util.c revision 6062a8dc
1/*
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
10 *	      Manfred Spraul <manfred@colorfullife.com>
11 * Oct 2002 - One lock per IPC id. RCU ipc_free for lock-free grow_ary().
12 *            Mingming Cao <cmm@us.ibm.com>
13 * Mar 2006 - support for audit of ipc object properties
14 *            Dustin Kirkland <dustin.kirkland@us.ibm.com>
15 * Jun 2006 - namespaces ssupport
16 *            OpenVZ, SWsoft Inc.
17 *            Pavel Emelianov <xemul@openvz.org>
18 */
19
20#include <linux/mm.h>
21#include <linux/shm.h>
22#include <linux/init.h>
23#include <linux/msg.h>
24#include <linux/vmalloc.h>
25#include <linux/slab.h>
26#include <linux/notifier.h>
27#include <linux/capability.h>
28#include <linux/highuid.h>
29#include <linux/security.h>
30#include <linux/rcupdate.h>
31#include <linux/workqueue.h>
32#include <linux/seq_file.h>
33#include <linux/proc_fs.h>
34#include <linux/audit.h>
35#include <linux/nsproxy.h>
36#include <linux/rwsem.h>
37#include <linux/memory.h>
38#include <linux/ipc_namespace.h>
39
40#include <asm/unistd.h>
41
42#include "util.h"
43
44struct ipc_proc_iface {
45	const char *path;
46	const char *header;
47	int ids;
48	int (*show)(struct seq_file *, void *);
49};
50
51static void ipc_memory_notifier(struct work_struct *work)
52{
53	ipcns_notify(IPCNS_MEMCHANGED);
54}
55
56static int ipc_memory_callback(struct notifier_block *self,
57				unsigned long action, void *arg)
58{
59	static DECLARE_WORK(ipc_memory_wq, ipc_memory_notifier);
60
61	switch (action) {
62	case MEM_ONLINE:    /* memory successfully brought online */
63	case MEM_OFFLINE:   /* or offline: it's time to recompute msgmni */
64		/*
65		 * This is done by invoking the ipcns notifier chain with the
66		 * IPC_MEMCHANGED event.
67		 * In order not to keep the lock on the hotplug memory chain
68		 * for too long, queue a work item that will, when waken up,
69		 * activate the ipcns notification chain.
70		 * No need to keep several ipc work items on the queue.
71		 */
72		if (!work_pending(&ipc_memory_wq))
73			schedule_work(&ipc_memory_wq);
74		break;
75	case MEM_GOING_ONLINE:
76	case MEM_GOING_OFFLINE:
77	case MEM_CANCEL_ONLINE:
78	case MEM_CANCEL_OFFLINE:
79	default:
80		break;
81	}
82
83	return NOTIFY_OK;
84}
85
86static struct notifier_block ipc_memory_nb = {
87	.notifier_call = ipc_memory_callback,
88	.priority = IPC_CALLBACK_PRI,
89};
90
91/**
92 *	ipc_init	-	initialise IPC subsystem
93 *
94 *	The various system5 IPC resources (semaphores, messages and shared
95 *	memory) are initialised
96 *	A callback routine is registered into the memory hotplug notifier
97 *	chain: since msgmni scales to lowmem this callback routine will be
98 *	called upon successful memory add / remove to recompute msmgni.
99 */
100
101static int __init ipc_init(void)
102{
103	sem_init();
104	msg_init();
105	shm_init();
106	register_hotmemory_notifier(&ipc_memory_nb);
107	register_ipcns_notifier(&init_ipc_ns);
108	return 0;
109}
110__initcall(ipc_init);
111
112/**
113 *	ipc_init_ids		-	initialise IPC identifiers
114 *	@ids: Identifier set
115 *
116 *	Set up the sequence range to use for the ipc identifier range (limited
117 *	below IPCMNI) then initialise the ids idr.
