util.c 20.9 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9
/*
 * linux/ipc/util.c
 * Copyright (C) 1992 Krishna Balasubramanian
 *
 * Sep 1997 - Call suser() last after "normal" permission checks so we
 *            get BSD style process accounting right.
 *            Occurs in several places in the IPC code.
 *            Chris Evans, <chris@ferret.lmh.ox.ac.uk>
 * Nov 1999 - ipc helper functions, unified SMP locking
10
 *	      Manfred Spraul <manfred@colorfullife.com>
L
Linus Torvalds 已提交
11 12
 * Oct 2002 - One lock per IPC id. RCU ipc_free for lock-free grow_ary().
 *            Mingming Cao <cmm@us.ibm.com>
S
Steve Grubb 已提交
13 14
 * Mar 2006 - support for audit of ipc object properties
 *            Dustin Kirkland <dustin.kirkland@us.ibm.com>
K
Kirill Korotaev 已提交
15 16 17
 * Jun 2006 - namespaces ssupport
 *            OpenVZ, SWsoft Inc.
 *            Pavel Emelianov <xemul@openvz.org>
18 19
 *
 * General sysv ipc locking scheme:
20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
 *	rcu_read_lock()
 *          obtain the ipc object (kern_ipc_perm) by looking up the id in an idr
 *	    tree.
 *	    - perform initial checks (capabilities, auditing and permission,
 *	      etc).
 *	    - perform read-only operations, such as STAT, INFO commands.
 *	      acquire the ipc lock (kern_ipc_perm.lock) through
 *	      ipc_lock_object()
 *		- perform data updates, such as SET, RMID commands and
 *		  mechanism-specific operations (semop/semtimedop,
 *		  msgsnd/msgrcv, shmat/shmdt).
 *	    drop the ipc lock, through ipc_unlock_object().
 *	rcu_read_unlock()
 *
 *  The ids->rwsem must be taken when:
 *	- creating, removing and iterating the existing entries in ipc
 *	  identifier sets.
 *	- iterating through files under /proc/sysvipc/
 *
 *  Note that sems have a special fast path that avoids kern_ipc_perm.lock -
 *  see sem_lock().
L
Linus Torvalds 已提交
41 42 43 44 45 46 47 48
 */

#include <linux/mm.h>
#include <linux/shm.h>
#include <linux/init.h>
#include <linux/msg.h>
#include <linux/vmalloc.h>
#include <linux/slab.h>
49
#include <linux/notifier.h>
50
#include <linux/capability.h>
L
Linus Torvalds 已提交
51 52 53 54
#include <linux/highuid.h>
#include <linux/security.h>
#include <linux/rcupdate.h>
#include <linux/workqueue.h>
55 56
#include <linux/seq_file.h>
#include <linux/proc_fs.h>
S
Steve Grubb 已提交
57
#include <linux/audit.h>
K
Kirill Korotaev 已提交
58
#include <linux/nsproxy.h>
N
Nadia Derbey 已提交
59
#include <linux/rwsem.h>
60
#include <linux/memory.h>
61
#include <linux/ipc_namespace.h>
L
Linus Torvalds 已提交
62 63 64 65 66

#include <asm/unistd.h>

#include "util.h"

67 68 69
struct ipc_proc_iface {
	const char *path;
	const char *header;
K
Kirill Korotaev 已提交
70
	int ids;
71 72 73
	int (*show)(struct seq_file *, void *);
};

L
Linus Torvalds 已提交
74
/**
D
Davidlohr Bueso 已提交
75
 * ipc_init - initialise ipc subsystem
L
Linus Torvalds 已提交
76
 *
D
Davidlohr Bueso 已提交
77 78 79 80 81 82
 * The various sysv ipc resources (semaphores, messages and shared
 * memory) are initialised.
 *
 * A callback routine is registered into the memory hotplug notifier
 * chain: since msgmni scales to lowmem this callback routine will be
 * called upon successful memory add / remove to recompute msmgni.
L
Linus Torvalds 已提交
83 84 85 86 87 88 89 90
 */
static int __init ipc_init(void)
{
	sem_init();
	msg_init();
	shm_init();
	return 0;
}
D
Davidlohr Bueso 已提交
91
device_initcall(ipc_init);
L
Linus Torvalds 已提交
92 93

/**
D
Davidlohr Bueso 已提交
94 95
 * ipc_init_ids	- initialise ipc identifiers
 * @ids: ipc identifier set
L
Linus Torvalds 已提交
96
 *
D
Davidlohr Bueso 已提交
97 98
 * Set up the sequence range to use for the ipc identifier range (limited
 * below IPCMNI) then initialise the ids idr.
L
Linus Torvalds 已提交
99
 */
N
Nadia Derbey 已提交
100
void ipc_init_ids(struct ipc_ids *ids)
L
Linus Torvalds 已提交
101 102 103
{
	ids->in_use = 0;
	ids->seq = 0;
104
	ids->next_id = -1;
105
	init_rwsem(&ids->rwsem);
N
Nadia Derbey 已提交
106
	idr_init(&ids->ipcs_idr);
L
Linus Torvalds 已提交
107 108
}

