blob: cb017c7b9370bb09ee501cd36b3db5534cefdd21 [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>
Pavel Emelyanovae5e1b22008-02-08 04:18:22 -080036#include <linux/ipc_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070037
38#include <asm/unistd.h>
39
40#include "util.h"
41
Mike Waychisonae781772005-09-06 15:17:09 -070042struct ipc_proc_iface {
43 const char *path;
44 const char *header;
Kirill Korotaev73ea4132006-10-02 02:18:20 -070045 int ids;
Mike Waychisonae781772005-09-06 15:17:09 -070046 int (*show)(struct seq_file *, void *);
47};
48
Kirill Korotaev73ea4132006-10-02 02:18:20 -070049struct ipc_namespace init_ipc_ns = {
50 .kref = {
51 .refcount = ATOMIC_INIT(2),
52 },
53};
54
Linus Torvalds1da177e2005-04-16 15:20:36 -070055/**
56 * ipc_init - initialise IPC subsystem
57 *
58 * The various system5 IPC resources (semaphores, messages and shared
Robert P. J. Day72fd4a32007-02-10 01:45:59 -080059 * memory) are initialised
Linus Torvalds1da177e2005-04-16 15:20:36 -070060 */
61
62static int __init ipc_init(void)
63{
64 sem_init();
65 msg_init();
66 shm_init();
67 return 0;
68}
69__initcall(ipc_init);
70
71/**
72 * ipc_init_ids - initialise IPC identifiers
73 * @ids: Identifier set
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 *
Nadia Derbey7ca7e562007-10-18 23:40:48 -070075 * Set up the sequence range to use for the ipc identifier range (limited
76 * below IPCMNI) then initialise the ids idr.
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 */
78
Nadia Derbey7ca7e562007-10-18 23:40:48 -070079void ipc_init_ids(struct ipc_ids *ids)
Linus Torvalds1da177e2005-04-16 15:20:36 -070080{
Nadia Derbey3e148c72007-10-18 23:40:54 -070081 init_rwsem(&ids->rw_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -070082
Linus Torvalds1da177e2005-04-16 15:20:36 -070083 ids->in_use = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070084 ids->seq = 0;
85 {
86 int seq_limit = INT_MAX/SEQ_MULTIPLIER;
87 if(seq_limit > USHRT_MAX)
88 ids->seq_max = USHRT_MAX;
89 else
90 ids->seq_max = seq_limit;
91 }
92
Nadia Derbey7ca7e562007-10-18 23:40:48 -070093 idr_init(&ids->ipcs_idr);
Linus Torvalds1da177e2005-04-16 15:20:36 -070094}
95
Mike Waychisonae781772005-09-06 15:17:09 -070096#ifdef CONFIG_PROC_FS
Arjan van de Ven9a321442007-02-12 00:55:35 -080097static const struct file_operations sysvipc_proc_fops;
Mike Waychisonae781772005-09-06 15:17:09 -070098/**
Robert P. J. Day72fd4a32007-02-10 01:45:59 -080099 * ipc_init_proc_interface - Create a proc interface for sysipc types using a seq_file interface.
Mike Waychisonae781772005-09-06 15:17:09 -0700100 * @path: Path in procfs
101 * @header: Banner to be printed at the beginning of the file.
102 * @ids: ipc id table to iterate.
103 * @show: show routine.
104 */
105void __init ipc_init_proc_interface(const char *path, const char *header,
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700106 int ids, int (*show)(struct seq_file *, void *))
Mike Waychisonae781772005-09-06 15:17:09 -0700107{
108 struct proc_dir_entry *pde;
109 struct ipc_proc_iface *iface;
110
111 iface = kmalloc(sizeof(*iface), GFP_KERNEL);
112 if (!iface)
113 return;
114 iface->path = path;
115 iface->header = header;
116 iface->ids = ids;
117 iface->show = show;
118
119 pde = create_proc_entry(path,
120 S_IRUGO, /* world readable */
121 NULL /* parent dir */);
122 if (pde) {
123 pde->data = iface;
124 pde->proc_fops = &sysvipc_proc_fops;
125 } else {
126 kfree(iface);
127 }
128}
129#endif
130
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131/**
132 * ipc_findkey - find a key in an ipc identifier set
133 * @ids: Identifier set
134 * @key: The key to find
135 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700136 * Requires ipc_ids.rw_mutex locked.
