blob: b42fbd58973a02f2d59a29204a23c5323bcd960c [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/ipc/util.c
3 * Copyright (C) 1992 Krishna Balasubramanian
4 *
5 * Sep 1997 - Call suser() last after "normal" permission checks so we
6 * get BSD style process accounting right.
7 * Occurs in several places in the IPC code.
8 * Chris Evans, <chris@ferret.lmh.ox.ac.uk>
9 * Nov 1999 - ipc helper functions, unified SMP locking
Christian Kujau624dffc2006-01-15 02:43:54 +010010 * Manfred Spraul <manfred@colorfullife.com>
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * Oct 2002 - One lock per IPC id. RCU ipc_free for lock-free grow_ary().
12 * Mingming Cao <cmm@us.ibm.com>
Steve Grubb073115d2006-04-02 17:07:33 -040013 * Mar 2006 - support for audit of ipc object properties
14 * Dustin Kirkland <dustin.kirkland@us.ibm.com>
Kirill Korotaev73ea4132006-10-02 02:18:20 -070015 * Jun 2006 - namespaces ssupport
16 * OpenVZ, SWsoft Inc.
17 * Pavel Emelianov <xemul@openvz.org>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018 */
19
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/mm.h>
21#include <linux/shm.h>
22#include <linux/init.h>
23#include <linux/msg.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/vmalloc.h>
25#include <linux/slab.h>
Randy.Dunlapc59ede72006-01-11 12:17:46 -080026#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/highuid.h>
28#include <linux/security.h>
29#include <linux/rcupdate.h>
30#include <linux/workqueue.h>
Mike Waychisonae781772005-09-06 15:17:09 -070031#include <linux/seq_file.h>
32#include <linux/proc_fs.h>
Steve Grubb073115d2006-04-02 17:07:33 -040033#include <linux/audit.h>
Kirill Korotaev73ea4132006-10-02 02:18:20 -070034#include <linux/nsproxy.h>
Nadia Derbey3e148c72007-10-18 23:40:54 -070035#include <linux/rwsem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070036
37#include <asm/unistd.h>
38
39#include "util.h"
40
Mike Waychisonae781772005-09-06 15:17:09 -070041struct ipc_proc_iface {
42 const char *path;
43 const char *header;
Kirill Korotaev73ea4132006-10-02 02:18:20 -070044 int ids;
Mike Waychisonae781772005-09-06 15:17:09 -070045 int (*show)(struct seq_file *, void *);
46};
47
Kirill Korotaev73ea4132006-10-02 02:18:20 -070048struct ipc_namespace init_ipc_ns = {
49 .kref = {
50 .refcount = ATOMIC_INIT(2),
51 },
52};
53
Kirill Korotaev73ea4132006-10-02 02:18:20 -070054static struct ipc_namespace *clone_ipc_ns(struct ipc_namespace *old_ns)
55{
56 int err;
57 struct ipc_namespace *ns;
58
59 err = -ENOMEM;
60 ns = kmalloc(sizeof(struct ipc_namespace), GFP_KERNEL);
61 if (ns == NULL)
62 goto err_mem;
63
64 err = sem_init_ns(ns);
65 if (err)
66 goto err_sem;
67 err = msg_init_ns(ns);
68 if (err)
69 goto err_msg;
70 err = shm_init_ns(ns);
71 if (err)
72 goto err_shm;
73
74 kref_init(&ns->kref);
75 return ns;
76
77err_shm:
78 msg_exit_ns(ns);
79err_msg:
80 sem_exit_ns(ns);
81err_sem:
82 kfree(ns);
83err_mem:
84 return ERR_PTR(err);
85}
86
Badari Pulavartye3222c42007-05-08 00:25:21 -070087struct ipc_namespace *copy_ipcs(unsigned long flags, struct ipc_namespace *ns)
Kirill Korotaev73ea4132006-10-02 02:18:20 -070088{
Kirill Korotaev73ea4132006-10-02 02:18:20 -070089 struct ipc_namespace *new_ns;
Kirill Korotaev73ea4132006-10-02 02:18:20 -070090
Badari Pulavartye3222c42007-05-08 00:25:21 -070091 BUG_ON(!ns);
92 get_ipc_ns(ns);
Kirill Korotaev73ea4132006-10-02 02:18:20 -070093
94 if (!(flags & CLONE_NEWIPC))
Badari Pulavartye3222c42007-05-08 00:25:21 -070095 return ns;
Kirill Korotaev73ea4132006-10-02 02:18:20 -070096
Badari Pulavartye3222c42007-05-08 00:25:21 -070097 new_ns = clone_ipc_ns(ns);