109
#ifdef CONFIG_PROC_FS
110
static const struct file_operations sysvipc_proc_fops;
111
/**
D
Davidlohr Bueso 已提交
112 113 114 115 116
 * ipc_init_proc_interface -  create a proc interface for sysipc types using a seq_file interface.
 * @path: Path in procfs
 * @header: Banner to be printed at the beginning of the file.
 * @ids: ipc id table to iterate.
 * @show: show routine.
117 118
 */
void __init ipc_init_proc_interface(const char *path, const char *header,
K
Kirill Korotaev 已提交
119
		int ids, int (*show)(struct seq_file *, void *))
120 121 122 123 124 125 126 127 128 129 130 131
{
	struct proc_dir_entry *pde;
	struct ipc_proc_iface *iface;

	iface = kmalloc(sizeof(*iface), GFP_KERNEL);
	if (!iface)
		return;
	iface->path	= path;
	iface->header	= header;
	iface->ids	= ids;
	iface->show	= show;

132 133 134 135 136
	pde = proc_create_data(path,
			       S_IRUGO,        /* world readable */
			       NULL,           /* parent dir */
			       &sysvipc_proc_fops,
			       iface);
137
	if (!pde)
138 139 140 141
		kfree(iface);
}
#endif

L
Linus Torvalds 已提交
142
/**
D
Davidlohr Bueso 已提交
143 144 145
 * ipc_findkey	- find a key in an ipc identifier set
 * @ids: ipc identifier set
 * @key: key to find
P
Paul McQuade 已提交
146
 *
D
Davidlohr Bueso 已提交
147 148 149 150
 * Returns the locked pointer to the ipc structure if found or NULL
 * otherwise. If key is found ipc points to the owning ipc structure
 *
 * Called with ipc_ids.rwsem held.
L
Linus Torvalds 已提交
151
 */
N
Nadia Derbey 已提交
152
static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
L
Linus Torvalds 已提交
153
{
N
Nadia Derbey 已提交
154 155 156
	struct kern_ipc_perm *ipc;
	int next_id;
	int total;
L
Linus Torvalds 已提交
157

N
Nadia Derbey 已提交
158 159 160 161
	for (total = 0, next_id = 0; total < ids->in_use; next_id++) {
		ipc = idr_find(&ids->ipcs_idr, next_id);

		if (ipc == NULL)
L
Linus Torvalds 已提交
162
			continue;
N
Nadia Derbey 已提交
163 164 165 166 167 168

		if (ipc->key != key) {
			total++;
			continue;
		}

D
Davidlohr Bueso 已提交
169 170
		rcu_read_lock();
		ipc_lock_object(ipc);
N
Nadia Derbey 已提交
171
		return ipc;
L
Linus Torvalds 已提交
172
	}
N
Nadia Derbey 已提交
173 174

	return NULL;
L
Linus Torvalds 已提交
175 176
}

N
Nadia Derbey 已提交
177
/**
D
Davidlohr Bueso 已提交
178 179
 * ipc_get_maxid - get the last assigned id
 * @ids: ipc identifier set
N
Nadia Derbey 已提交
180
 *
D
Davidlohr Bueso 已提交
181
 * Called with ipc_ids.rwsem held.
L
Linus Torvalds 已提交
182
 */
N
Nadia Derbey 已提交
183 184 185 186 187 188 189 190
int ipc_get_maxid(struct ipc_ids *ids)
{
	struct kern_ipc_perm *ipc;
	int max_id = -1;
	int total, id;

	if (ids->in_use == 0)
		return -1;
L
Linus Torvalds 已提交
191

N
Nadia Derbey 已提交
192 193 194 195 196 197 198 199 200 201 202 203 204
	if (ids->in_use == IPCMNI)
		return IPCMNI - 1;

	/* Look for the last assigned id */
	total = 0;
	for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
		ipc = idr_find(&ids->ipcs_idr, id);
		if (ipc != NULL) {
			max_id = id;
			total++;
		}
	}
	return max_id;
L
Linus Torvalds 已提交
205 206 207
}