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700137 * Returns the LOCKED pointer to the ipc structure if found or NULL
138 * if not.
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700139 * If key is found ipc points to the owning ipc structure
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140 */
141
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700142static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700144 struct kern_ipc_perm *ipc;
145 int next_id;
146 int total;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700148 for (total = 0, next_id = 0; total < ids->in_use; next_id++) {
149 ipc = idr_find(&ids->ipcs_idr, next_id);
150
151 if (ipc == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 continue;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700153
154 if (ipc->key != key) {
155 total++;
156 continue;
157 }
158
159 ipc_lock_by_ptr(ipc);
160 return ipc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700162
163 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164}
165
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700166/**
167 * ipc_get_maxid - get the last assigned id
168 * @ids: IPC identifier set
169 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700170 * Called with ipc_ids.rw_mutex held.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 */
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700172
173int ipc_get_maxid(struct ipc_ids *ids)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700175 struct kern_ipc_perm *ipc;
176 int max_id = -1;
177 int total, id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700179 if (ids->in_use == 0)
180 return -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700182 if (ids->in_use == IPCMNI)
183 return IPCMNI - 1;
184
185 /* Look for the last assigned id */
186 total = 0;
187 for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
188 ipc = idr_find(&ids->ipcs_idr, id);
189 if (ipc != NULL) {
190 max_id = id;
191 total++;
192 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700193 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700194 return max_id;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195}
196
197/**
198 * ipc_addid - add an IPC identifier
199 * @ids: IPC identifier set
200 * @new: new IPC permission set
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700201 * @size: limit for the number of used ids
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700203 * Add an entry 'new' to the IPC ids idr. The permissions object is
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 * initialised and the first free entry is set up and the id assigned
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700205 * is returned. The 'new' entry is returned in a locked state on success.
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700206 * On failure the entry is not locked and a negative err-code is returned.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700208 * Called with ipc_ids.rw_mutex held as a writer.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 */
210
211int ipc_addid(struct ipc_ids* ids, struct kern_ipc_perm* new, int size)
212{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700213 int id, err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700215 if (size > IPCMNI)
216 size = IPCMNI;
217
218 if (ids->in_use >= size)
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700219 return -ENOSPC;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700220
221 err = idr_get_new(&ids->ipcs_idr, new, &id);
222 if (err)
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700223 return err;
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700224
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225 ids->in_use++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226
227 new->cuid = new->uid = current->euid;
228 new->gid = new->cgid = current->egid;
229
230 new->seq = ids->seq++;
231 if(ids->seq > ids->seq_max)
232 ids->seq = 0;
233
Pierre Peiffer48dea402008-04-29 01:00:35 -0700234 new->id = ipc_buildid(id, new->seq);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235 spin_lock_init(&new->lock);
236 new->deleted = 0;
237 rcu_read_lock();
238 spin_lock(&new->lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 return id;
240}
241
242/**
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700243 * ipcget_new - create a new ipc object
244 * @ns: namespace
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700245 * @ids: IPC identifer set
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700246 * @ops: the actual creation routine to call
247 * @params: its parameters
248 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700249 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
250 * when the key is IPC_PRIVATE.
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700251 */
Pavel Emelyanovb2d75cd2008-02-08 04:18:54 -0800252static int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700253 struct ipc_ops *ops, struct ipc_params *params)
254{
255 int err;
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700256retry:
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700257 err = idr_pre_get(&ids->ipcs_idr, GFP_KERNEL);
258
259 if (!err)
260 return -ENOMEM;
261
Nadia Derbey3e148c72007-10-18 23:40:54 -0700262 down_write(&ids->rw_mutex);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700263 err = ops->getnew(ns, params);
Nadia Derbey3e148c72007-10-18 23:40:54 -0700264 up_write(&ids->rw_mutex);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700265
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700266 if (err == -EAGAIN)
267 goto retry;
268
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700269 return err;
270}
271
272/**
273 * ipc_check_perms - check security and permissions for an IPC
274 * @ipcp: ipc permission set
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700275 * @ops: the actual security routine to call
276 * @params: its parameters
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700277 *
278 * This routine is called by sys_msgget(), sys_semget() and sys_shmget()
279 * when the key is not IPC_PRIVATE and that key already exists in the
280 * ids IDR.