118 */
119
120void ipc_init_ids(struct ipc_ids *ids)
121{
122	init_rwsem(&ids->rw_mutex);
123
124	ids->in_use = 0;
125	ids->seq = 0;
126	ids->next_id = -1;
127	{
128		int seq_limit = INT_MAX/SEQ_MULTIPLIER;
129		if (seq_limit > USHRT_MAX)
130			ids->seq_max = USHRT_MAX;
131		 else
132		 	ids->seq_max = seq_limit;
133	}
134
135	idr_init(&ids->ipcs_idr);
136}
137
138#ifdef CONFIG_PROC_FS
139static const struct file_operations sysvipc_proc_fops;
140/**
141 *	ipc_init_proc_interface	-  Create a proc interface for sysipc types using a seq_file interface.
142 *	@path: Path in procfs
143 *	@header: Banner to be printed at the beginning of the file.
144 *	@ids: ipc id table to iterate.
145 *	@show: show routine.
146 */
147void __init ipc_init_proc_interface(const char *path, const char *header,
148		int ids, int (*show)(struct seq_file *, void *))
149{
150	struct proc_dir_entry *pde;
151	struct ipc_proc_iface *iface;
152
153	iface = kmalloc(sizeof(*iface), GFP_KERNEL);
154	if (!iface)
155		return;
156	iface->path	= path;
157	iface->header	= header;
158	iface->ids	= ids;
159	iface->show	= show;
160
161	pde = proc_create_data(path,
162			       S_IRUGO,        /* world readable */
163			       NULL,           /* parent dir */
164			       &sysvipc_proc_fops,
165			       iface);
166	if (!pde) {
167		kfree(iface);
168	}
169}
170#endif
171
172/**
173 *	ipc_findkey	-	find a key in an ipc identifier set
174 *	@ids: Identifier set
175 *	@key: The key to find
176 *
177 *	Requires ipc_ids.rw_mutex locked.
178 *	Returns the LOCKED pointer to the ipc structure if found or NULL
179 *	if not.
180 *	If key is found ipc points to the owning ipc structure
181 */
182
183static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
184{
185	struct kern_ipc_perm *ipc;
186	int next_id;
187	int total;
188
189	for (total = 0, next_id = 0; total < ids->in_use; next_id++) {
190		ipc = idr_find(&ids->ipcs_idr, next_id);
191
192		if (ipc == NULL)
193			continue;
194
195		if (ipc->key != key) {
196			total++;
197			continue;
198		}
199
200		ipc_lock_by_ptr(ipc);
201		return ipc;
202	}
203
204	return NULL;
205}
206
207/**
208 *	ipc_get_maxid 	-	get the last assigned id
209 *	@ids: IPC identifier set
210 *
211 *	Called with ipc_ids.rw_mutex held.
212 */
213
214int ipc_get_maxid(struct ipc_ids *ids)
215{
216	struct kern_ipc_perm *ipc;
217	int max_id = -1;
218	int total, id;
219
220	if (ids->in_use == 0)
221		return -1;
222
223	if (ids->in_use == IPCMNI)
224		return IPCMNI - 1;
225
226	/* Look for the last assigned id */
227	total = 0;
228	for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
229		ipc = idr_find(&ids->ipcs_idr, id);
230		if (ipc != NULL) {
231			max_id = id;
232			total++;
233		}
234	}
235	return max_id;
236}
237
238/**
239 *	ipc_addid 	-	add an IPC identifier
240 *	@ids: IPC identifier set
241 *	@new: new IPC permission set
242 *	@size: limit for the number of used ids
243 *
244 *	Add an entry 'new' to the IPC ids idr. The permissions object is
245 *	initialised and the first free entry is set up and the id assigned
246 *	is returned. The 'new' entry is returned in a locked state on success.
247 *	On failure the entry is not locked and a negative err-code is returned.
248 *
249 *	Called with ipc_ids.rw_mutex held as a writer.