/**
D
Davidlohr Bueso 已提交
208 209 210 211
 * ipc_addid - add an ipc identifier
 * @ids: ipc identifier set
 * @new: new ipc permission set
 * @size: limit for the number of used ids
L
Linus Torvalds 已提交
212
 *
D
Davidlohr Bueso 已提交
213 214 215 216
 * Add an entry 'new' to the ipc ids idr. The permissions object is
 * initialised and the first free entry is set up and the id assigned
 * is returned. The 'new' entry is returned in a locked state on success.
 * On failure the entry is not locked and a negative err-code is returned.
L
Linus Torvalds 已提交
217
 *
D
Davidlohr Bueso 已提交
218
 * Called with writer ipc_ids.rwsem held.
L
Linus Torvalds 已提交
219
 */
M
Manfred Spraul 已提交
220
int ipc_addid(struct ipc_ids *ids, struct kern_ipc_perm *new, int size)
L
Linus Torvalds 已提交
221
{
222 223
	kuid_t euid;
	kgid_t egid;
T
Tejun Heo 已提交
224
	int id;
225
	int next_id = ids->next_id;
L
Linus Torvalds 已提交
226

N
Nadia Derbey 已提交
227 228 229 230
	if (size > IPCMNI)
		size = IPCMNI;

	if (ids->in_use >= size)
231
		return -ENOSPC;
N
Nadia Derbey 已提交
232

T
Tejun Heo 已提交
233 234
	idr_preload(GFP_KERNEL);

235
	spin_lock_init(&new->lock);
236
	new->deleted = false;
237 238 239
	rcu_read_lock();
	spin_lock(&new->lock);

240 241 242 243
	current_euid_egid(&euid, &egid);
	new->cuid = new->uid = euid;
	new->gid = new->cgid = egid;

T
Tejun Heo 已提交
244 245 246 247 248
	id = idr_alloc(&ids->ipcs_idr, new,
		       (next_id < 0) ? 0 : ipcid_to_idx(next_id), 0,
		       GFP_NOWAIT);
	idr_preload_end();
	if (id < 0) {
249 250
		spin_unlock(&new->lock);
		rcu_read_unlock();
T
Tejun Heo 已提交
251
		return id;
252
	}
N
Nadia Derbey 已提交
253

L
Linus Torvalds 已提交
254 255
	ids->in_use++;

256 257
	if (next_id < 0) {
		new->seq = ids->seq++;
258
		if (ids->seq > IPCID_SEQ_MAX)
259 260 261 262 263
			ids->seq = 0;
	} else {
		new->seq = ipcid_to_seqx(next_id);
		ids->next_id = -1;
	}
L
Linus Torvalds 已提交
264

265
	new->id = ipc_buildid(id, new->seq);
L
Linus Torvalds 已提交
266 267 268
	return id;
}

N
Nadia Derbey 已提交
269
/**
D
Davidlohr Bueso 已提交
270 271
 * ipcget_new -	create a new ipc object
 * @ns: ipc namespace
272
 * @ids: ipc identifier set
D
Davidlohr Bueso 已提交
273 274 275 276 277
 * @ops: the actual creation routine to call
 * @params: its parameters
 *
 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
 * when the key is IPC_PRIVATE.
N
Nadia Derbey 已提交
278
 */
279
static int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
M
Mathias Krause 已提交
280
		const struct ipc_ops *ops, struct ipc_params *params)
N
Nadia Derbey 已提交
281 282 283
{
	int err;

D
Davidlohr Bueso 已提交
284
	down_write(&ids->rwsem);
N
Nadia Derbey 已提交
285
	err = ops->getnew(ns, params);
D
Davidlohr Bueso 已提交
286
	up_write(&ids->rwsem);
N
Nadia Derbey 已提交
287 288 289 290
	return err;
}

/**
D
Davidlohr Bueso 已提交
291 292 293 294 295
 * ipc_check_perms - check security and permissions for an ipc object
 * @ns: ipc namespace
 * @ipcp: ipc permission set
 * @ops: the actual security routine to call
 * @params: its parameters
N
Nadia Derbey 已提交
296
 *
D
Davidlohr Bueso 已提交
297 298 299
 * This routine is called by sys_msgget(), sys_semget() and sys_shmget()
 * when the key is not IPC_PRIVATE and that key already exists in the
 * ds IDR.
N
Nadia Derbey 已提交
300
 *
D
Davidlohr Bueso 已提交
301
 * On success, the ipc id is returned.
N
Nadia Derbey 已提交
302
 *
D
Davidlohr Bueso 已提交
303
 * It is called with ipc_ids.rwsem and ipcp->lock held.
N
Nadia Derbey 已提交
304
 */
305 306
static int ipc_check_perms(struct ipc_namespace *ns,
			   struct kern_ipc_perm *ipcp,
M
Mathias Krause 已提交
307
			   const struct ipc_ops *ops,
308
			   struct ipc_params *params)
N
Nadia Derbey 已提交
309 310 311
{
	int err;