281 *
282 * On success, the IPC id is returned.
283 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700284 * It is called with ipc_ids.rw_mutex and ipcp->lock held.
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700285 */
286static int ipc_check_perms(struct kern_ipc_perm *ipcp, struct ipc_ops *ops,
287 struct ipc_params *params)
288{
289 int err;
290
291 if (ipcperms(ipcp, params->flg))
292 err = -EACCES;
293 else {
294 err = ops->associate(ipcp, params->flg);
295 if (!err)
296 err = ipcp->id;
297 }
298
299 return err;
300}
301
302/**
303 * ipcget_public - get an ipc object or create a new one
304 * @ns: namespace
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700305 * @ids: IPC identifer set
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700306 * @ops: the actual creation routine to call
307 * @params: its parameters
308 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700309 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
310 * when the key is not IPC_PRIVATE.
311 * It adds a new entry if the key is not found and does some permission
312 * / security checkings if the key is found.
313 *
314 * On success, the ipc id is returned.
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700315 */
Pavel Emelyanovb2d75cd2008-02-08 04:18:54 -0800316static int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700317 struct ipc_ops *ops, struct ipc_params *params)
318{
319 struct kern_ipc_perm *ipcp;
320 int flg = params->flg;
321 int err;
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700322retry:
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700323 err = idr_pre_get(&ids->ipcs_idr, GFP_KERNEL);
324
Nadia Derbey3e148c72007-10-18 23:40:54 -0700325 /*
326 * Take the lock as a writer since we are potentially going to add
327 * a new entry + read locks are not "upgradable"
328 */
329 down_write(&ids->rw_mutex);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700330 ipcp = ipc_findkey(ids, params->key);
331 if (ipcp == NULL) {
332 /* key not used */
333 if (!(flg & IPC_CREAT))
334 err = -ENOENT;
335 else if (!err)
336 err = -ENOMEM;
337 else
338 err = ops->getnew(ns, params);
339 } else {
340 /* ipc object has been locked by ipc_findkey() */
341
342 if (flg & IPC_CREAT && flg & IPC_EXCL)
343 err = -EEXIST;
344 else {
345 err = 0;
346 if (ops->more_checks)
347 err = ops->more_checks(ipcp, params);
348 if (!err)
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700349 /*
350 * ipc_check_perms returns the IPC id on
351 * success
352 */
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700353 err = ipc_check_perms(ipcp, ops, params);
354 }
355 ipc_unlock(ipcp);
356 }
Nadia Derbey3e148c72007-10-18 23:40:54 -0700357 up_write(&ids->rw_mutex);
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700358
Pierre Peiffer283bb7f2007-10-18 23:40:57 -0700359 if (err == -EAGAIN)
360 goto retry;
361
Nadia Derbey7748dbf2007-10-18 23:40:49 -0700362 return err;
363}
364
365
366/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 * ipc_rmid - remove an IPC identifier
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700368 * @ids: IPC identifier set
369 * @ipcp: ipc perm structure containing the identifier to remove
Linus Torvalds1da177e2005-04-16 15:20:36 -0700370 *
Nadia Derbey3e148c72007-10-18 23:40:54 -0700371 * ipc_ids.rw_mutex (as a writer) and the spinlock for this ID are held
372 * before this function is called, and remain locked on the exit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700373 */
374
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700375void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376{
Nadia Derbeyce621f52007-10-18 23:40:52 -0700377 int lid = ipcid_to_idx(ipcp->id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700379 idr_remove(&ids->ipcs_idr, lid);
380
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 ids->in_use--;
382
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700383 ipcp->deleted = 1;
384
385 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386}
387
388/**
389 * ipc_alloc - allocate ipc space
390 * @size: size desired
391 *
392 * Allocate memory from the appropriate pools and return a pointer to it.