Kirill Korotaev73ea4132006-10-02 02:18:20 -070098
Badari Pulavartye3222c42007-05-08 00:25:21 -070099 put_ipc_ns(ns);
100 return new_ns;
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700101}
102
103void free_ipc_ns(struct kref *kref)
104{
105 struct ipc_namespace *ns;
106
107 ns = container_of(kref, struct ipc_namespace, kref);
108 sem_exit_ns(ns);
109 msg_exit_ns(ns);
110 shm_exit_ns(ns);
111 kfree(ns);
112}
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700113
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114/**
115 * ipc_init - initialise IPC subsystem
116 *
117 * The various system5 IPC resources (semaphores, messages and shared
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800118 * memory) are initialised
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119 */
120
121static int __init ipc_init(void)
122{
123 sem_init();
124 msg_init();
125 shm_init();
126 return 0;
127}
128__initcall(ipc_init);
129
130/**
131 * ipc_init_ids - initialise IPC identifiers
132 * @ids: Identifier set
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133 *
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700134 * Set up the sequence range to use for the ipc identifier range (limited
135 * below IPCMNI) then initialise the ids idr.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 */
137
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700138void ipc_init_ids(struct ipc_ids *ids)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139{
Nadia Derbey3e148c72007-10-18 23:40:54 -0700140 init_rwsem(&ids->rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 ids->in_use = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 ids->seq = 0;
144 {
145 int seq_limit = INT_MAX/SEQ_MULTIPLIER;
146 if(seq_limit > USHRT_MAX)
147 ids->seq_max = USHRT_MAX;
148 else
149 ids->seq_max = seq_limit;
150 }
151
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700152 idr_init(&ids->ipcs_idr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153}
154
Mike Waychisonae781772005-09-06 15:17:09 -0700155#ifdef CONFIG_PROC_FS
Arjan van de Ven9a321442007-02-12 00:55:35 -0800156static const struct file_operations sysvipc_proc_fops;
Mike Waychisonae781772005-09-06 15:17:09 -0700157/**
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800158 * ipc_init_proc_interface - Create a proc interface for sysipc types using a seq_file interface.
Mike Waychisonae781772005-09-06 15:17:09 -0700159 * @path: Path in procfs
160 * @header: Banner to be printed at the beginning of the file.
161 * @ids: ipc id table to iterate.
162 * @show: show routine.
163 */
164void __init ipc_init_proc_interface(const char *path, const char *header,
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700165 int ids, int (*show)(struct seq_file *, void *))
Mike Waychisonae781772005-09-06 15:17:09 -0700166{
167 struct proc_dir_entry *pde;
168 struct ipc_proc_iface *iface;
169
170 iface = kmalloc(sizeof(*iface), GFP_KERNEL);
171 if (!iface)
172 return;
173 iface->path = path;
174 iface->header = header;
175 iface->ids = ids;
176 iface->show = show;
177
178 pde = create_proc_entry(path,
179 S_IRUGO, /* world readable */
180 NULL /* parent dir */);
181 if (pde) {
182 pde->data = iface;
183 pde->proc_fops = &sysvipc_proc_fops;
184 } else {
185 kfree(iface);
186 }
187}
188#endif
189
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190/**
191 * ipc_findkey - find a key in an ipc identifier set
192 * @ids: Identifier set
193 * @key: The key to find
194 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700195 * Requires ipc_ids.rw_mutex locked.
