utsname __user * name) |
| 1854 | { |
| 1855 | int errno = 0; |
| 1856 | |
| 1857 | down_read(&uts_sem); |
Serge E. Hallyn | e9ff399 | 2006-10-02 02:18:11 -0700 | [diff] [blame] | 1858 | if (copy_to_user(name, utsname(), sizeof *name)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1859 | errno = -EFAULT; |
| 1860 | up_read(&uts_sem); |
| 1861 | return errno; |
| 1862 | } |
| 1863 | |
| 1864 | asmlinkage long sys_sethostname(char __user *name, int len) |
| 1865 | { |
| 1866 | int errno; |
| 1867 | char tmp[__NEW_UTS_LEN]; |
| 1868 | |
| 1869 | if (!capable(CAP_SYS_ADMIN)) |
| 1870 | return -EPERM; |
| 1871 | if (len < 0 || len > __NEW_UTS_LEN) |
| 1872 | return -EINVAL; |
| 1873 | down_write(&uts_sem); |
| 1874 | errno = -EFAULT; |
| 1875 | if (!copy_from_user(tmp, name, len)) { |
Serge E. Hallyn | e9ff399 | 2006-10-02 02:18:11 -0700 | [diff] [blame] | 1876 | memcpy(utsname()->nodename, tmp, len); |
| 1877 | utsname()->nodename[len] = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1878 | errno = 0; |
| 1879 | } |
| 1880 | up_write(&uts_sem); |
| 1881 | return errno; |
| 1882 | } |
| 1883 | |
| 1884 | #ifdef __ARCH_WANT_SYS_GETHOSTNAME |
| 1885 | |
| 1886 | asmlinkage long sys_gethostname(char __user *name, int len) |
| 1887 | { |
| 1888 | int i, errno; |
| 1889 | |
| 1890 | if (len < 0) |
| 1891 | return -EINVAL; |
| 1892 | down_read(&uts_sem); |
Serge E. Hallyn | e9ff399 | 2006-10-02 02:18:11 -0700 | [diff] [blame] | 1893 | i = 1 + strlen(utsname()->nodename); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1894 | if (i > len) |
| 1895 | i = len; |
| 1896 | errno = 0; |
Serge E. Hallyn | e9ff399 | 2006-10-02 02:18:11 -0700 | [diff] [blame] | 1897 | if (copy_to_user(name, utsname()->nodename, i)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1898 | errno = -EFAULT; |
| 1899 | up_read(&uts_sem); |
| 1900 | return errno; |
| 1901 | } |
| 1902 | |
| 1903 | #endif |
| 1904 | |
| 1905 | /* |
| 1906 | * Only setdomainname; getdomainname can be implemented by calling |
| 1907 | * uname() |
| 1908 | */ |
| 1909 | asmlinkage long sys_setdomainname(char __user *name, int len) |
| 1910 | { |
| 1911 | int errno; |
| 1912 | char tmp[__NEW_UTS_LEN]; |
| 1913 | |
| 1914 | if (!capable(CAP_SYS_ADMIN)) |
| 1915 | return -EPERM; |
| 1916 | if (len < 0 || len > __NEW_UTS_LEN) |
| 1917 | return -EINVAL; |
| 1918 | |
| 1919 | down_write(&uts_sem); |
| 1920 | errno = -EFAULT; |
| 1921 | if (!copy_from_user(tmp, name, len)) { |
Serge E. Hallyn | e9ff399 | 2006-10-02 02:18:11 -0700 | [diff] [blame] | 1922 | memcpy(utsname()->domainname, tmp, len); |
| 1923 | utsname()->domainname[len] = 0; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1924 | errno = 0; |
| 1925 | } |
| 1926 | up_write(&uts_sem); |
| 1927 | return errno; |
| 1928 | } |
| 1929 | |
| 1930 | asmlinkage long sys_getrlimit(unsigned int resource, struct rlimit __user *rlim) |
| 1931 | { |
| 1932 | if (resource >= RLIM_NLIMITS) |
| 1933 | return -EINVAL; |
| 1934 | else { |
| 1935 | struct rlimit value; |
| 1936 | task_lock(current->group_leader); |
| 1937 | value = current->signal->rlim[resource]; |