393 * NULL is returned if the allocation fails
394 */
395
396void* ipc_alloc(int size)
397{
398 void* out;
399 if(size > PAGE_SIZE)
400 out = vmalloc(size);
401 else
402 out = kmalloc(size, GFP_KERNEL);
403 return out;
404}
405
406/**
407 * ipc_free - free ipc space
408 * @ptr: pointer returned by ipc_alloc
409 * @size: size of block
410 *
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800411 * Free a block created with ipc_alloc(). The caller must know the size
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 * used in the allocation call.
413 */
414
415void ipc_free(void* ptr, int size)
416{
417 if(size > PAGE_SIZE)
418 vfree(ptr);
419 else
420 kfree(ptr);
421}
422
423/*
424 * rcu allocations:
425 * There are three headers that are prepended to the actual allocation:
426 * - during use: ipc_rcu_hdr.
427 * - during the rcu grace period: ipc_rcu_grace.
428 * - [only if vmalloc]: ipc_rcu_sched.
429 * Their lifetime doesn't overlap, thus the headers share the same memory.
430 * Unlike a normal union, they are right-aligned, thus some container_of
431 * forward/backward casting is necessary:
432 */
433struct ipc_rcu_hdr
434{
435 int refcount;
436 int is_vmalloc;
437 void *data[0];
438};
439
440
441struct ipc_rcu_grace
442{
443 struct rcu_head rcu;
444 /* "void *" makes sure alignment of following data is sane. */
445 void *data[0];
446};
447
448struct ipc_rcu_sched
449{
450 struct work_struct work;
451 /* "void *" makes sure alignment of following data is sane. */
452 void *data[0];
453};
454
455#define HDRLEN_KMALLOC (sizeof(struct ipc_rcu_grace) > sizeof(struct ipc_rcu_hdr) ? \
456 sizeof(struct ipc_rcu_grace) : sizeof(struct ipc_rcu_hdr))
457#define HDRLEN_VMALLOC (sizeof(struct ipc_rcu_sched) > HDRLEN_KMALLOC ? \
458 sizeof(struct ipc_rcu_sched) : HDRLEN_KMALLOC)
459
460static inline int rcu_use_vmalloc(int size)
461{
462 /* Too big for a single page? */
463 if (HDRLEN_KMALLOC + size > PAGE_SIZE)
464 return 1;
465 return 0;
466}
467
468/**
469 * ipc_rcu_alloc - allocate ipc and rcu space
470 * @size: size desired
471 *
472 * Allocate memory for the rcu header structure + the object.
473 * Returns the pointer to the object.
474 * NULL is returned if the allocation fails.
475 */
476
477void* ipc_rcu_alloc(int size)
478{
479 void* out;
480 /*
481 * We prepend the allocation with the rcu struct, and
482 * workqueue if necessary (for vmalloc).
483 */
484 if (rcu_use_vmalloc(size)) {
485 out = vmalloc(HDRLEN_VMALLOC + size);
486 if (out) {
487 out += HDRLEN_VMALLOC;
488 container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 1;
489 container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
490 }
491 } else {
492 out = kmalloc(HDRLEN_KMALLOC + size, GFP_KERNEL);
493 if (out) {
494 out += HDRLEN_KMALLOC;
495 container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 0;
496 container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
497 }
498 }
499
500 return out;
501}
502
503void ipc_rcu_getref(void *ptr)
504{
505 container_of(ptr, struct ipc_rcu_hdr, data)->refcount++;
506}
507
David Howells65f27f32006-11-22 14:55:48 +0000508static void ipc_do_vfree(struct work_struct *work)
509{
510 vfree(container_of(work, struct ipc_rcu_sched, work));
511}
512
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513/**
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800514 * ipc_schedule_free - free ipc + rcu space
515 * @head: RCU callback structure for queued work
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 *
517 * Since RCU callback function is called in bh,
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800518 * we need to defer the vfree to schedule_work().