312
	if (ipcperms(ns, ipcp, params->flg))
N
Nadia Derbey 已提交
313 314 315 316 317 318 319 320 321 322 323
		err = -EACCES;
	else {
		err = ops->associate(ipcp, params->flg);
		if (!err)
			err = ipcp->id;
	}

	return err;
}

/**
D
Davidlohr Bueso 已提交
324 325
 * ipcget_public - get an ipc object or create a new one
 * @ns: ipc namespace
326
 * @ids: ipc identifier set
D
Davidlohr Bueso 已提交
327 328 329 330 331 332 333 334 335
 * @ops: the actual creation routine to call
 * @params: its parameters
 *
 * This routine is called by sys_msgget, sys_semget() and sys_shmget()
 * when the key is not IPC_PRIVATE.
 * It adds a new entry if the key is not found and does some permission
 * / security checkings if the key is found.
 *
 * On success, the ipc id is returned.
N
Nadia Derbey 已提交
336
 */
337
static int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
M
Mathias Krause 已提交
338
		const struct ipc_ops *ops, struct ipc_params *params)
N
Nadia Derbey 已提交
339 340 341 342 343
{
	struct kern_ipc_perm *ipcp;
	int flg = params->flg;
	int err;

N
Nadia Derbey 已提交
344 345 346 347
	/*
	 * Take the lock as a writer since we are potentially going to add
	 * a new entry + read locks are not "upgradable"
	 */
D
Davidlohr Bueso 已提交
348
	down_write(&ids->rwsem);
N
Nadia Derbey 已提交
349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365
	ipcp = ipc_findkey(ids, params->key);
	if (ipcp == NULL) {
		/* key not used */
		if (!(flg & IPC_CREAT))
			err = -ENOENT;
		else
			err = ops->getnew(ns, params);
	} else {
		/* ipc object has been locked by ipc_findkey() */

		if (flg & IPC_CREAT && flg & IPC_EXCL)
			err = -EEXIST;
		else {
			err = 0;
			if (ops->more_checks)
				err = ops->more_checks(ipcp, params);
			if (!err)
N
Nadia Derbey 已提交
366 367 368 369
				/*
				 * ipc_check_perms returns the IPC id on
				 * success
				 */
370
				err = ipc_check_perms(ns, ipcp, ops, params);
N
Nadia Derbey 已提交
371 372 373
		}
		ipc_unlock(ipcp);
	}
D
Davidlohr Bueso 已提交
374
	up_write(&ids->rwsem);
N
Nadia Derbey 已提交
375 376 377 378 379

	return err;
}


L
Linus Torvalds 已提交
380
/**
D
Davidlohr Bueso 已提交
381 382 383
 * ipc_rmid - remove an ipc identifier
 * @ids: ipc identifier set
 * @ipcp: ipc perm structure containing the identifier to remove
L
Linus Torvalds 已提交
384
 *
D
Davidlohr Bueso 已提交
385 386
 * ipc_ids.rwsem (as a writer) and the spinlock for this ID are held
 * before this function is called, and remain locked on the exit.
L
Linus Torvalds 已提交
387
 */
N
Nadia Derbey 已提交
388
void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
L
Linus Torvalds 已提交
389
{
390
	int lid = ipcid_to_idx(ipcp->id);
N
Nadia Derbey 已提交
391 392

	idr_remove(&ids->ipcs_idr, lid);
L
Linus Torvalds 已提交
393
	ids->in_use--;
394
	ipcp->deleted = true;
L
Linus Torvalds 已提交
395 396 397
}

/**
D
Davidlohr Bueso 已提交
398 399
 * ipc_alloc -	allocate ipc space
 * @size: size desired
L
Linus Torvalds 已提交
400
 *
D
Davidlohr Bueso 已提交
401 402
 * Allocate memory from the appropriate pools and return a pointer to it.
 * NULL is returned if the allocation fails
L
Linus Torvalds 已提交
403
 */
404
void *ipc_alloc(int size)
L
Linus Torvalds 已提交
405
{
M
Michal Hocko 已提交
406
	return kvmalloc(size, GFP_KERNEL);
L
Linus Torvalds 已提交
407 408 409
}

/**
D
Davidlohr Bueso 已提交
410 411
 * ipc_free - free ipc space
 * @ptr: pointer returned by ipc_alloc
L
Linus Torvalds 已提交
412
 *
413
 * Free a block created with ipc_alloc().
L
Linus Torvalds 已提交
414
 */
415
void ipc_free(void *ptr)
L
Linus Torvalds 已提交
416
{
417
	kvfree(ptr);
L
Linus Torvalds 已提交
418 419 420
}