| 1938 | task_unlock(current->group_leader); |
| 1939 | return copy_to_user(rlim, &value, sizeof(*rlim)) ? -EFAULT : 0; |
| 1940 | } |
| 1941 | } |
| 1942 | |
| 1943 | #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT |
| 1944 | |
| 1945 | /* |
| 1946 | * Back compatibility for getrlimit. Needed for some apps. |
| 1947 | */ |
| 1948 | |
| 1949 | asmlinkage long sys_old_getrlimit(unsigned int resource, struct rlimit __user *rlim) |
| 1950 | { |
| 1951 | struct rlimit x; |
| 1952 | if (resource >= RLIM_NLIMITS) |
| 1953 | return -EINVAL; |
| 1954 | |
| 1955 | task_lock(current->group_leader); |
| 1956 | x = current->signal->rlim[resource]; |
| 1957 | task_unlock(current->group_leader); |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1958 | if (x.rlim_cur > 0x7FFFFFFF) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1959 | x.rlim_cur = 0x7FFFFFFF; |
Cal Peake | 756184b | 2006-09-30 23:27:24 -0700 | [diff] [blame] | 1960 | if (x.rlim_max > 0x7FFFFFFF) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1961 | x.rlim_max = 0x7FFFFFFF; |
| 1962 | return copy_to_user(rlim, &x, sizeof(x))?-EFAULT:0; |
| 1963 | } |
| 1964 | |
| 1965 | #endif |
| 1966 | |
| 1967 | asmlinkage long sys_setrlimit(unsigned int resource, struct rlimit __user *rlim) |
| 1968 | { |
| 1969 | struct rlimit new_rlim, *old_rlim; |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 1970 | unsigned long it_prof_secs; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1971 | int retval; |
| 1972 | |
| 1973 | if (resource >= RLIM_NLIMITS) |
| 1974 | return -EINVAL; |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 1975 | if (copy_from_user(&new_rlim, rlim, sizeof(*rlim))) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1976 | return -EFAULT; |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 1977 | if (new_rlim.rlim_cur > new_rlim.rlim_max) |
| 1978 | return -EINVAL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1979 | old_rlim = current->signal->rlim + resource; |
| 1980 | if ((new_rlim.rlim_max > old_rlim->rlim_max) && |
| 1981 | !capable(CAP_SYS_RESOURCE)) |
| 1982 | return -EPERM; |
| 1983 | if (resource == RLIMIT_NOFILE && new_rlim.rlim_max > NR_OPEN) |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 1984 | return -EPERM; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1985 | |
| 1986 | retval = security_task_setrlimit(resource, &new_rlim); |
| 1987 | if (retval) |
| 1988 | return retval; |
| 1989 | |
Tom Alsberg | 9926e4c | 2007-05-08 00:30:31 -0700 | [diff] [blame] | 1990 | if (resource == RLIMIT_CPU && new_rlim.rlim_cur == 0) { |
| 1991 | /* |
| 1992 | * The caller is asking for an immediate RLIMIT_CPU |
| 1993 | * expiry. But we use the zero value to mean "it was |
| 1994 | * never set". So let's cheat and make it one second |
| 1995 | * instead |
| 1996 | */ |
| 1997 | new_rlim.rlim_cur = 1; |
| 1998 | } |
| 1999 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2000 | task_lock(current->group_leader); |
| 2001 | *old_rlim = new_rlim; |
| 2002 | task_unlock(current->group_leader); |
| 2003 | |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 2004 | if (resource != RLIMIT_CPU) |