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 */
520static void ipc_schedule_free(struct rcu_head *head)
521{
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700522 struct ipc_rcu_grace *grace;
523 struct ipc_rcu_sched *sched;
524
525 grace = container_of(head, struct ipc_rcu_grace, rcu);
526 sched = container_of(&(grace->data[0]), struct ipc_rcu_sched,
527 data[0]);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528
David Howells65f27f32006-11-22 14:55:48 +0000529 INIT_WORK(&sched->work, ipc_do_vfree);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 schedule_work(&sched->work);
531}
532
533/**
Randy Dunlap1e5d5332005-11-07 01:01:06 -0800534 * ipc_immediate_free - free ipc + rcu space
535 * @head: RCU callback structure that contains pointer to be freed
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 *
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800537 * Free from the RCU callback context.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 */
539static void ipc_immediate_free(struct rcu_head *head)
540{
541 struct ipc_rcu_grace *free =
542 container_of(head, struct ipc_rcu_grace, rcu);
543 kfree(free);
544}
545
546void ipc_rcu_putref(void *ptr)
547{
548 if (--container_of(ptr, struct ipc_rcu_hdr, data)->refcount > 0)
549 return;
550
551 if (container_of(ptr, struct ipc_rcu_hdr, data)->is_vmalloc) {
552 call_rcu(&container_of(ptr, struct ipc_rcu_grace, data)->rcu,
553 ipc_schedule_free);
554 } else {
555 call_rcu(&container_of(ptr, struct ipc_rcu_grace, data)->rcu,
556 ipc_immediate_free);
557 }
558}
559
560/**
561 * ipcperms - check IPC permissions
562 * @ipcp: IPC permission set
563 * @flag: desired permission set.
564 *
565 * Check user, group, other permissions for access
566 * to ipc resources. return 0 if allowed
567 */
568
569int ipcperms (struct kern_ipc_perm *ipcp, short flag)
570{ /* flag will most probably be 0 or S_...UGO from <linux/stat.h> */
Steve Grubb073115d2006-04-02 17:07:33 -0400571 int requested_mode, granted_mode, err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572
Steve Grubb073115d2006-04-02 17:07:33 -0400573 if (unlikely((err = audit_ipc_obj(ipcp))))
574 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 requested_mode = (flag >> 6) | (flag >> 3) | flag;
576 granted_mode = ipcp->mode;
577 if (current->euid == ipcp->cuid || current->euid == ipcp->uid)
578 granted_mode >>= 6;
579 else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
580 granted_mode >>= 3;
581 /* is there some bit set in requested_mode but not in granted_mode? */
582 if ((requested_mode & ~granted_mode & 0007) &&
583 !capable(CAP_IPC_OWNER))
584 return -1;
585
586 return security_ipc_permission(ipcp, flag);
587}
588
589/*
590 * Functions to convert between the kern_ipc_perm structure and the
591 * old/new ipc_perm structures
592 */
593
594/**
595 * kernel_to_ipc64_perm - convert kernel ipc permissions to user
596 * @in: kernel permissions
597 * @out: new style IPC permissions
598 *
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800599 * Turn the kernel object @in into a set of permissions descriptions
600 * for returning to userspace (@out).
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 */
602
603
604void kernel_to_ipc64_perm (struct kern_ipc_perm *in, struct ipc64_perm *out)
605{
606 out->key = in->key;
607 out->uid = in->uid;
608 out->gid = in->gid;
609 out->cuid = in->cuid;
610 out->cgid = in->cgid;
611 out->mode = in->mode;
612 out->seq = in->seq;
613}
614
615/**
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700616 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 * @in: new style IPC permissions
618 * @out: old style IPC permissions
619 *
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800620 * Turn the new style permissions object @in into a compatibility
621 * object and store it into the @out pointer.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 */
623
624void ipc64_perm_to_ipc_perm (struct ipc64_perm *in, struct ipc_perm *out)
625{
626 out->key = in->key;
627 SET_UID(out->uid, in->uid);
628 SET_GID(out->gid, in->gid);
629 SET_UID(out->cuid, in->cuid);
630 SET_GID(out->cgid, in->cgid);
631 out->mode = in->mode;
632 out->seq = in->seq;
633}
634
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700635/**
Nadia Derbey3e148c72007-10-18 23:40:54 -0700636 * ipc_lock - Lock an ipc structure without rw_mutex held
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700637 * @ids: IPC identifier set
638 * @id: ipc id to look for
639 *
640 * Look for an id in the ipc ids idr and lock the associated ipc object.