250 */
251
252int ipc_addid(struct ipc_ids* ids, struct kern_ipc_perm* new, int size)
253{
254	kuid_t euid;
255	kgid_t egid;
256	int id;
257	int next_id = ids->next_id;
258
259	if (size > IPCMNI)
260		size = IPCMNI;
261
262	if (ids->in_use >= size)
263		return -ENOSPC;
264
265	idr_preload(GFP_KERNEL);
266
267	spin_lock_init(&new->lock);
268	new->deleted = 0;
269	rcu_read_lock();
270	spin_lock(&new->lock);
271
272	id = idr_alloc(&ids->ipcs_idr, new,
273		       (next_id < 0) ? 0 : ipcid_to_idx(next_id), 0,
274		       GFP_NOWAIT);
275	idr_preload_end();
276	if (id < 0) {
277		spin_unlock(&new->lock);
278		rcu_read_unlock();
279		return id;
280	}
281
282	ids->in_use++;
283
284	current_euid_egid(&euid, &egid);
285	new->cuid = new->uid = euid;
286	new->gid = new->cgid = egid;
287
288	if (next_id < 0) {
289		new->seq = ids->seq++;
290		if (ids->seq > ids->seq_max)
291			ids->seq = 0;
292	} else {
293		new->seq = ipcid_to_seqx(next_id);
294		ids->next_id = -1;
295	}
296
297	new->id = ipc_buildid(id, new->seq);
298	return id;
299}
300
301/**
302 *	ipcget_new	-	create a new ipc object
303 *	@ns: namespace
304 *	@ids: IPC identifer set
305 *	@ops: the actual creation routine to call
306 *	@params: its parameters
307 *
308 *	This routine is called by sys_msgget, sys_semget() and sys_shmget()
309 *	when the key is IPC_PRIVATE.
310 */
311static int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
312		struct ipc_ops *ops, struct ipc_params *params)
313{
314	int err;
315
316	down_write(&ids->rw_mutex);
317	err = ops->getnew(ns, params);
318	up_write(&ids->rw_mutex);
319	return err;
320}
321
322/**
323 *	ipc_check_perms	-	check security and permissions for an IPC
324 *	@ns: IPC namespace
325 *	@ipcp: ipc permission set
326 *	@ops: the actual security routine to call
327 *	@params: its parameters
328 *
329 *	This routine is called by sys_msgget(), sys_semget() and sys_shmget()
330 *      when the key is not IPC_PRIVATE and that key already exists in the
331 *      ids IDR.
332 *
333 *	On success, the IPC id is returned.
334 *
335 *	It is called with ipc_ids.rw_mutex and ipcp->lock held.
336 */
337static int ipc_check_perms(struct ipc_namespace *ns,
338			   struct kern_ipc_perm *ipcp,
339			   struct ipc_ops *ops,
340			   struct ipc_params *params)
341{
342	int err;
343
344	if (ipcperms(ns, ipcp, params->flg))
345		err = -EACCES;
346	else {
347		err = ops->associate(ipcp, params->flg);
348		if (!err)
349			err = ipcp->id;
350	}
351
352	return err;
353}
354
355/**
356 *	ipcget_public	-	get an ipc object or create a new one
357 *	@ns: namespace
358 *	@ids: IPC identifer set
359 *	@ops: the actual creation routine to call
360 *	@params: its parameters
361 *
362 *	This routine is called by sys_msgget, sys_semget() and sys_shmget()
363 *	when the key is not IPC_PRIVATE.
364 *	It adds a new entry if the key is not found and does some permission
365 *      / security checkings if the key is found.
366 *
367 *	On success, the ipc id is returned.
368 */
369static int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
370		struct ipc_ops *ops, struct ipc_params *params)
371{
372	struct kern_ipc_perm *ipcp;
373	int flg = params->flg;
374	int err;
375
376	/*
377	 * Take the lock as a writer since we are potentially going to add
378	 * a new entry + read locks are not "upgradable"
379	 */
380	down_write(&ids->rw_mutex);
381	ipcp = ipc_findkey(ids, params->key);
382	if (ipcp == NULL) {
383		/* key not used */
384		if (!(flg & IPC_CREAT))
385			err = -ENOENT;
386		else
387			err = ops->getnew(ns, params);
388	} else {
389		/* ipc object has been locked by ipc_findkey() */
390
391		if (flg & IPC_CREAT && flg & IPC_EXCL)
392			err = -EEXIST;
393		else {
394			err = 0;
395			if (ops->more_checks)
396				err = ops->more_checks(ipcp, params);
397			if (!err)
398				/*
399				 * ipc_check_perms returns the IPC id on
400				 * success
401				 */
402				err = ipc_check_perms(ns, ipcp, ops, params);
403		}
404		ipc_unlock(ipcp);
405	}
406	up_write(&ids->rw_mutex);
407
408	return err;
409}
410
411
412/**
413 *	ipc_rmid	-	remove an IPC identifier
414 *	@ids: IPC identifier set
415 *	@ipcp: ipc perm structure containing the identifier to remove
416 *
417 *	ipc_ids.rw_mutex (as a writer) and the spinlock for this ID are held
418 *	before this function is called, and remain locked on the exit.