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700196 * Returns the LOCKED pointer to the ipc structure if found or NULL
197 * if not.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700198 * If key is found ipc points to the owning ipc structure
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199 */
200
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700201static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700203 struct kern_ipc_perm *ipc;
204 int next_id;
205 int total;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700207 for (total = 0, next_id = 0; total < ids->in_use; next_id++) {
208 ipc = idr_find(&ids->ipcs_idr, next_id);
209
210 if (ipc == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211 continue;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700212
213 if (ipc->key != key) {
214 total++;
215 continue;
216 }
217
218 ipc_lock_by_ptr(ipc);
219 return ipc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700220 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700221
222 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700223}
224
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700225/**
226 * ipc_get_maxid - get the last assigned id
227 * @ids: IPC identifier set
228 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700229 * Called with ipc_ids.rw_mutex held.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 */
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700231
232int ipc_get_maxid(struct ipc_ids *ids)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700234 struct kern_ipc_perm *ipc;
235 int max_id = -1;
236 int total, id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700238 if (ids->in_use == 0)
239 return -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700241 if (ids->in_use == IPCMNI)
242 return IPCMNI - 1;
243
244 /* Look for the last assigned id */
245 total = 0;
246 for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
247 ipc = idr_find(&ids->ipcs_idr, id);
248 if (ipc != NULL) {
249 max_id = id;
250 total++;
251 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700253 return max_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254}
255
256/**
257 * ipc_addid - add an IPC identifier
258 * @ids: IPC identifier set
259 * @new: new IPC permission set
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700260 * @size: limit for the number of used ids
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700262 * Add an entry 'new' to the IPC ids idr. The permissions object is
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 * initialised and the first free entry is set up and the id assigned
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700264 * is returned. The 'new' entry is returned in a locked state on success.
265 * On failure the entry is not locked and -1 is returned.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700267 * Called with ipc_ids.rw_mutex held as a writer.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268 */
269
270int ipc_addid(struct ipc_ids* ids, struct kern_ipc_perm* new, int size)
271{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700272 int id, err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700274 if (size > IPCMNI)
275 size = IPCMNI;
276
277 if (ids->in_use >= size)
278 return -1;
279
280 err = idr_get_new(&ids->ipcs_idr, new, &id);
281 if (err)
282 return -1;
283
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284 ids->in_use++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285
286 new->cuid = new->uid = current->euid;
287 new->gid = new->cgid = current->egid;
288
289 new->seq = ids->seq++;
290 if(ids->seq > ids->seq_max)
291 ids->seq = 0;
292
293 spin_lock_init(&new->lock);
294 new->deleted = 0;
295 rcu_read_lock();
296 spin_lock(&new->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 return id;
298}
299
300/**
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700301 * ipcget_new - create a new ipc object
302 * @ns: namespace
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700303 * @ids: IPC identifer set
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700304 * @ops: the actual creation routine to call
305 * @params: its parameters
306 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700307 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
308 * when the key is IPC_PRIVATE.
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700309 */
310int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
311 struct ipc_ops *ops, struct ipc_params *params)
312{
313 int err;
314
315 err = idr_pre_get(&ids->ipcs_idr, GFP_KERNEL);
316
317 if (!err)
318 return -ENOMEM;
319
Nadia Derbey3e148c72007-10-18 23:40:54 -0700320 down_write(&ids->rw_mutex);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700321 err = ops->getnew(ns, params);
Nadia Derbey3e148c72007-10-18 23:40:54 -0700322 up_write(&ids->rw_mutex);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700323
324 return err;
325}
326
327/**
328 * ipc_check_perms - check security and permissions for an IPC
329 * @ipcp: ipc permission set
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700330 * @ops: the actual security routine to call
331 * @params: its parameters
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700332 *
333 * This routine is called by sys_msgget(), sys_semget() and sys_shmget()
334 * when the key is not IPC_PRIVATE and that key already exists in the
335 * ids IDR.
336 *
337 * On success, the IPC id is returned.
338 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700339 * It is called with ipc_ids.rw_mutex and ipcp->lock held.
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700340 */
341static int ipc_check_perms(struct kern_ipc_perm *ipcp, struct ipc_ops *ops,
342 struct ipc_params *params)
343{
344 int err;
345
346 if (ipcperms(ipcp, params->flg))
347 err = -EACCES;
348 else {
349 err = ops->associate(ipcp, params->flg);
350 if (!err)
351 err = ipcp->id;
352 }
353
354 return err;
355}
356
357/**
358 * ipcget_public - get an ipc object or create a new one
359 * @ns: namespace
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700360 * @ids: IPC identifer set
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700361 * @ops: the actual creation routine to call
362 * @params: its parameters
363 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700364 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
365 * when the key is not IPC_PRIVATE.