| 2005 | goto out; |
Andrew Morton | d3561f7 | 2006-03-24 03:18:36 -0800 | [diff] [blame] | 2006 | |
| 2007 | /* |
| 2008 | * RLIMIT_CPU handling. Note that the kernel fails to return an error |
| 2009 | * code if it rejected the user's attempt to set RLIMIT_CPU. This is a |
| 2010 | * very long-standing error, and fixing it now risks breakage of |
| 2011 | * applications, so we live with it |
| 2012 | */ |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 2013 | if (new_rlim.rlim_cur == RLIM_INFINITY) |
| 2014 | goto out; |
| 2015 | |
| 2016 | it_prof_secs = cputime_to_secs(current->signal->it_prof_expires); |
| 2017 | if (it_prof_secs == 0 || new_rlim.rlim_cur <= it_prof_secs) { |
Andrew Morton | e066111 | 2006-03-24 03:18:35 -0800 | [diff] [blame] | 2018 | unsigned long rlim_cur = new_rlim.rlim_cur; |
| 2019 | cputime_t cputime; |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 2020 | |
Andrew Morton | e066111 | 2006-03-24 03:18:35 -0800 | [diff] [blame] | 2021 | cputime = secs_to_cputime(rlim_cur); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2022 | read_lock(&tasklist_lock); |
| 2023 | spin_lock_irq(¤t->sighand->siglock); |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 2024 | set_process_cpu_timer(current, CPUCLOCK_PROF, &cputime, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2025 | spin_unlock_irq(¤t->sighand->siglock); |
| 2026 | read_unlock(&tasklist_lock); |
| 2027 | } |
Andrew Morton | ec9e16b | 2006-03-24 03:18:34 -0800 | [diff] [blame] | 2028 | out: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2029 | return 0; |
| 2030 | } |
| 2031 | |
| 2032 | /* |
| 2033 | * It would make sense to put struct rusage in the task_struct, |
| 2034 | * except that would make the task_struct be *really big*. After |
| 2035 | * task_struct gets moved into malloc'ed memory, it would |
| 2036 | * make sense to do this. It will make moving the rest of the information |
| 2037 | * a lot simpler! (Which we're not doing right now because we're not |
| 2038 | * measuring them yet). |
| 2039 | * |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2040 | * When sampling multiple threads for RUSAGE_SELF, under SMP we might have |
| 2041 | * races with threads incrementing their own counters. But since word |
| 2042 | * reads are atomic, we either get new values or old values and we don't |
| 2043 | * care which for the sums. We always take the siglock to protect reading |
| 2044 | * the c* fields from p->signal from races with exit.c updating those |
| 2045 | * fields when reaping, so a sample either gets all the additions of a |
| 2046 | * 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] | 2047 | * |
Ravikiran G Thirumalai | de047c1 | 2006-06-22 14:47:26 -0700 | [diff] [blame] | 2048 | * Locking: |
| 2049 | * We need to take the siglock for CHILDEREN, SELF and BOTH |
| 2050 | * for the cases current multithreaded, non-current single threaded |
| 2051 | * non-current multithreaded. Thread traversal is now safe with |
| 2052 | * the siglock held. |