419 */
420
421void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
422{
423	int lid = ipcid_to_idx(ipcp->id);
424
425	idr_remove(&ids->ipcs_idr, lid);
426
427	ids->in_use--;
428
429	ipcp->deleted = 1;
430
431	return;
432}
433
434/**
435 *	ipc_alloc	-	allocate ipc space
436 *	@size: size desired
437 *
438 *	Allocate memory from the appropriate pools and return a pointer to it.
439 *	NULL is returned if the allocation fails
440 */
441
442void *ipc_alloc(int size)
443{
444	void *out;
445	if(size > PAGE_SIZE)
446		out = vmalloc(size);
447	else
448		out = kmalloc(size, GFP_KERNEL);
449	return out;
450}
451
452/**
453 *	ipc_free        -       free ipc space
454 *	@ptr: pointer returned by ipc_alloc
455 *	@size: size of block
456 *
457 *	Free a block created with ipc_alloc(). The caller must know the size
458 *	used in the allocation call.
459 */
460
461void ipc_free(void* ptr, int size)
462{
463	if(size > PAGE_SIZE)
464		vfree(ptr);
465	else
466		kfree(ptr);
467}
468
469/*
470 * rcu allocations:
471 * There are three headers that are prepended to the actual allocation:
472 * - during use: ipc_rcu_hdr.
473 * - during the rcu grace period: ipc_rcu_grace.
474 * - [only if vmalloc]: ipc_rcu_sched.
475 * Their lifetime doesn't overlap, thus the headers share the same memory.
476 * Unlike a normal union, they are right-aligned, thus some container_of
477 * forward/backward casting is necessary:
478 */
479struct ipc_rcu_hdr
480{
481	atomic_t refcount;
482	int is_vmalloc;
483	void *data[0];
484};
485
486
487struct ipc_rcu_grace
488{
489	struct rcu_head rcu;
490	/* "void *" makes sure alignment of following data is sane. */
491	void *data[0];
492};
493
494struct ipc_rcu_sched
495{
496	struct work_struct work;
497	/* "void *" makes sure alignment of following data is sane. */
498	void *data[0];
499};
500
501#define HDRLEN_KMALLOC		(sizeof(struct ipc_rcu_grace) > sizeof(struct ipc_rcu_hdr) ? \
502					sizeof(struct ipc_rcu_grace) : sizeof(struct ipc_rcu_hdr))
503#define HDRLEN_VMALLOC		(sizeof(struct ipc_rcu_sched) > HDRLEN_KMALLOC ? \
504					sizeof(struct ipc_rcu_sched) : HDRLEN_KMALLOC)
505
506static inline int rcu_use_vmalloc(int size)
507{
508	/* Too big for a single page? */
509	if (HDRLEN_KMALLOC + size > PAGE_SIZE)
510		return 1;
511	return 0;
512}
513
514/**
515 *	ipc_rcu_alloc	-	allocate ipc and rcu space
516 *	@size: size desired
517 *
518 *	Allocate memory for the rcu header structure +  the object.
519 *	Returns the pointer to the object or NULL upon failure.
520 */
521void *ipc_rcu_alloc(int size)
522{
523	void *out;
524
525	/*
526	 * We prepend the allocation with the rcu struct, and
527	 * workqueue if necessary (for vmalloc).