/**
421
 * ipc_rcu_alloc - allocate ipc space
D
Davidlohr Bueso 已提交
422
 * @size: size desired
L
Linus Torvalds 已提交
423
 *
424 425
 * Allocate memory for an ipc object.
 * The first member must be struct kern_ipc_perm.
L
Linus Torvalds 已提交
426
 */
427
struct kern_ipc_perm *ipc_rcu_alloc(int size)
L
Linus Torvalds 已提交
428
{
429
	/*
430
	 * We prepend the allocation with the rcu struct
L
Linus Torvalds 已提交
431
	 */
432
	struct kern_ipc_perm *out = ipc_alloc(size);
433 434
	if (unlikely(!out))
		return NULL;
435 436

	memset(out, 0, size);
437
	atomic_set(&out->refcount, 1);
438
	return out;
L
Linus Torvalds 已提交
439 440
}

441
int ipc_rcu_getref(struct kern_ipc_perm *ptr)
L
Linus Torvalds 已提交
442
{
443
	return atomic_inc_not_zero(&ptr->refcount);
444 445
}

446 447
void ipc_rcu_putref(struct kern_ipc_perm *ptr,
			void (*func)(struct rcu_head *head))
L
Linus Torvalds 已提交
448
{
449
	if (!atomic_dec_and_test(&ptr->refcount))
L
Linus Torvalds 已提交
450 451
		return;

452
	call_rcu(&ptr->rcu, func);
D
Davidlohr Bueso 已提交
453 454
}

455
void ipc_rcu_free(struct rcu_head *h)
D
Davidlohr Bueso 已提交
456
{
457
	struct kern_ipc_perm *ptr = container_of(h, struct kern_ipc_perm, rcu);
D
Davidlohr Bueso 已提交
458

459
	kvfree(ptr);
L
Linus Torvalds 已提交
460 461 462
}

/**
D
Davidlohr Bueso 已提交
463 464 465 466
 * ipcperms - check ipc permissions
 * @ns: ipc namespace
 * @ipcp: ipc permission set
 * @flag: desired permission set
L
Linus Torvalds 已提交
467
 *
D
Davidlohr Bueso 已提交
468 469
 * Check user, group, other permissions for access
 * to ipc resources. return 0 if allowed
470
 *
471
 * @flag will most probably be 0 or ``S_...UGO`` from <linux/stat.h>
L
Linus Torvalds 已提交
472
 */
473 474
int ipcperms(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp, short flag)
{
475
	kuid_t euid = current_euid();
A
Al Viro 已提交
476
	int requested_mode, granted_mode;
L
Linus Torvalds 已提交
477

A
Al Viro 已提交
478
	audit_ipc_obj(ipcp);
L
Linus Torvalds 已提交
479 480
	requested_mode = (flag >> 6) | (flag >> 3) | flag;
	granted_mode = ipcp->mode;
481 482
	if (uid_eq(euid, ipcp->cuid) ||
	    uid_eq(euid, ipcp->uid))
L
Linus Torvalds 已提交
483 484 485 486
		granted_mode >>= 6;
	else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
		granted_mode >>= 3;
	/* is there some bit set in requested_mode but not in granted_mode? */
P
Paul McQuade 已提交
487
	if ((requested_mode & ~granted_mode & 0007) &&
488
	    !ns_capable(ns->user_ns, CAP_IPC_OWNER))
L
Linus Torvalds 已提交
489 490 491 492 493 494 495 496 497 498 499
		return -1;

	return security_ipc_permission(ipcp, flag);
}

/*
 * Functions to convert between the kern_ipc_perm structure and the
 * old/new ipc_perm structures
 */

/**
D
Davidlohr Bueso 已提交
500 501 502
 * kernel_to_ipc64_perm	- convert kernel ipc permissions to user
 * @in: kernel permissions
 * @out: new style ipc permissions
L
Linus Torvalds 已提交
503
 *
D
Davidlohr Bueso 已提交
504 505
 * Turn the kernel object @in into a set of permissions descriptions
 * for returning to userspace (@out).
L
Linus Torvalds 已提交
506
 */
M
Manfred Spraul 已提交
507
void kernel_to_ipc64_perm(struct kern_ipc_perm *in, struct ipc64_perm *out)
L
Linus Torvalds 已提交
508 509
{
	out->key	= in->key;
510 511 512 513
	out->uid	= from_kuid_munged(current_user_ns(), in->uid);
	out->gid	= from_kgid_munged(current_user_ns(), in->gid);
	out->cuid	= from_kuid_munged(current_user_ns(), in->cuid);
	out->cgid	= from_kgid_munged(current_user_ns(), in->cgid);
L
Linus Torvalds 已提交
514 515 516 517 518
	out->mode	= in->mode;
	out->seq	= in->seq;
}