| 2053 | * Strictly speaking, we donot need to take the siglock if we are current and |
| 2054 | * single threaded, as no one else can take our signal_struct away, no one |
| 2055 | * else can reap the children to update signal->c* counters, and no one else |
| 2056 | * can race with the signal-> fields. If we do not take any lock, the |
| 2057 | * signal-> fields could be read out of order while another thread was just |
| 2058 | * exiting. So we should place a read memory barrier when we avoid the lock. |
| 2059 | * On the writer side, write memory barrier is implied in __exit_signal |
| 2060 | * as __exit_signal releases the siglock spinlock after updating the signal-> |
| 2061 | * 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] | 2062 | * |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2063 | */ |
| 2064 | |
| 2065 | static void k_getrusage(struct task_struct *p, int who, struct rusage *r) |
| 2066 | { |
| 2067 | struct task_struct *t; |
| 2068 | unsigned long flags; |
| 2069 | cputime_t utime, stime; |
| 2070 | |
| 2071 | memset((char *) r, 0, sizeof *r); |
Oleg Nesterov | 0f59cc4 | 2006-01-08 01:05:15 -0800 | [diff] [blame] | 2072 | utime = stime = cputime_zero; |
| 2073 | |
Ravikiran G Thirumalai | de047c1 | 2006-06-22 14:47:26 -0700 | [diff] [blame] | 2074 | rcu_read_lock(); |
| 2075 | if (!lock_task_sighand(p, &flags)) { |
| 2076 | rcu_read_unlock(); |
| 2077 | return; |
| 2078 | } |
Ravikiran G Thirumalai | 2dd0ebc | 2006-03-23 03:00:13 -0800 | [diff] [blame] | 2079 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2080 | switch (who) { |
Oleg Nesterov | 0f59cc4 | 2006-01-08 01:05:15 -0800 | [diff] [blame] | 2081 | case RUSAGE_BOTH: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2082 | case RUSAGE_CHILDREN: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2083 | utime = p->signal->cutime; |
| 2084 | stime = p->signal->cstime; |
| 2085 | r->ru_nvcsw = p->signal->cnvcsw; |
| 2086 | r->ru_nivcsw = p->signal->cnivcsw; |
| 2087 | r->ru_minflt = p->signal->cmin_flt; |
| 2088 | r->ru_majflt = p->signal->cmaj_flt; |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 2089 | r->ru_inblock = p->signal->cinblock; |
| 2090 | r->ru_oublock = p->signal->coublock; |
Oleg Nesterov | 0f59cc4 | 2006-01-08 01:05:15 -0800 | [diff] [blame] | 2091 | |
| 2092 | if (who == RUSAGE_CHILDREN) |
| 2093 | break; |
| 2094 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2095 | case RUSAGE_SELF: |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2096 | utime = cputime_add(utime, p->signal->utime); |
| 2097 | stime = cputime_add(stime, p->signal->stime); |
| 2098 | r->ru_nvcsw += p->signal->nvcsw; |
| 2099 | r->ru_nivcsw += p->signal->nivcsw; |
| 2100 | r->ru_minflt += p->signal->min_flt; |
| 2101 | r->ru_majflt += p->signal->maj_flt; |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 2102 | r->ru_inblock += p->signal->inblock; |
| 2103 | r->ru_oublock += p->signal->oublock; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2104 | t = p; |
| 2105 | do { |
| 2106 | utime = cputime_add(utime, t->utime); |
| 2107 | stime = cputime_add(stime, t->stime); |
| 2108 | r->ru_nvcsw += t->nvcsw; |
