util.c revision 8f68fa2d
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	int 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.
520 *	NULL is returned if the allocation fails.
521 */
522
523void* ipc_rcu_alloc(int size)
524{
525	void* out;
526	/*
527	 * We prepend the allocation with the rcu struct, and
528	 * workqueue if necessary (for vmalloc).
529	 */
530	if (rcu_use_vmalloc(size)) {
531		out = vmalloc(HDRLEN_VMALLOC + size);
532		if (out) {
533			out += HDRLEN_VMALLOC;
534			container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 1;
535			container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
536		}
537	} else {
538		out = kmalloc(HDRLEN_KMALLOC + size, GFP_KERNEL);
539		if (out) {
540			out += HDRLEN_KMALLOC;
541			container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 0;
542			container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
543		}
544	}
545
546	return out;
547}
548
549void ipc_rcu_getref(void *ptr)
550{
551	container_of(ptr, struct ipc_rcu_hdr, data)->refcount++;
552}
553
554static void ipc_do_vfree(struct work_struct *work)
555{
556	vfree(container_of(work, struct ipc_rcu_sched, work));
557}
558
559/**
560 * ipc_schedule_free - free ipc + rcu space
561 * @head: RCU callback structure for queued work
562 *
563 * Since RCU callback function is called in bh,
564 * we need to defer the vfree to schedule_work().
565 */
566static void ipc_schedule_free(struct rcu_head *head)
567{
568	struct ipc_rcu_grace *grace;
569	struct ipc_rcu_sched *sched;
570
571	grace = container_of(head, struct ipc_rcu_grace, rcu);
572	sched = container_of(&(grace->data[0]), struct ipc_rcu_sched,
573				data[0]);
574
575	INIT_WORK(&sched->work, ipc_do_vfree);
576	schedule_work(&sched->work);
577}
578
579void ipc_rcu_putref(void *ptr)
580{
581	if (--container_of(ptr, struct ipc_rcu_hdr, data)->refcount > 0)
582		return;
583
584	if (container_of(ptr, struct ipc_rcu_hdr, data)->is_vmalloc) {
585		call_rcu(&container_of(ptr, struct ipc_rcu_grace, data)->rcu,
586				ipc_schedule_free);
587	} else {
588		kfree_rcu(container_of(ptr, struct ipc_rcu_grace, data), rcu);
589	}
590}
591
592/**
593 *	ipcperms	-	check IPC permissions
594 *	@ns: IPC namespace
595 *	@ipcp: IPC permission set
596 *	@flag: desired permission set.
597 *
598 *	Check user, group, other permissions for access
599 *	to ipc resources. return 0 if allowed
600 *
601 * 	@flag will most probably be 0 or S_...UGO from <linux/stat.h>
602 */
603
604int ipcperms(struct ipc_namespace *ns, struct kern_ipc_perm *ipcp, short flag)
605{
606	kuid_t euid = current_euid();
607	int requested_mode, granted_mode;
608
609	audit_ipc_obj(ipcp);
610	requested_mode = (flag >> 6) | (flag >> 3) | flag;
611	granted_mode = ipcp->mode;
612	if (uid_eq(euid, ipcp->cuid) ||
613	    uid_eq(euid, ipcp->uid))
614		granted_mode >>= 6;
615	else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
616		granted_mode >>= 3;
617	/* is there some bit set in requested_mode but not in granted_mode? */
618	if ((requested_mode & ~granted_mode & 0007) &&
619	    !ns_capable(ns->user_ns, CAP_IPC_OWNER))
620		return -1;
621
622	return security_ipc_permission(ipcp, flag);
623}
624
625/*
626 * Functions to convert between the kern_ipc_perm structure and the
627 * old/new ipc_perm structures
628 */
629
630/**
631 *	kernel_to_ipc64_perm	-	convert kernel ipc permissions to user
632 *	@in: kernel permissions
633 *	@out: new style IPC permissions
634 *
635 *	Turn the kernel object @in into a set of permissions descriptions
636 *	for returning to userspace (@out).
637 */
638
639
640void kernel_to_ipc64_perm (struct kern_ipc_perm *in, struct ipc64_perm *out)
641{
642	out->key	= in->key;
643	out->uid	= from_kuid_munged(current_user_ns(), in->uid);
644	out->gid	= from_kgid_munged(current_user_ns(), in->gid);
645	out->cuid	= from_kuid_munged(current_user_ns(), in->cuid);
646	out->cgid	= from_kgid_munged(current_user_ns(), in->cgid);
647	out->mode	= in->mode;
648	out->seq	= in->seq;
649}
650
651/**
652 *	ipc64_perm_to_ipc_perm	-	convert new ipc permissions to old
653 *	@in: new style IPC permissions
654 *	@out: old style IPC permissions
655 *
656 *	Turn the new style permissions object @in into a compatibility
657 *	object and store it into the @out pointer.
658 */
659
660void ipc64_perm_to_ipc_perm (struct ipc64_perm *in, struct ipc_perm *out)
661{
662	out->key	= in->key;
663	SET_UID(out->uid, in->uid);
664	SET_GID(out->gid, in->gid);
665	SET_UID(out->cuid, in->cuid);
666	SET_GID(out->cgid, in->cgid);
667	out->mode	= in->mode;
668	out->seq	= in->seq;
669}
670
671/**
672 * ipc_lock - Lock an ipc structure without rw_mutex held
673 * @ids: IPC identifier set
674 * @id: ipc id to look for
675 *
676 * Look for an id in the ipc ids idr and lock the associated ipc object.