/**
D
Davidlohr Bueso 已提交
519 520 521
 * ipc64_perm_to_ipc_perm - convert new ipc permissions to old
 * @in: new style ipc permissions
 * @out: old style ipc permissions
L
Linus Torvalds 已提交
522
 *
D
Davidlohr Bueso 已提交
523 524
 * Turn the new style permissions object @in into a compatibility
 * object and store it into the @out pointer.
L
Linus Torvalds 已提交
525
 */
M
Manfred Spraul 已提交
526
void ipc64_perm_to_ipc_perm(struct ipc64_perm *in, struct ipc_perm *out)
L
Linus Torvalds 已提交
527 528 529 530 531 532 533 534 535 536
{
	out->key	= in->key;
	SET_UID(out->uid, in->uid);
	SET_GID(out->gid, in->gid);
	SET_UID(out->cuid, in->cuid);
	SET_GID(out->cgid, in->cgid);
	out->mode	= in->mode;
	out->seq	= in->seq;
}

537 538 539 540 541 542 543 544 545 546
/**
 * ipc_obtain_object
 * @ids: ipc identifier set
 * @id: ipc id to look for
 *
 * Look for an id in the ipc ids idr and return associated ipc object.
 *
 * Call inside the RCU critical section.
 * The ipc object is *not* locked on exit.
 */
D
Davidlohr Bueso 已提交
547
struct kern_ipc_perm *ipc_obtain_object_idr(struct ipc_ids *ids, int id)
548 549 550 551 552 553 554 555 556 557 558
{
	struct kern_ipc_perm *out;
	int lid = ipcid_to_idx(id);

	out = idr_find(&ids->ipcs_idr, lid);
	if (!out)
		return ERR_PTR(-EINVAL);

	return out;
}

N
Nadia Derbey 已提交
559
/**
D
Davidlohr Bueso 已提交
560 561
 * ipc_lock - lock an ipc structure without rwsem held
 * @ids: ipc identifier set
N
Nadia Derbey 已提交
562 563 564 565
 * @id: ipc id to look for
 *
 * Look for an id in the ipc ids idr and lock the associated ipc object.
 *
566
 * The ipc object is locked on successful exit.
N
Nadia Derbey 已提交
567
 */
N
Nadia Derbey 已提交
568
struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
L
Linus Torvalds 已提交
569
{
N
Nadia Derbey 已提交
570
	struct kern_ipc_perm *out;
L
Linus Torvalds 已提交
571 572

	rcu_read_lock();
D
Davidlohr Bueso 已提交
573
	out = ipc_obtain_object_idr(ids, id);
574
	if (IS_ERR(out))
575
		goto err;
N
Nadia Derbey 已提交
576

L
Linus Torvalds 已提交
577
	spin_lock(&out->lock);
578

579 580 581 582 583
	/*
	 * ipc_rmid() may have already freed the ID while ipc_lock()
	 * was spinning: here verify that the structure is still valid.
	 * Upon races with RMID, return -EIDRM, thus indicating that
	 * the ID points to a removed identifier.
L
Linus Torvalds 已提交
584
	 */
585
	if (ipc_valid_object(out))
586 587 588
		return out;

	spin_unlock(&out->lock);
589 590
	out = ERR_PTR(-EIDRM);
err:
591 592 593
	rcu_read_unlock();
	return out;
}
N
Nadia Derbey 已提交
594

595 596 597 598 599
/**
 * ipc_obtain_object_check
 * @ids: ipc identifier set
 * @id: ipc id to look for
 *
D
Davidlohr Bueso 已提交
600
 * Similar to ipc_obtain_object_idr() but also checks
601 602 603 604 605 606 607
 * the ipc object reference counter.
 *
 * Call inside the RCU critical section.
 * The ipc object is *not* locked on exit.
 */
struct kern_ipc_perm *ipc_obtain_object_check(struct ipc_ids *ids, int id)
{
D
Davidlohr Bueso 已提交
608
	struct kern_ipc_perm *out = ipc_obtain_object_idr(ids, id);
609 610 611 612 613

	if (IS_ERR(out))
		goto out;

	if (ipc_checkid(out, id))
614
		return ERR_PTR(-EINVAL);
615
out:
L
Linus Torvalds 已提交
616 617 618
	return out;
}

619 620
/**
 * ipcget - Common sys_*get() code
621
 * @ns: namespace
D
Davidlohr Bueso 已提交
622 623 624 625
 * @ids: ipc identifier set
 * @ops: operations to be called on ipc object creation, permission checks
 *       and further checks
 * @params: the parameters needed by the previous operations.
626 627 628 629
 *
 * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
 */
int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
M
Mathias Krause 已提交
630
			const struct ipc_ops *ops, struct ipc_params *params)
631 632 633 634 635 636 637
{
	if (params->key == IPC_PRIVATE)
		return ipcget_new(ns, ids, ops, params);
	else
		return ipcget_public(ns, ids, ops, params);
}