528	 */
529	if (rcu_use_vmalloc(size)) {
530		out = vmalloc(HDRLEN_VMALLOC + size);
531		if (!out)
532			goto done;
533
534		out += HDRLEN_VMALLOC;
535		container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 1;
536	} else {
537		out = kmalloc(HDRLEN_KMALLOC + size, GFP_KERNEL);
538		if (!out)
539			goto done;
540
541		out += HDRLEN_KMALLOC;
542		container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 0;
543	}
544
545	/* set reference counter no matter what kind of allocation was done */
546	atomic_set(&container_of(out, struct ipc_rcu_hdr, data)->refcount, 1);
547done:
548	return out;
549}
550
551int ipc_rcu_getref(void *ptr)
552{
553	return atomic_inc_not_zero(&container_of(ptr, struct ipc_rcu_hdr, data)->refcount);
554}
555
556static void ipc_do_vfree(struct work_struct *work)
557{
558	vfree(container_of(work, struct ipc_rcu_sched, work));
559}
560
561/**
562 * ipc_schedule_free - free ipc + rcu space
563 * @head: RCU callback structure for queued work
564 *
565 * Since RCU callback function is called in bh,
566 * we need to defer the vfree to schedule_work().
567 */
568static void ipc_schedule_free(struct rcu_head *head)
569{
570	struct ipc_rcu_grace *grace;
571	struct ipc_rcu_sched *sched;
572
573	grace = container_of(head, struct ipc_rcu_grace, rcu);
574	sched = container_of(&(grace->data[0]), struct ipc_rcu_sched,
575				data[0]);
576
577	INIT_WORK(&sched->work, ipc_do_vfree);
578	schedule_work(&sched->work);
579}
580
581void ipc_rcu_putref(void *ptr)
582{
583	if (!atomic_dec_and_test(&container_of(ptr, struct ipc_rcu_hdr, data)->refcount))
584		return;
585
586	if (container_of(ptr, struct ipc_rcu_hdr, data)->is_vmalloc) {
587		call_rcu(&container_of(ptr, struct ipc_rcu_grace, data)->rcu,
588				ipc_schedule_free);
589	} else {
590		kfree_rcu(container_of(ptr, struct ipc_rcu_grace, data), rcu);
591	}
592}
593
594/**
595 *	ipcperms	-	check IPC permissions
596 *	@ns: IPC namespace
597 *	@ipcp: IPC permission set
598 *	@flag: desired permission set.
599 *
600 *	Check user, group, other permissions for access
601 *	to ipc resources. return 0 if allowed
602 *
603 * 	@flag will most probably be 0 or S_...UGO from <linux/stat.h>
604 */
605
606int ipcperms(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp, short flag)
607{
608	kuid_t euid = current_euid();
609	int requested_mode, granted_mode;
610
611	audit_ipc_obj(ipcp);
612	requested_mode = (flag >> 6) | (flag >> 3) | flag;
613	granted_mode = ipcp->mode;
614	if (uid_eq(euid, ipcp->cuid) ||
615	    uid_eq(euid, ipcp->uid))
616		granted_mode >>= 6;
617	else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
618		granted_mode >>= 3;
619	/* is there some bit set in requested_mode but not in granted_mode? */
620	if ((requested_mode & ~granted_mode & 0007) &&
621	    !ns_capable(ns->user_ns, CAP_IPC_OWNER))
622		return -1;
623
624	return security_ipc_permission(ipcp, flag);
625}
626
627/*
628 * Functions to convert between the kern_ipc_perm structure and the
629 * old/new ipc_perm structures
630 */
631
632/**
633 *	kernel_to_ipc64_perm	-	convert kernel ipc permissions to user
634 *	@in: kernel permissions
635 *	@out: new style IPC permissions
636 *
637 *	Turn the kernel object @in into a set of permissions descriptions
638 *	for returning to userspace (@out).
639 */
640
641
642void kernel_to_ipc64_perm (struct kern_ipc_perm *in, struct ipc64_perm *out)
643{
644	out->key	= in->key;
645	out->uid	= from_kuid_munged(current_user_ns(), in->uid);
646	out->gid	= from_kgid_munged(current_user_ns(), in->gid);
647	out->cuid	= from_kuid_munged(current_user_ns(), in->cuid);
648	out->cgid	= from_kgid_munged(current_user_ns(), in->cgid);
649	out->mode	= in->mode;
650	out->seq	= in->seq;
651}
652
653/**
654 *	ipc64_perm_to_ipc_perm	-	convert new ipc permissions to old
655 *	@in: new style IPC permissions
656 *	@out: old style IPC permissions
657 *
658 *	Turn the new style permissions object @in into a compatibility
659 *	object and store it into the @out pointer.