641 *
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700642 * The ipc object is locked on exit.
Nadia Derbey3e148c72007-10-18 23:40:54 -0700643 *
644 * This is the routine that should be called when the rw_mutex is not already
645 * held, i.e. idr tree not protected: it protects the idr tree in read mode
646 * during the idr_find().
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700647 */
648
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700649struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700651 struct kern_ipc_perm *out;
Nadia Derbeyce621f52007-10-18 23:40:52 -0700652 int lid = ipcid_to_idx(id);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653
Nadia Derbey3e148c72007-10-18 23:40:54 -0700654 down_read(&ids->rw_mutex);
655
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656 rcu_read_lock();
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700657 out = idr_find(&ids->ipcs_idr, lid);
658 if (out == NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 rcu_read_unlock();
Nadia Derbey3e148c72007-10-18 23:40:54 -0700660 up_read(&ids->rw_mutex);
Nadia Derbey023a5352007-10-18 23:40:51 -0700661 return ERR_PTR(-EINVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700663
Nadia Derbey3e148c72007-10-18 23:40:54 -0700664 up_read(&ids->rw_mutex);
665
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666 spin_lock(&out->lock);
667
668 /* ipc_rmid() may have already freed the ID while ipc_lock
669 * was spinning: here verify that the structure is still valid
670 */
671 if (out->deleted) {
672 spin_unlock(&out->lock);
673 rcu_read_unlock();
Nadia Derbey023a5352007-10-18 23:40:51 -0700674 return ERR_PTR(-EINVAL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 }
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700676
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 return out;
678}
679
Nadia Derbey3e148c72007-10-18 23:40:54 -0700680/**
681 * ipc_lock_down - Lock an ipc structure with rw_sem held
682 * @ids: IPC identifier set
683 * @id: ipc id to look for
684 *
685 * Look for an id in the ipc ids idr and lock the associated ipc object.
686 *
687 * The ipc object is locked on exit.
688 *
689 * This is the routine that should be called when the rw_mutex is already
690 * held, i.e. idr tree protected.
691 */
692
693struct kern_ipc_perm *ipc_lock_down(struct ipc_ids *ids, int id)
694{
695 struct kern_ipc_perm *out;
696 int lid = ipcid_to_idx(id);
697
698 rcu_read_lock();
699 out = idr_find(&ids->ipcs_idr, lid);
700 if (out == NULL) {
701 rcu_read_unlock();
702 return ERR_PTR(-EINVAL);
703 }
704
705 spin_lock(&out->lock);
706
707 /*
708 * No need to verify that the structure is still valid since the
709 * rw_mutex is held.
710 */
711 return out;
712}
713
Pavel Emelyanovb2d75cd2008-02-08 04:18:54 -0800714struct kern_ipc_perm *ipc_lock_check_down(struct ipc_ids *ids, int id)
715{
716 struct kern_ipc_perm *out;
717
718 out = ipc_lock_down(ids, id);
719 if (IS_ERR(out))
720 return out;
721
722 if (ipc_checkid(out, id)) {
723 ipc_unlock(out);
724 return ERR_PTR(-EIDRM);
725 }
726
727 return out;
728}
729
730struct kern_ipc_perm *ipc_lock_check(struct ipc_ids *ids, int id)
731{
732 struct kern_ipc_perm *out;
733
734 out = ipc_lock(ids, id);
735 if (IS_ERR(out))
736 return out;
737
738 if (ipc_checkid(out, id)) {
739 ipc_unlock(out);
740 return ERR_PTR(-EIDRM);
741 }
742
743 return out;
744}
745
746/**
747 * ipcget - Common sys_*get() code
748 * @ns : namsepace
749 * @ids : IPC identifier set
750 * @ops : operations to be called on ipc object creation, permission checks
751 * and further checks
752 * @params : the parameters needed by the previous operations.