638
/**
D
Davidlohr Bueso 已提交
639
 * ipc_update_perm - update the permissions of an ipc object
640 641 642
 * @in:  the permission given as input.
 * @out: the permission of the ipc to set.
 */
643
int ipc_update_perm(struct ipc64_perm *in, struct kern_ipc_perm *out)
644
{
645 646 647 648 649 650 651
	kuid_t uid = make_kuid(current_user_ns(), in->uid);
	kgid_t gid = make_kgid(current_user_ns(), in->gid);
	if (!uid_valid(uid) || !gid_valid(gid))
		return -EINVAL;

	out->uid = uid;
	out->gid = gid;
652 653
	out->mode = (out->mode & ~S_IRWXUGO)
		| (in->mode & S_IRWXUGO);
654 655

	return 0;
656 657
}

658
/**
D
Davidlohr Bueso 已提交
659
 * ipcctl_pre_down_nolock - retrieve an ipc and check permissions for some IPC_XXX cmd
D
Davidlohr Bueso 已提交
660
 * @ns:  ipc namespace
661 662 663 664 665 666 667 668
 * @ids:  the table of ids where to look for the ipc
 * @id:   the id of the ipc to retrieve
 * @cmd:  the cmd to check
 * @perm: the permission to set
 * @extra_perm: one extra permission parameter used by msq
 *
 * This function does some common audit and permissions check for some IPC_XXX
 * cmd and is called from semctl_down, shmctl_down and msgctl_down.
669 670 671 672 673 674
 * It must be called without any lock held and:
 *
 *   - retrieves the ipc with the given id in the given table.
 *   - performs some audit and permission check, depending on the given cmd
 *   - returns a pointer to the ipc object or otherwise, the corresponding
 *     error.
675
 *
D
Davidlohr Bueso 已提交
676
 * Call holding the both the rwsem and the rcu read lock.
677
 */
678
struct kern_ipc_perm *ipcctl_pre_down_nolock(struct ipc_namespace *ns,
D
Davidlohr Bueso 已提交
679 680
					struct ipc_ids *ids, int id, int cmd,
					struct ipc64_perm *perm, int extra_perm)
681
{
682
	kuid_t euid;
683 684
	int err = -EPERM;
	struct kern_ipc_perm *ipcp;
685

686
	ipcp = ipc_obtain_object_check(ids, id);
687 688
	if (IS_ERR(ipcp)) {
		err = PTR_ERR(ipcp);
689
		goto err;
690 691
	}

A
Al Viro 已提交
692
	audit_ipc_obj(ipcp);
A
Al Viro 已提交
693 694
	if (cmd == IPC_SET)
		audit_ipc_set_perm(extra_perm, perm->uid,
695
				   perm->gid, perm->mode);
696 697

	euid = current_euid();
698
	if (uid_eq(euid, ipcp->cuid) || uid_eq(euid, ipcp->uid)  ||
699
	    ns_capable(ns->user_ns, CAP_SYS_ADMIN))
700 701
		return ipcp; /* successful lookup */
err:
702 703 704
	return ERR_PTR(err);
}

705
#ifdef CONFIG_ARCH_WANT_IPC_PARSE_VERSION
L
Linus Torvalds 已提交
706 707 708


/**
D
Davidlohr Bueso 已提交
709 710
 * ipc_parse_version - ipc call version
 * @cmd: pointer to command
L
Linus Torvalds 已提交
711
 *
D
Davidlohr Bueso 已提交
712 713 714
 * Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
 * The @cmd value is turned from an encoding command and version into
 * just the command code.
L
Linus Torvalds 已提交
715
 */
M
Manfred Spraul 已提交
716
int ipc_parse_version(int *cmd)
L
Linus Torvalds 已提交
717 718 719 720 721 722 723 724 725
{
	if (*cmd & IPC_64) {
		*cmd ^= IPC_64;
		return IPC_64;
	} else {
		return IPC_OLD;
	}
}

726
#endif /* CONFIG_ARCH_WANT_IPC_PARSE_VERSION */
727 728

#ifdef CONFIG_PROC_FS
729 730 731 732 733
struct ipc_proc_iter {
	struct ipc_namespace *ns;
	struct ipc_proc_iface *iface;
};

N
Nadia Derbey 已提交
734 735 736
/*
 * This routine locks the ipc structure found at least at position pos.
 */
737 738
static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
					      loff_t *new_pos)
739
{
N
Nadia Derbey 已提交
740 741
	struct kern_ipc_perm *ipc;
	int total, id;
K
Kirill Korotaev 已提交
742

N
Nadia Derbey 已提交
743 744 745 746 747 748
	total = 0;
	for (id = 0; id < pos && total < ids->in_use; id++) {
		ipc = idr_find(&ids->ipcs_idr, id);
		if (ipc != NULL)
			total++;
	}
749