366 * It adds a new entry if the key is not found and does some permission
367 * / security checkings if the key is found.
368 *
369 * On success, the ipc id is returned.
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700370 */
371int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
372 struct ipc_ops *ops, struct ipc_params *params)
373{
374 struct kern_ipc_perm *ipcp;
375 int flg = params->flg;
376 int err;
377
378 err = idr_pre_get(&ids->ipcs_idr, GFP_KERNEL);
379
Nadia Derbey3e148c72007-10-18 23:40:54 -0700380 /*
381 * Take the lock as a writer since we are potentially going to add
382 * a new entry + read locks are not "upgradable"
383 */
384 down_write(&ids->rw_mutex);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700385 ipcp = ipc_findkey(ids, params->key);
386 if (ipcp == NULL) {
387 /* key not used */
388 if (!(flg & IPC_CREAT))
389 err = -ENOENT;
390 else if (!err)
391 err = -ENOMEM;
392 else
393 err = ops->getnew(ns, params);
394 } else {
395 /* ipc object has been locked by ipc_findkey() */
396
397 if (flg & IPC_CREAT && flg & IPC_EXCL)
398 err = -EEXIST;
399 else {
400 err = 0;
401 if (ops->more_checks)
402 err = ops->more_checks(ipcp, params);
403 if (!err)
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700404 /*
405 * ipc_check_perms returns the IPC id on
406 * success
407 */
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700408 err = ipc_check_perms(ipcp, ops, params);
409 }
410 ipc_unlock(ipcp);
411 }
Nadia Derbey3e148c72007-10-18 23:40:54 -0700412 up_write(&ids->rw_mutex);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700413
414 return err;
415}
416
417
418/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 * ipc_rmid - remove an IPC identifier
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700420 * @ids: IPC identifier set
421 * @ipcp: ipc perm structure containing the identifier to remove
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700423 * ipc_ids.rw_mutex (as a writer) and the spinlock for this ID are held
424 * before this function is called, and remain locked on the exit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 */
426
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700427void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428{
Nadia Derbeyce621f52007-10-18 23:40:52 -0700429 int lid = ipcid_to_idx(ipcp->id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700431 idr_remove(&ids->ipcs_idr, lid);
432
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 ids->in_use--;
434
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700435 ipcp->deleted = 1;
436
437 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438}
439
440/**
441 * ipc_alloc - allocate ipc space
442 * @size: size desired
443 *
444 * Allocate memory from the appropriate pools and return a pointer to it.
445 * NULL is returned if the allocation fails
446 */
447
448void* ipc_alloc(int size)
449{
450 void* out;
451 if(size > PAGE_SIZE)
452 out = vmalloc(size);
453 else
454 out = kmalloc(size, GFP_KERNEL);
455 return out;
456}
457
458/**
459 * ipc_free - free ipc space
460 * @ptr: pointer returned by ipc_alloc
461 * @size: size of block
462 *
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800463 * Free a block created with ipc_alloc(). The caller must know the size
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 * used in the allocation call.
465 */
466
467void ipc_free(void* ptr, int size)
468{
469 if(size > PAGE_SIZE)
470 vfree(ptr);
471 else
472 kfree(ptr);
473}
474
475/*
476 * rcu allocations:
477 * There are three headers that are prepended to the actual allocation:
478 * - during use: ipc_rcu_hdr.
479 * - during the rcu grace period: ipc_rcu_grace.
480 * - [only if vmalloc]: ipc_rcu_sched.
481 * Their lifetime doesn't overlap, thus the headers share the same memory.