660 */
661
662void ipc64_perm_to_ipc_perm (struct ipc64_perm *in, struct ipc_perm *out)
663{
664	out->key	= in->key;
665	SET_UID(out->uid, in->uid);
666	SET_GID(out->gid, in->gid);
667	SET_UID(out->cuid, in->cuid);
668	SET_GID(out->cgid, in->cgid);
669	out->mode	= in->mode;
670	out->seq	= in->seq;
671}
672
673/**
674 * ipc_obtain_object
675 * @ids: ipc identifier set
676 * @id: ipc id to look for
677 *
678 * Look for an id in the ipc ids idr and return associated ipc object.
679 *
680 * Call inside the RCU critical section.
681 * The ipc object is *not* locked on exit.
682 */
683struct kern_ipc_perm *ipc_obtain_object(struct ipc_ids *ids, int id)
684{
685	struct kern_ipc_perm *out;
686	int lid = ipcid_to_idx(id);
687
688	out = idr_find(&ids->ipcs_idr, lid);
689	if (!out)
690		return ERR_PTR(-EINVAL);
691
692	return out;
693}
694
695/**
696 * ipc_lock - Lock an ipc structure without rw_mutex held
697 * @ids: IPC identifier set
698 * @id: ipc id to look for
699 *
700 * Look for an id in the ipc ids idr and lock the associated ipc object.
701 *
702 * The ipc object is locked on successful exit.
703 */
704struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
705{
706	struct kern_ipc_perm *out;
707
708	rcu_read_lock();
709	out = ipc_obtain_object(ids, id);
710	if (IS_ERR(out))
711		goto err1;
712
713	spin_lock(&out->lock);
714
715	/* ipc_rmid() may have already freed the ID while ipc_lock
716	 * was spinning: here verify that the structure is still valid
717	 */
718	if (!out->deleted)
719		return out;
720
721	spin_unlock(&out->lock);
722	out = ERR_PTR(-EINVAL);
723err1:
724	rcu_read_unlock();
725	return out;
726}
727
728/**
729 * ipc_obtain_object_check
730 * @ids: ipc identifier set
731 * @id: ipc id to look for
732 *
733 * Similar to ipc_obtain_object() but also checks
734 * the ipc object reference counter.
735 *
736 * Call inside the RCU critical section.
737 * The ipc object is *not* locked on exit.
738 */
739struct kern_ipc_perm *ipc_obtain_object_check(struct ipc_ids *ids, int id)
740{
741	struct kern_ipc_perm *out = ipc_obtain_object(ids, id);
742
743	if (IS_ERR(out))
744		goto out;
745
746	if (ipc_checkid(out, id))
747		return ERR_PTR(-EIDRM);
748out:
749	return out;
750}
751
752struct kern_ipc_perm *ipc_lock_check(struct ipc_ids *ids, int id)
753{
754	struct kern_ipc_perm *out;
755
756	out = ipc_lock(ids, id);
757	if (IS_ERR(out))
758		return out;
759
760	if (ipc_checkid(out, id)) {
761		ipc_unlock(out);
762		return ERR_PTR(-EIDRM);
763	}
764
765	return out;
766}
767
768/**
769 * ipcget - Common sys_*get() code
770 * @ns : namsepace
771 * @ids : IPC identifier set
772 * @ops : operations to be called on ipc object creation, permission checks
773 *        and further checks
774 * @params : the parameters needed by the previous operations.
775 *
776 * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
777 */
778int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
779			struct ipc_ops *ops, struct ipc_params *params)
780{
781	if (params->key == IPC_PRIVATE)
782		return ipcget_new(ns, ids, ops, params);
783	else
784		return ipcget_public(ns, ids, ops, params);
785}
786
787/**
788 * ipc_update_perm - update the permissions of an IPC.
789 * @in:  the permission given as input.
790 * @out: the permission of the ipc to set.