N
Nadia Derbey 已提交
750 751
	if (total >= ids->in_use)
		return NULL;
752

M
Manfred Spraul 已提交
753
	for (; pos < IPCMNI; pos++) {
N
Nadia Derbey 已提交
754 755 756
		ipc = idr_find(&ids->ipcs_idr, pos);
		if (ipc != NULL) {
			*new_pos = pos + 1;
D
Davidlohr Bueso 已提交
757 758
			rcu_read_lock();
			ipc_lock_object(ipc);
759 760 761 762 763 764 765 766
			return ipc;
		}
	}

	/* Out of range - return NULL to terminate iteration */
	return NULL;
}

N
Nadia Derbey 已提交
767 768 769 770 771 772 773 774 775 776
static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
{
	struct ipc_proc_iter *iter = s->private;
	struct ipc_proc_iface *iface = iter->iface;
	struct kern_ipc_perm *ipc = it;

	/* If we had an ipc id locked before, unlock it */
	if (ipc && ipc != SEQ_START_TOKEN)
		ipc_unlock(ipc);

777
	return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
N
Nadia Derbey 已提交
778 779
}

780
/*
N
Nadia Derbey 已提交
781 782
 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
783 784 785
 */
static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
{
786 787
	struct ipc_proc_iter *iter = s->private;
	struct ipc_proc_iface *iface = iter->iface;
K
Kirill Korotaev 已提交
788 789
	struct ipc_ids *ids;

790
	ids = &iter->ns->ids[iface->ids];
791 792 793 794 795

	/*
	 * Take the lock - this will be released by the corresponding
	 * call to stop().
	 */
D
Davidlohr Bueso 已提交
796
	down_read(&ids->rwsem);
797 798 799 800 801 802 803 804 805 806

	/* pos < 0 is invalid */
	if (*pos < 0)
		return NULL;

	/* pos == 0 means header */
	if (*pos == 0)
		return SEQ_START_TOKEN;

	/* Find the (pos-1)th ipc */
N
Nadia Derbey 已提交
807
	return sysvipc_find_ipc(ids, *pos - 1, pos);
808 809 810 811 812
}

static void sysvipc_proc_stop(struct seq_file *s, void *it)
{
	struct kern_ipc_perm *ipc = it;
813 814
	struct ipc_proc_iter *iter = s->private;
	struct ipc_proc_iface *iface = iter->iface;
K
Kirill Korotaev 已提交
815
	struct ipc_ids *ids;
816

N
Nadia Derbey 已提交
817
	/* If we had a locked structure, release it */
818 819 820
	if (ipc && ipc != SEQ_START_TOKEN)
		ipc_unlock(ipc);

821
	ids = &iter->ns->ids[iface->ids];
822
	/* Release the lock we took in start() */
D
Davidlohr Bueso 已提交
823
	up_read(&ids->rwsem);
824 825 826 827
}

static int sysvipc_proc_show(struct seq_file *s, void *it)
{
828 829
	struct ipc_proc_iter *iter = s->private;
	struct ipc_proc_iface *iface = iter->iface;
830

831 832 833 834
	if (it == SEQ_START_TOKEN) {
		seq_puts(s, iface->header);
		return 0;
	}
835 836 837 838

	return iface->show(s, it);
}

J
James Morris 已提交
839
static const struct seq_operations sysvipc_proc_seqops = {
840 841 842 843 844 845
	.start = sysvipc_proc_start,
	.stop  = sysvipc_proc_stop,
	.next  = sysvipc_proc_next,
	.show  = sysvipc_proc_show,
};

846 847 848 849
static int sysvipc_proc_open(struct inode *inode, struct file *file)
{
	struct ipc_proc_iter *iter;

850
	iter = __seq_open_private(file, &sysvipc_proc_seqops, sizeof(*iter));
851
	if (!iter)
852
		return -ENOMEM;
853

A
Al Viro 已提交
854
	iter->iface = PDE_DATA(inode);
855
	iter->ns    = get_ipc_ns(current->nsproxy->ipc_ns);
856 857

	return 0;
858 859 860 861 862 863 864 865
}

static int sysvipc_proc_release(struct inode *inode, struct file *file)
{
	struct seq_file *seq = file->private_data;
	struct ipc_proc_iter *iter = seq->private;
	put_ipc_ns(iter->ns);
	return seq_release_private(inode, file);
866 867
}

868
static const struct file_operations sysvipc_proc_fops = {
869 870 871
	.open    = sysvipc_proc_open,
	.read    = seq_read,
	.llseek  = seq_lseek,
872
	.release = sysvipc_proc_release,
873 874
};
#endif /* CONFIG_PROC_FS */