482 * Unlike a normal union, they are right-aligned, thus some container_of
483 * forward/backward casting is necessary:
484 */
485struct ipc_rcu_hdr
486{
487 int refcount;
488 int is_vmalloc;
489 void *data[0];
490};
491
492
493struct ipc_rcu_grace
494{
495 struct rcu_head rcu;
496 /* "void *" makes sure alignment of following data is sane. */
497 void *data[0];
498};
499
500struct ipc_rcu_sched
501{
502 struct work_struct work;
503 /* "void *" makes sure alignment of following data is sane. */
504 void *data[0];
505};
506
507#define HDRLEN_KMALLOC (sizeof(struct ipc_rcu_grace) > sizeof(struct ipc_rcu_hdr) ? \
508 sizeof(struct ipc_rcu_grace) : sizeof(struct ipc_rcu_hdr))
509#define HDRLEN_VMALLOC (sizeof(struct ipc_rcu_sched) > HDRLEN_KMALLOC ? \
510 sizeof(struct ipc_rcu_sched) : HDRLEN_KMALLOC)
511
512static inline int rcu_use_vmalloc(int size)
513{
514 /* Too big for a single page? */
515 if (HDRLEN_KMALLOC + size > PAGE_SIZE)
516 return 1;
517 return 0;
518}
519
520/**
521 * ipc_rcu_alloc - allocate ipc and rcu space
522 * @size: size desired
523 *
524 * Allocate memory for the rcu header structure + the object.
525 * Returns the pointer to the object.
526 * NULL is returned if the allocation fails.
527 */
528
529void* ipc_rcu_alloc(int size)
530{
531 void* out;
532 /*
533 * We prepend the allocation with the rcu struct, and
534 * workqueue if necessary (for vmalloc).
535 */
536 if (rcu_use_vmalloc(size)) {
537 out = vmalloc(HDRLEN_VMALLOC + size);
538 if (out) {
539 out += HDRLEN_VMALLOC;
540 container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 1;
541 container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
542 }
543 } else {
544 out = kmalloc(HDRLEN_KMALLOC + size, GFP_KERNEL);
545 if (out) {
546 out += HDRLEN_KMALLOC;
547 container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 0;
548 container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
549 }
550 }
551
552 return out;
553}
554
555void ipc_rcu_getref(void *ptr)
556{
557 container_of(ptr, struct ipc_rcu_hdr, data)->refcount++;
558}
559
David Howells65f27f32006-11-22 14:55:48 +0000560static void ipc_do_vfree(struct work_struct *work)
561{
562 vfree(container_of(work, struct ipc_rcu_sched, work));
563}
564
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565/**
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800566 * ipc_schedule_free - free ipc + rcu space
567 * @head: RCU callback structure for queued work
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 *
569 * Since RCU callback function is called in bh,
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800570 * we need to defer the vfree to schedule_work().
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 */
572static void ipc_schedule_free(struct rcu_head *head)
573{
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700574 struct ipc_rcu_grace *grace;
575 struct ipc_rcu_sched *sched;
576
577 grace = container_of(head, struct ipc_rcu_grace, rcu);
578 sched = container_of(&(grace->data[0]), struct ipc_rcu_sched,
579 data[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580
David Howells65f27f32006-11-22 14:55:48 +0000581 INIT_WORK(&sched->work, ipc_do_vfree);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 schedule_work(&sched->work);
583}
584
585/**
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800586 * ipc_immediate_free - free ipc + rcu space
587 * @head: RCU callback structure that contains pointer to be freed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 *
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800589 * Free from the RCU callback context.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 */
591static void ipc_immediate_free(struct rcu_head *head)
592{
593 struct ipc_rcu_grace *free =
594 container_of(head, struct ipc_rcu_grace, rcu);
595 kfree(free);
596}
597
598void ipc_rcu_putref(void *ptr)
599{
600 if (--container_of(ptr, struct ipc_rcu_hdr, data)->refcount > 0)
601 return;
602
603 if (container_of(ptr, struct ipc_rcu_hdr, data)->is_vmalloc) {
604 call_rcu(&container_of(ptr, struct ipc_rcu_grace, data)->rcu,
605 ipc_schedule_free);
606 } else {
607 call_rcu(&container_of(ptr, struct ipc_rcu_grace, data)->rcu,
608 ipc_immediate_free);
609 }
610}
611
612/**
613 * ipcperms - check IPC permissions
614 * @ipcp: IPC permission set
615 * @flag: desired permission set.