791 */
792int ipc_update_perm(struct ipc64_perm *in, struct kern_ipc_perm *out)
793{
794	kuid_t uid = make_kuid(current_user_ns(), in->uid);
795	kgid_t gid = make_kgid(current_user_ns(), in->gid);
796	if (!uid_valid(uid) || !gid_valid(gid))
797		return -EINVAL;
798
799	out->uid = uid;
800	out->gid = gid;
801	out->mode = (out->mode & ~S_IRWXUGO)
802		| (in->mode & S_IRWXUGO);
803
804	return 0;
805}
806
807/**
808 * ipcctl_pre_down - retrieve an ipc and check permissions for some IPC_XXX cmd
809 * @ns:  the ipc namespace
810 * @ids:  the table of ids where to look for the ipc
811 * @id:   the id of the ipc to retrieve
812 * @cmd:  the cmd to check
813 * @perm: the permission to set
814 * @extra_perm: one extra permission parameter used by msq
815 *
816 * This function does some common audit and permissions check for some IPC_XXX
817 * cmd and is called from semctl_down, shmctl_down and msgctl_down.
818 * It must be called without any lock held and
819 *  - retrieves the ipc with the given id in the given table.
820 *  - performs some audit and permission check, depending on the given cmd
821 *  - returns the ipc with both ipc and rw_mutex locks held in case of success
822 *    or an err-code without any lock held otherwise.
823 */
824struct kern_ipc_perm *ipcctl_pre_down(struct ipc_namespace *ns,
825				      struct ipc_ids *ids, int id, int cmd,
826				      struct ipc64_perm *perm, int extra_perm)
827{
828	struct kern_ipc_perm *ipcp;
829
830	ipcp = ipcctl_pre_down_nolock(ns, ids, id, cmd, perm, extra_perm);
831	if (IS_ERR(ipcp))
832		goto out;
833
834	spin_lock(&ipcp->lock);
835out:
836	return ipcp;
837}
838
839struct kern_ipc_perm *ipcctl_pre_down_nolock(struct ipc_namespace *ns,
840					     struct ipc_ids *ids, int id, int cmd,
841					     struct ipc64_perm *perm, int extra_perm)
842{
843	kuid_t euid;
844	int err = -EPERM;
845	struct kern_ipc_perm *ipcp;
846
847	down_write(&ids->rw_mutex);
848	rcu_read_lock();
849
850	ipcp = ipc_obtain_object_check(ids, id);
851	if (IS_ERR(ipcp)) {
852		err = PTR_ERR(ipcp);
853		goto out_up;
854	}
855
856	audit_ipc_obj(ipcp);
857	if (cmd == IPC_SET)
858		audit_ipc_set_perm(extra_perm, perm->uid,
859				   perm->gid, perm->mode);
860
861	euid = current_euid();
862	if (uid_eq(euid, ipcp->cuid) || uid_eq(euid, ipcp->uid)  ||
863	    ns_capable(ns->user_ns, CAP_SYS_ADMIN))
864		return ipcp;
865
866out_up:
867	/*
868	 * Unsuccessful lookup, unlock and return
869	 * the corresponding error.
870	 */
871	rcu_read_unlock();
872	up_write(&ids->rw_mutex);
873
874	return ERR_PTR(err);
875}
876
877#ifdef CONFIG_ARCH_WANT_IPC_PARSE_VERSION
878
879
880/**
881 *	ipc_parse_version	-	IPC call version
882 *	@cmd: pointer to command
883 *
884 *	Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
885 *	The @cmd value is turned from an encoding command and version into
886 *	just the command code.
887 */
888
889int ipc_parse_version (int *cmd)
890{
891	if (*cmd & IPC_64) {
892		*cmd ^= IPC_64;
893		return IPC_64;
894	} else {
895		return IPC_OLD;
896	}
897}
898
899#endif /* CONFIG_ARCH_WANT_IPC_PARSE_VERSION */
900
901#ifdef CONFIG_PROC_FS
902struct ipc_proc_iter {
903	struct ipc_namespace *ns;
904	struct ipc_proc_iface *iface;
905};
906
907/*
908 * This routine locks the ipc structure found at least at position pos.