| 2109 | r->ru_nivcsw += t->nivcsw; |
| 2110 | r->ru_minflt += t->min_flt; |
| 2111 | r->ru_majflt += t->maj_flt; |
Eric Dumazet | 6eaeeab | 2007-05-10 22:22:37 -0700 | [diff] [blame] | 2112 | r->ru_inblock += task_io_get_inblock(t); |
| 2113 | r->ru_oublock += task_io_get_oublock(t); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2114 | t = next_thread(t); |
| 2115 | } while (t != p); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2116 | break; |
Oleg Nesterov | 0f59cc4 | 2006-01-08 01:05:15 -0800 | [diff] [blame] | 2117 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2118 | default: |
| 2119 | BUG(); |
| 2120 | } |
Oleg Nesterov | 0f59cc4 | 2006-01-08 01:05:15 -0800 | [diff] [blame] | 2121 | |
Ravikiran G Thirumalai | de047c1 | 2006-06-22 14:47:26 -0700 | [diff] [blame] | 2122 | unlock_task_sighand(p, &flags); |
| 2123 | rcu_read_unlock(); |
| 2124 | |
Oleg Nesterov | 0f59cc4 | 2006-01-08 01:05:15 -0800 | [diff] [blame] | 2125 | cputime_to_timeval(utime, &r->ru_utime); |
| 2126 | cputime_to_timeval(stime, &r->ru_stime); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2127 | } |
| 2128 | |
| 2129 | int getrusage(struct task_struct *p, int who, struct rusage __user *ru) |
| 2130 | { |
| 2131 | struct rusage r; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2132 | k_getrusage(p, who, &r); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2133 | return copy_to_user(ru, &r, sizeof(r)) ? -EFAULT : 0; |
| 2134 | } |
| 2135 | |
| 2136 | asmlinkage long sys_getrusage(int who, struct rusage __user *ru) |
| 2137 | { |
| 2138 | if (who != RUSAGE_SELF && who != RUSAGE_CHILDREN) |
| 2139 | return -EINVAL; |
| 2140 | return getrusage(current, who, ru); |
| 2141 | } |
| 2142 | |
| 2143 | asmlinkage long sys_umask(int mask) |
| 2144 | { |
| 2145 | mask = xchg(¤t->fs->umask, mask & S_IRWXUGO); |
| 2146 | return mask; |
| 2147 | } |
| 2148 | |
| 2149 | asmlinkage long sys_prctl(int option, unsigned long arg2, unsigned long arg3, |
| 2150 | unsigned long arg4, unsigned long arg5) |
| 2151 | { |
| 2152 | long error; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2153 | |
| 2154 | error = security_task_prctl(option, arg2, arg3, arg4, arg5); |
| 2155 | if (error) |
| 2156 | return error; |
| 2157 | |
| 2158 | switch (option) { |
| 2159 | case PR_SET_PDEATHSIG: |
Jesper Juhl | 0730ded | 2005-09-06 15:17:37 -0700 | [diff] [blame] | 2160 | if (!valid_signal(arg2)) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2161 | error = -EINVAL; |
| 2162 | break; |
| 2163 | } |
Jesper Juhl | 0730ded | 2005-09-06 15:17:37 -0700 | [diff] [blame] | 2164 | current->pdeath_signal = arg2; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2165 | break; |
| 2166 | case PR_GET_PDEATHSIG: |
| 2167 | error = put_user(current->pdeath_signal, (int __user *)arg2); |
| 2168 | break; |
| 2169 | case PR_GET_DUMPABLE: |
Michael Kerrisk | 2030c0f | 2005-09-16 19:28:02 -0700 | [diff] [blame] | 2170 | error = current->mm->dumpable; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2171 | break; |
| 2172 | case PR_SET_DUMPABLE: |