616 *
617 * Check user, group, other permissions for access
618 * to ipc resources. return 0 if allowed
619 */
620
621int ipcperms (struct kern_ipc_perm *ipcp, short flag)
622{ /* flag will most probably be 0 or S_...UGO from <linux/stat.h> */
Steve Grubb073115d2006-04-02 17:07:33 -0400623 int requested_mode, granted_mode, err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624
Steve Grubb073115d2006-04-02 17:07:33 -0400625 if (unlikely((err = audit_ipc_obj(ipcp))))
626 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 requested_mode = (flag >> 6) | (flag >> 3) | flag;
628 granted_mode = ipcp->mode;
629 if (current->euid == ipcp->cuid || current->euid == ipcp->uid)
630 granted_mode >>= 6;
631 else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
632 granted_mode >>= 3;
633 /* is there some bit set in requested_mode but not in granted_mode? */
634 if ((requested_mode & ~granted_mode & 0007) &&
635 !capable(CAP_IPC_OWNER))
636 return -1;
637
638 return security_ipc_permission(ipcp, flag);
639}
640
641/*
642 * Functions to convert between the kern_ipc_perm structure and the
643 * old/new ipc_perm structures
644 */
645
646/**
647 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
648 * @in: kernel permissions
649 * @out: new style IPC permissions
650 *
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800651 * Turn the kernel object @in into a set of permissions descriptions
652 * for returning to userspace (@out).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 */
654
655
656void kernel_to_ipc64_perm (struct kern_ipc_perm *in, struct ipc64_perm *out)
657{
658 out->key = in->key;
659 out->uid = in->uid;
660 out->gid = in->gid;
661 out->cuid = in->cuid;
662 out->cgid = in->cgid;
663 out->mode = in->mode;
664 out->seq = in->seq;
665}
666
667/**
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700668 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 * @in: new style IPC permissions
670 * @out: old style IPC permissions
671 *
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800672 * Turn the new style permissions object @in into a compatibility
673 * object and store it into the @out pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674 */
675
676void ipc64_perm_to_ipc_perm (struct ipc64_perm *in, struct ipc_perm *out)
677{
678 out->key = in->key;
679 SET_UID(out->uid, in->uid);
680 SET_GID(out->gid, in->gid);
681 SET_UID(out->cuid, in->cuid);
682 SET_GID(out->cgid, in->cgid);
683 out->mode = in->mode;
684 out->seq = in->seq;
685}
686
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700687/**
Nadia Derbey3e148c72007-10-18 23:40:54 -0700688 * ipc_lock - Lock an ipc structure without rw_mutex held
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700689 * @ids: IPC identifier set
690 * @id: ipc id to look for
691 *
692 * Look for an id in the ipc ids idr and lock the associated ipc object.
693 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700694 * The ipc object is locked on exit.
Nadia Derbey3e148c72007-10-18 23:40:54 -0700695 *
696 * This is the routine that should be called when the rw_mutex is not already
697 * held, i.e. idr tree not protected: it protects the idr tree in read mode
698 * during the idr_find().
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700699 */
700
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700701struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700703 struct kern_ipc_perm *out;
Nadia Derbeyce621f52007-10-18 23:40:52 -0700704 int lid = ipcid_to_idx(id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705
Nadia Derbey3e148c72007-10-18 23:40:54 -0700706 down_read(&ids->rw_mutex);
707
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 rcu_read_lock();
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700709 out = idr_find(&ids->ipcs_idr, lid);
710 if (out == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711 rcu_read_unlock();
Nadia Derbey3e148c72007-10-18 23:40:54 -0700712 up_read(&ids->rw_mutex);
Nadia Derbey023a5352007-10-18 23:40:51 -0700713 return ERR_PTR(-EINVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700715
Nadia Derbey3e148c72007-10-18 23:40:54 -0700716 up_read(&ids->rw_mutex);
717
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 spin_lock(&out->lock);
719
720 /* ipc_rmid() may have already freed the ID while ipc_lock
721 * was spinning: here verify that the structure is still valid
722 */
723 if (out->deleted) {
724 spin_unlock(&out->lock);
725 rcu_read_unlock();
Nadia Derbey023a5352007-10-18 23:40:51 -0700726 return ERR_PTR(-EINVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700728
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 return out;
730}
731
Nadia Derbey3e148c72007-10-18 23:40:54 -0700732/**
733 * ipc_lock_down - Lock an ipc structure with rw_sem held
734 * @ids: IPC identifier set
735 * @id: ipc id to look for
736 *
737 * Look for an id in the ipc ids idr and lock the associated ipc object.