753 *
754 * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
755 */
756int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
757 struct ipc_ops *ops, struct ipc_params *params)
758{
759 if (params->key == IPC_PRIVATE)
760 return ipcget_new(ns, ids, ops, params);
761 else
762 return ipcget_public(ns, ids, ops, params);
763}
764
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765#ifdef __ARCH_WANT_IPC_PARSE_VERSION
766
767
768/**
769 * ipc_parse_version - IPC call version
770 * @cmd: pointer to command
771 *
772 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
Robert P. J. Day72fd4a32007-02-10 01:45:59 -0800773 * The @cmd value is turned from an encoding command and version into
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 * just the command code.
775 */
776
777int ipc_parse_version (int *cmd)
778{
779 if (*cmd & IPC_64) {
780 *cmd ^= IPC_64;
781 return IPC_64;
782 } else {
783 return IPC_OLD;
784 }
785}
786
787#endif /* __ARCH_WANT_IPC_PARSE_VERSION */
Mike Waychisonae781772005-09-06 15:17:09 -0700788
789#ifdef CONFIG_PROC_FS
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800790struct ipc_proc_iter {
791 struct ipc_namespace *ns;
792 struct ipc_proc_iface *iface;
793};
794
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700795/*
796 * This routine locks the ipc structure found at least at position pos.
797 */
Adrian Bunkb524b9a2008-02-06 01:36:28 -0800798static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
799 loff_t *new_pos)
Mike Waychisonae781772005-09-06 15:17:09 -0700800{
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700801 struct kern_ipc_perm *ipc;
802 int total, id;
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700803
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700804 total = 0;
805 for (id = 0; id < pos && total < ids->in_use; id++) {
806 ipc = idr_find(&ids->ipcs_idr, id);
807 if (ipc != NULL)
808 total++;
809 }
Mike Waychisonae781772005-09-06 15:17:09 -0700810
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700811 if (total >= ids->in_use)
812 return NULL;
Mike Waychisonae781772005-09-06 15:17:09 -0700813
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700814 for ( ; pos < IPCMNI; pos++) {
815 ipc = idr_find(&ids->ipcs_idr, pos);
816 if (ipc != NULL) {
817 *new_pos = pos + 1;
818 ipc_lock_by_ptr(ipc);
Mike Waychisonae781772005-09-06 15:17:09 -0700819 return ipc;
820 }
821 }
822
823 /* Out of range - return NULL to terminate iteration */
824 return NULL;
825}
826
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700827static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
828{
829 struct ipc_proc_iter *iter = s->private;
830 struct ipc_proc_iface *iface = iter->iface;
831 struct kern_ipc_perm *ipc = it;
832
833 /* If we had an ipc id locked before, unlock it */
834 if (ipc && ipc != SEQ_START_TOKEN)
835 ipc_unlock(ipc);
836
Pierre Peiffered2ddbf2008-02-08 04:18:57 -0800837 return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700838}
839
Mike Waychisonae781772005-09-06 15:17:09 -0700840/*
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700841 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
842 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
Mike Waychisonae781772005-09-06 15:17:09 -0700843 */
844static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
845{
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800846 struct ipc_proc_iter *iter = s->private;
847 struct ipc_proc_iface *iface = iter->iface;
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700848 struct ipc_ids *ids;
849
Pierre Peiffered2ddbf2008-02-08 04:18:57 -0800850 ids = &iter->ns->ids[iface->ids];
Mike Waychisonae781772005-09-06 15:17:09 -0700851
852 /*
853 * Take the lock - this will be released by the corresponding
854 * call to stop().