Marcel Holtmann | abf75a5 | 2006-07-12 13:12:00 +0200 | [diff] [blame] | 2173 | if (arg2 < 0 || arg2 > 1) { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2174 | error = -EINVAL; |
| 2175 | break; |
| 2176 | } |
| 2177 | current->mm->dumpable = arg2; |
| 2178 | break; |
| 2179 | |
| 2180 | case PR_SET_UNALIGN: |
| 2181 | error = SET_UNALIGN_CTL(current, arg2); |
| 2182 | break; |
| 2183 | case PR_GET_UNALIGN: |
| 2184 | error = GET_UNALIGN_CTL(current, arg2); |
| 2185 | break; |
| 2186 | case PR_SET_FPEMU: |
| 2187 | error = SET_FPEMU_CTL(current, arg2); |
| 2188 | break; |
| 2189 | case PR_GET_FPEMU: |
| 2190 | error = GET_FPEMU_CTL(current, arg2); |
| 2191 | break; |
| 2192 | case PR_SET_FPEXC: |
| 2193 | error = SET_FPEXC_CTL(current, arg2); |
| 2194 | break; |
| 2195 | case PR_GET_FPEXC: |
| 2196 | error = GET_FPEXC_CTL(current, arg2); |
| 2197 | break; |
| 2198 | case PR_GET_TIMING: |
| 2199 | error = PR_TIMING_STATISTICAL; |
| 2200 | break; |
| 2201 | case PR_SET_TIMING: |
| 2202 | if (arg2 == PR_TIMING_STATISTICAL) |
| 2203 | error = 0; |
| 2204 | else |
| 2205 | error = -EINVAL; |
| 2206 | break; |
| 2207 | |
| 2208 | case PR_GET_KEEPCAPS: |
| 2209 | if (current->keep_capabilities) |
| 2210 | error = 1; |
| 2211 | break; |
| 2212 | case PR_SET_KEEPCAPS: |
| 2213 | if (arg2 != 0 && arg2 != 1) { |
| 2214 | error = -EINVAL; |
| 2215 | break; |
| 2216 | } |
| 2217 | current->keep_capabilities = arg2; |
| 2218 | break; |
| 2219 | case PR_SET_NAME: { |
| 2220 | struct task_struct *me = current; |
| 2221 | unsigned char ncomm[sizeof(me->comm)]; |
| 2222 | |
| 2223 | ncomm[sizeof(me->comm)-1] = 0; |
| 2224 | if (strncpy_from_user(ncomm, (char __user *)arg2, |
| 2225 | sizeof(me->comm)-1) < 0) |
| 2226 | return -EFAULT; |
| 2227 | set_task_comm(me, ncomm); |
| 2228 | return 0; |
| 2229 | } |
| 2230 | case PR_GET_NAME: { |
| 2231 | struct task_struct *me = current; |
| 2232 | unsigned char tcomm[sizeof(me->comm)]; |
| 2233 | |
| 2234 | get_task_comm(tcomm, me); |
| 2235 | if (copy_to_user((char __user *)arg2, tcomm, sizeof(tcomm))) |
| 2236 | return -EFAULT; |
| 2237 | return 0; |
| 2238 | } |
Anton Blanchard | 651d765 | 2006-06-07 16:10:19 +1000 | [diff] [blame] | 2239 | case PR_GET_ENDIAN: |
| 2240 | error = GET_ENDIAN(current, arg2); |
| 2241 | break; |
| 2242 | case PR_SET_ENDIAN: |
| 2243 | error = SET_ENDIAN(current, arg2); |
| 2244 | break; |
| 2245 | |
Andrea Arcangeli | 1d9d02f | 2007-07-15 23:41:32 -0700 | [diff] [blame] | 2246 | case PR_GET_SECCOMP: |
| 2247 | error = prctl_get_seccomp(); |
| 2248 | break; |
| 2249 | case PR_SET_SECCOMP: |
| 2250 | error = prctl_set_seccomp(arg2); |
| 2251 | break; |
| 2252 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2253 | default: |
| 2254 | error = -EINVAL; |
| 2255 | break; |
| 2256 | } |
| 2257 | return error; |
| 2258 | } |
Andi Kleen | 3cfc348 | 2006-09-26 10:52:28 +0200 | [diff] [blame] | 2259 | |
| 2260 | asmlinkage long sys_getcpu(unsigned __user *cpup, unsigned __user *nodep, |
| 2261 | struct getcpu_cache __user *cache) |
| 2262 | { |
| 2263 | int err = 0; |
| 2264 | int cpu = raw_smp_processor_id(); |
| 2265 | if (cpup) |