738 *
739 * The ipc object is locked on exit.
740 *
741 * This is the routine that should be called when the rw_mutex is already
742 * held, i.e. idr tree protected.
743 */
744
745struct kern_ipc_perm *ipc_lock_down(struct ipc_ids *ids, int id)
746{
747 struct kern_ipc_perm *out;
748 int lid = ipcid_to_idx(id);
749
750 rcu_read_lock();
751 out = idr_find(&ids->ipcs_idr, lid);
752 if (out == NULL) {
753 rcu_read_unlock();
754 return ERR_PTR(-EINVAL);
755 }
756
757 spin_lock(&out->lock);
758
759 /*
760 * No need to verify that the structure is still valid since the
761 * rw_mutex is held.
762 */
763 return out;
764}
765
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766#ifdef __ARCH_WANT_IPC_PARSE_VERSION
767
768
769/**
770 * ipc_parse_version - IPC call version
771 * @cmd: pointer to command
772 *
773 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800774 * The @cmd value is turned from an encoding command and version into
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775 * just the command code.
776 */
777
778int ipc_parse_version (int *cmd)
779{
780 if (*cmd & IPC_64) {
781 *cmd ^= IPC_64;
782 return IPC_64;
783 } else {
784 return IPC_OLD;
785 }
786}
787
788#endif /* __ARCH_WANT_IPC_PARSE_VERSION */
Mike Waychisonae781772005-09-06 15:17:09 -0700789
790#ifdef CONFIG_PROC_FS
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800791struct ipc_proc_iter {
792 struct ipc_namespace *ns;
793 struct ipc_proc_iface *iface;
794};
795
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700796/*
797 * This routine locks the ipc structure found at least at position pos.
798 */
799struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
800 loff_t *new_pos)
Mike Waychisonae781772005-09-06 15:17:09 -0700801{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700802 struct kern_ipc_perm *ipc;
803 int total, id;
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700804
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700805 total = 0;
806 for (id = 0; id < pos && total < ids->in_use; id++) {
807 ipc = idr_find(&ids->ipcs_idr, id);
808 if (ipc != NULL)
809 total++;
810 }
Mike Waychisonae781772005-09-06 15:17:09 -0700811
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700812 if (total >= ids->in_use)
813 return NULL;
Mike Waychisonae781772005-09-06 15:17:09 -0700814
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700815 for ( ; pos < IPCMNI; pos++) {
816 ipc = idr_find(&ids->ipcs_idr, pos);
817 if (ipc != NULL) {
818 *new_pos = pos + 1;
819 ipc_lock_by_ptr(ipc);
Mike Waychisonae781772005-09-06 15:17:09 -0700820 return ipc;
821 }
822 }
823
824 /* Out of range - return NULL to terminate iteration */
825 return NULL;
826}
827
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700828static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
829{
830 struct ipc_proc_iter *iter = s->private;
831 struct ipc_proc_iface *iface = iter->iface;
832 struct kern_ipc_perm *ipc = it;
833
834 /* If we had an ipc id locked before, unlock it */
835 if (ipc && ipc != SEQ_START_TOKEN)
836 ipc_unlock(ipc);
837
838 return sysvipc_find_ipc(iter->ns->ids[iface->ids], *pos, pos);
839}
840
Mike Waychisonae781772005-09-06 15:17:09 -0700841/*
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700842 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
843 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
Mike Waychisonae781772005-09-06 15:17:09 -0700844 */
845static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
846{
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800847 struct ipc_proc_iter *iter = s->private;
848 struct ipc_proc_iface *iface = iter->iface;
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700849 struct ipc_ids *ids;
850
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800851 ids = iter->ns->ids[iface->ids];
Mike Waychisonae781772005-09-06 15:17:09 -0700852
853 /*
854 * Take the lock - this will be released by the corresponding
855 * call to stop().