909 */
910static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
911					      loff_t *new_pos)
912{
913	struct kern_ipc_perm *ipc;
914	int total, id;
915
916	total = 0;
917	for (id = 0; id < pos && total < ids->in_use; id++) {
918		ipc = idr_find(&ids->ipcs_idr, id);
919		if (ipc != NULL)
920			total++;
921	}
922
923	if (total >= ids->in_use)
924		return NULL;
925
926	for ( ; pos < IPCMNI; pos++) {
927		ipc = idr_find(&ids->ipcs_idr, pos);
928		if (ipc != NULL) {
929			*new_pos = pos + 1;
930			ipc_lock_by_ptr(ipc);
931			return ipc;
932		}
933	}
934
935	/* Out of range - return NULL to terminate iteration */
936	return NULL;
937}
938
939static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
940{
941	struct ipc_proc_iter *iter = s->private;
942	struct ipc_proc_iface *iface = iter->iface;
943	struct kern_ipc_perm *ipc = it;
944
945	/* If we had an ipc id locked before, unlock it */
946	if (ipc && ipc != SEQ_START_TOKEN)
947		ipc_unlock(ipc);
948
949	return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
950}
951
952/*
953 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
954 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
955 */
956static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
957{
958	struct ipc_proc_iter *iter = s->private;
959	struct ipc_proc_iface *iface = iter->iface;
960	struct ipc_ids *ids;
961
962	ids = &iter->ns->ids[iface->ids];
963
964	/*
965	 * Take the lock - this will be released by the corresponding
966	 * call to stop().
967	 */
968	down_read(&ids->rw_mutex);
969
970	/* pos < 0 is invalid */
971	if (*pos < 0)
972		return NULL;
973
974	/* pos == 0 means header */
975	if (*pos == 0)
976		return SEQ_START_TOKEN;
977
978	/* Find the (pos-1)th ipc */
979	return sysvipc_find_ipc(ids, *pos - 1, pos);
980}
981
982static void sysvipc_proc_stop(struct seq_file *s, void *it)
983{
984	struct kern_ipc_perm *ipc = it;
985	struct ipc_proc_iter *iter = s->private;
986	struct ipc_proc_iface *iface = iter->iface;
987	struct ipc_ids *ids;
988
989	/* If we had a locked structure, release it */
990	if (ipc && ipc != SEQ_START_TOKEN)
991		ipc_unlock(ipc);
992
993	ids = &iter->ns->ids[iface->ids];
994	/* Release the lock we took in start() */
995	up_read(&ids->rw_mutex);
996}
997
998static int sysvipc_proc_show(struct seq_file *s, void *it)
999{
1000	struct ipc_proc_iter *iter = s->private;
1001	struct ipc_proc_iface *iface = iter->iface;
1002
1003	if (it == SEQ_START_TOKEN)
1004		return seq_puts(s, iface->header);
1005
1006	return iface->show(s, it);
1007}
1008
1009static const struct seq_operations sysvipc_proc_seqops = {
1010	.start = sysvipc_proc_start,
1011	.stop  = sysvipc_proc_stop,
1012	.next  = sysvipc_proc_next,
1013	.show  = sysvipc_proc_show,
1014};
1015
1016static int sysvipc_proc_open(struct inode *inode, struct file *file)
1017{
1018	int ret;
1019	struct seq_file *seq;
1020	struct ipc_proc_iter *iter;
1021
1022	ret = -ENOMEM;
1023	iter = kmalloc(sizeof(*iter), GFP_KERNEL);
1024	if (!iter)
1025		goto out;
1026
1027	ret = seq_open(file, &sysvipc_proc_seqops);
1028	if (ret)
1029		goto out_kfree;
1030
1031	seq = file->private_data;
1032	seq->private = iter;
1033
1034	iter->iface = PDE(inode)->data;
1035	iter->ns    = get_ipc_ns(current->nsproxy->ipc_ns);
1036out:
1037	return ret;
1038out_kfree:
1039	kfree(iter);
1040	goto out;
1041}
1042
1043static int sysvipc_proc_release(struct inode *inode, struct file *file)
1044{
1045	struct seq_file *seq = file->private_data;
1046	struct ipc_proc_iter *iter = seq->private;
1047	put_ipc_ns(iter->ns);
1048	return seq_release_private(inode, file);
1049}
1050
1051static const struct file_operations sysvipc_proc_fops = {
1052	.open    = sysvipc_proc_open,
1053	.read    = seq_read,
1054	.llseek  = seq_lseek,
1055	.release = sysvipc_proc_release,
1056};
1057#endif /* CONFIG_PROC_FS */
1058