| 2266 | err |= put_user(cpu, cpup); |
| 2267 | if (nodep) |
| 2268 | err |= put_user(cpu_to_node(cpu), nodep); |
| 2269 | if (cache) { |
| 2270 | /* |
| 2271 | * The cache is not needed for this implementation, |
| 2272 | * but make sure user programs pass something |
| 2273 | * valid. vsyscall implementations can instead make |
| 2274 | * good use of the cache. Only use t0 and t1 because |
| 2275 | * these are available in both 32bit and 64bit ABI (no |
| 2276 | * need for a compat_getcpu). 32bit has enough |
| 2277 | * padding |
| 2278 | */ |
| 2279 | unsigned long t0, t1; |
Andi Kleen | 34596dc | 2006-09-30 01:47:55 +0200 | [diff] [blame] | 2280 | get_user(t0, &cache->blob[0]); |
| 2281 | get_user(t1, &cache->blob[1]); |
Andi Kleen | 3cfc348 | 2006-09-26 10:52:28 +0200 | [diff] [blame] | 2282 | t0++; |
| 2283 | t1++; |
Andi Kleen | 34596dc | 2006-09-30 01:47:55 +0200 | [diff] [blame] | 2284 | put_user(t0, &cache->blob[0]); |
| 2285 | put_user(t1, &cache->blob[1]); |
Andi Kleen | 3cfc348 | 2006-09-26 10:52:28 +0200 | [diff] [blame] | 2286 | } |
| 2287 | return err ? -EFAULT : 0; |
| 2288 | } |
Jeremy Fitzhardinge | 10a0a8d | 2007-07-17 18:37:02 -0700 | [diff] [blame^] | 2289 | |
| 2290 | char poweroff_cmd[POWEROFF_CMD_PATH_LEN] = "/sbin/poweroff"; |
| 2291 | |
| 2292 | static void argv_cleanup(char **argv, char **envp) |
| 2293 | { |
| 2294 | argv_free(argv); |
| 2295 | } |
| 2296 | |
| 2297 | /** |
| 2298 | * orderly_poweroff - Trigger an orderly system poweroff |
| 2299 | * @force: force poweroff if command execution fails |
| 2300 | * |
| 2301 | * This may be called from any context to trigger a system shutdown. |
| 2302 | * If the orderly shutdown fails, it will force an immediate shutdown. |
| 2303 | */ |
| 2304 | int orderly_poweroff(bool force) |
| 2305 | { |
| 2306 | int argc; |
| 2307 | char **argv = argv_split(GFP_ATOMIC, poweroff_cmd, &argc); |
| 2308 | static char *envp[] = { |
| 2309 | "HOME=/", |
| 2310 | "PATH=/sbin:/bin:/usr/sbin:/usr/bin", |
| 2311 | NULL |
| 2312 | }; |
| 2313 | int ret = -ENOMEM; |
| 2314 | struct subprocess_info *info; |
| 2315 | |
| 2316 | if (argv == NULL) { |
| 2317 | printk(KERN_WARNING "%s failed to allocate memory for \"%s\"\n", |
| 2318 | __func__, poweroff_cmd); |
| 2319 | goto out; |
| 2320 | } |
| 2321 | |
| 2322 | info = call_usermodehelper_setup(argv[0], argv, envp); |
| 2323 | if (info == NULL) { |
| 2324 | argv_free(argv); |
| 2325 | goto out; |
| 2326 | } |
| 2327 | |
| 2328 | call_usermodehelper_setcleanup(info, argv_cleanup); |
| 2329 | |
| 2330 | ret = call_usermodehelper_exec(info, -1); |
| 2331 | |
| 2332 | out: |
| 2333 | if (ret && force) { |
| 2334 | printk(KERN_WARNING "Failed to start orderly shutdown: " |
| 2335 | "forcing the issue\n"); |
| 2336 | |
| 2337 | /* I guess this should try to kick off some daemon to |
| 2338 | sync and poweroff asap. Or not even bother syncing |
| 2339 | if we're doing an emergency shutdown? */ |
| 2340 | emergency_sync(); |
| 2341 | kernel_power_off(); |
| 2342 | } |
| 2343 | |
| 2344 | return ret; |
| 2345 | } |
| 2346 | EXPORT_SYMBOL_GPL(orderly_poweroff); |