856 */
Nadia Derbey3e148c72007-10-18 23:40:54 -0700857 down_read(&ids->rw_mutex);
Mike Waychisonae781772005-09-06 15:17:09 -0700858
859 /* pos < 0 is invalid */
860 if (*pos < 0)
861 return NULL;
862
863 /* pos == 0 means header */
864 if (*pos == 0)
865 return SEQ_START_TOKEN;
866
867 /* Find the (pos-1)th ipc */
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700868 return sysvipc_find_ipc(ids, *pos - 1, pos);
Mike Waychisonae781772005-09-06 15:17:09 -0700869}
870
871static void sysvipc_proc_stop(struct seq_file *s, void *it)
872{
873 struct kern_ipc_perm *ipc = it;
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800874 struct ipc_proc_iter *iter = s->private;
875 struct ipc_proc_iface *iface = iter->iface;
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700876 struct ipc_ids *ids;
Mike Waychisonae781772005-09-06 15:17:09 -0700877
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700878 /* If we had a locked structure, release it */
Mike Waychisonae781772005-09-06 15:17:09 -0700879 if (ipc && ipc != SEQ_START_TOKEN)
880 ipc_unlock(ipc);
881
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800882 ids = iter->ns->ids[iface->ids];
Mike Waychisonae781772005-09-06 15:17:09 -0700883 /* Release the lock we took in start() */
Nadia Derbey3e148c72007-10-18 23:40:54 -0700884 up_read(&ids->rw_mutex);
Mike Waychisonae781772005-09-06 15:17:09 -0700885}
886
887static int sysvipc_proc_show(struct seq_file *s, void *it)
888{
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800889 struct ipc_proc_iter *iter = s->private;
890 struct ipc_proc_iface *iface = iter->iface;
Mike Waychisonae781772005-09-06 15:17:09 -0700891
892 if (it == SEQ_START_TOKEN)
893 return seq_puts(s, iface->header);
894
895 return iface->show(s, it);
896}
897
898static struct seq_operations sysvipc_proc_seqops = {
899 .start = sysvipc_proc_start,
900 .stop = sysvipc_proc_stop,
901 .next = sysvipc_proc_next,
902 .show = sysvipc_proc_show,
903};
904
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800905static int sysvipc_proc_open(struct inode *inode, struct file *file)
906{
Mike Waychisonae781772005-09-06 15:17:09 -0700907 int ret;
908 struct seq_file *seq;
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800909 struct ipc_proc_iter *iter;
910
911 ret = -ENOMEM;
912 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
913 if (!iter)
914 goto out;
Mike Waychisonae781772005-09-06 15:17:09 -0700915
916 ret = seq_open(file, &sysvipc_proc_seqops);
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800917 if (ret)
918 goto out_kfree;
919
920 seq = file->private_data;
921 seq->private = iter;
922
923 iter->iface = PDE(inode)->data;
924 iter->ns = get_ipc_ns(current->nsproxy->ipc_ns);
925out:
Mike Waychisonae781772005-09-06 15:17:09 -0700926 return ret;
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800927out_kfree:
928 kfree(iter);
929 goto out;
930}
931
932static int sysvipc_proc_release(struct inode *inode, struct file *file)
933{
934 struct seq_file *seq = file->private_data;
935 struct ipc_proc_iter *iter = seq->private;
936 put_ipc_ns(iter->ns);
937 return seq_release_private(inode, file);
Mike Waychisonae781772005-09-06 15:17:09 -0700938}
939
Arjan van de Ven9a321442007-02-12 00:55:35 -0800940static const struct file_operations sysvipc_proc_fops = {
Mike Waychisonae781772005-09-06 15:17:09 -0700941 .open = sysvipc_proc_open,
942 .read = seq_read,
943 .llseek = seq_lseek,
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800944 .release = sysvipc_proc_release,
Mike Waychisonae781772005-09-06 15:17:09 -0700945};
946#endif /* CONFIG_PROC_FS */