855 */
Nadia Derbey3e148c72007-10-18 23:40:54 -0700856 down_read(&ids->rw_mutex);
Mike Waychisonae781772005-09-06 15:17:09 -0700857
858 /* pos < 0 is invalid */
859 if (*pos < 0)
860 return NULL;
861
862 /* pos == 0 means header */
863 if (*pos == 0)
864 return SEQ_START_TOKEN;
865
866 /* Find the (pos-1)th ipc */
Nadia Derbey7ca7e562007-10-18 23:40:48 -0700867 return sysvipc_find_ipc(ids, *pos - 1, pos);
Mike Waychisonae781772005-09-06 15:17:09 -0700868}
869
870static void sysvipc_proc_stop(struct seq_file *s, void *it)
871{
872 struct kern_ipc_perm *ipc = it;
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800873 struct ipc_proc_iter *iter = s->private;
874 struct ipc_proc_iface *iface = iter->iface;
Kirill Korotaev73ea4132006-10-02 02:18:20 -0700875 struct ipc_ids *ids;
Mike Waychisonae781772005-09-06 15:17:09 -0700876
Nadia Derbeyf4566f02007-10-18 23:40:53 -0700877 /* If we had a locked structure, release it */
Mike Waychisonae781772005-09-06 15:17:09 -0700878 if (ipc && ipc != SEQ_START_TOKEN)
879 ipc_unlock(ipc);
880
Pierre Peiffered2ddbf2008-02-08 04:18:57 -0800881 ids = &iter->ns->ids[iface->ids];
Mike Waychisonae781772005-09-06 15:17:09 -0700882 /* Release the lock we took in start() */
Nadia Derbey3e148c72007-10-18 23:40:54 -0700883 up_read(&ids->rw_mutex);
Mike Waychisonae781772005-09-06 15:17:09 -0700884}
885
886static int sysvipc_proc_show(struct seq_file *s, void *it)
887{
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800888 struct ipc_proc_iter *iter = s->private;
889 struct ipc_proc_iface *iface = iter->iface;
Mike Waychisonae781772005-09-06 15:17:09 -0700890
891 if (it == SEQ_START_TOKEN)
892 return seq_puts(s, iface->header);
893
894 return iface->show(s, it);
895}
896
897static struct seq_operations sysvipc_proc_seqops = {
898 .start = sysvipc_proc_start,
899 .stop = sysvipc_proc_stop,
900 .next = sysvipc_proc_next,
901 .show = sysvipc_proc_show,
902};
903
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800904static int sysvipc_proc_open(struct inode *inode, struct file *file)
905{
Mike Waychisonae781772005-09-06 15:17:09 -0700906 int ret;
907 struct seq_file *seq;
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800908 struct ipc_proc_iter *iter;
909
910 ret = -ENOMEM;
911 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
912 if (!iter)
913 goto out;
Mike Waychisonae781772005-09-06 15:17:09 -0700914
915 ret = seq_open(file, &sysvipc_proc_seqops);
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800916 if (ret)
917 goto out_kfree;
918
919 seq = file->private_data;
920 seq->private = iter;
921
922 iter->iface = PDE(inode)->data;
923 iter->ns = get_ipc_ns(current->nsproxy->ipc_ns);
924out:
Mike Waychisonae781772005-09-06 15:17:09 -0700925 return ret;
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800926out_kfree:
927 kfree(iter);
928 goto out;
929}
930
931static int sysvipc_proc_release(struct inode *inode, struct file *file)
932{
933 struct seq_file *seq = file->private_data;
934 struct ipc_proc_iter *iter = seq->private;
935 put_ipc_ns(iter->ns);
936 return seq_release_private(inode, file);
Mike Waychisonae781772005-09-06 15:17:09 -0700937}
938
Arjan van de Ven9a321442007-02-12 00:55:35 -0800939static const struct file_operations sysvipc_proc_fops = {
Mike Waychisonae781772005-09-06 15:17:09 -0700940 .open = sysvipc_proc_open,
941 .read = seq_read,
942 .llseek = seq_lseek,
Eric W. Biedermanbc1fc6d2007-02-12 00:52:10 -0800943 .release = sysvipc_proc_release,
Mike Waychisonae781772005-09-06 15:17:09 -0700944};
945#endif /* CONFIG_PROC_FS */