677 *
678 * The ipc object is locked on exit.
679 */
680
681struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
682{
683	struct kern_ipc_perm *out;
684	int lid = ipcid_to_idx(id);
685
686	rcu_read_lock();
687	out = idr_find(&ids->ipcs_idr, lid);
688	if (out == NULL) {
689		rcu_read_unlock();
690		return ERR_PTR(-EINVAL);
691	}
692
693	spin_lock(&out->lock);
694
695	/* ipc_rmid() may have already freed the ID while ipc_lock
696	 * was spinning: here verify that the structure is still valid
697	 */
698	if (out->deleted) {
699		spin_unlock(&out->lock);
700		rcu_read_unlock();
701		return ERR_PTR(-EINVAL);
702	}
703
704	return out;
705}
706
707struct kern_ipc_perm *ipc_lock_check(struct ipc_ids *ids, int id)
708{
709	struct kern_ipc_perm *out;
710
711	out = ipc_lock(ids, id);
712	if (IS_ERR(out))
713		return out;
714
715	if (ipc_checkid(out, id)) {
716		ipc_unlock(out);
717		return ERR_PTR(-EIDRM);
718	}
719
720	return out;
721}
722
723/**
724 * ipcget - Common sys_*get() code
725 * @ns : namsepace
726 * @ids : IPC identifier set
727 * @ops : operations to be called on ipc object creation, permission checks
728 *        and further checks
729 * @params : the parameters needed by the previous operations.
730 *
731 * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
732 */
733int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
734			struct ipc_ops *ops, struct ipc_params *params)
735{
736	if (params->key == IPC_PRIVATE)
737		return ipcget_new(ns, ids, ops, params);
738	else
739		return ipcget_public(ns, ids, ops, params);
740}
741
742/**
743 * ipc_update_perm - update the permissions of an IPC.
744 * @in:  the permission given as input.
745 * @out: the permission of the ipc to set.
746 */
747int ipc_update_perm(struct ipc64_perm *in, struct kern_ipc_perm *out)
748{
749	kuid_t uid = make_kuid(current_user_ns(), in->uid);
750	kgid_t gid = make_kgid(current_user_ns(), in->gid);
751	if (!uid_valid(uid) || !gid_valid(gid))
752		return -EINVAL;
753
754	out->uid = uid;
755	out->gid = gid;
756	out->mode = (out->mode & ~S_IRWXUGO)
757		| (in->mode & S_IRWXUGO);
758
759	return 0;
760}
761
762/**
763 * ipcctl_pre_down - retrieve an ipc and check permissions for some IPC_XXX cmd
764 * @ns:  the ipc namespace
765 * @ids:  the table of ids where to look for the ipc
766 * @id:   the id of the ipc to retrieve
767 * @cmd:  the cmd to check
768 * @perm: the permission to set
769 * @extra_perm: one extra permission parameter used by msq
770 *
771 * This function does some common audit and permissions check for some IPC_XXX
772 * cmd and is called from semctl_down, shmctl_down and msgctl_down.
773 * It must be called without any lock held and
774 *  - retrieves the ipc with the given id in the given table.
775 *  - performs some audit and permission check, depending on the given cmd
776 *  - returns the ipc with both ipc and rw_mutex locks held in case of success
777 *    or an err-code without any lock held otherwise.
778 */
779struct kern_ipc_perm *ipcctl_pre_down(struct ipc_namespace *ns,
780				      struct ipc_ids *ids, int id, int cmd,
781				      struct ipc64_perm *perm, int extra_perm)
782{
783	struct kern_ipc_perm *ipcp;
784	kuid_t euid;
785	int err;
786
787	down_write(&ids->rw_mutex);
788	ipcp = ipc_lock_check(ids, id);
789	if (IS_ERR(ipcp)) {
790		err = PTR_ERR(ipcp);
791		goto out_up;
792	}
793
794	audit_ipc_obj(ipcp);
795	if (cmd == IPC_SET)
796		audit_ipc_set_perm(extra_perm, perm->uid,
797					 perm->gid, perm->mode);
798
799	euid = current_euid();
800	if (uid_eq(euid, ipcp->cuid) || uid_eq(euid, ipcp->uid)  ||
801	    ns_capable(ns->user_ns, CAP_SYS_ADMIN))
802		return ipcp;
803
804	err = -EPERM;
805	ipc_unlock(ipcp);
806out_up:
807	up_write(&ids->rw_mutex);
808	return ERR_PTR(err);
809}
810
811#ifdef CONFIG_ARCH_WANT_IPC_PARSE_VERSION
812
813
814/**
815 *	ipc_parse_version	-	IPC call version
816 *	@cmd: pointer to command
817 *
818 *	Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
819 *	The @cmd value is turned from an encoding command and version into
820 *	just the command code.
821 */
822
823int ipc_parse_version (int *cmd)
824{
825	if (*cmd & IPC_64) {
826		*cmd ^= IPC_64;
827		return IPC_64;
828	} else {
829		return IPC_OLD;
830	}
831}
832
833#endif /* CONFIG_ARCH_WANT_IPC_PARSE_VERSION */
834
835#ifdef CONFIG_PROC_FS
836struct ipc_proc_iter {
837	struct ipc_namespace *ns;
838	struct ipc_proc_iface *iface;
839};
840
841/*
842 * This routine locks the ipc structure found at least at position pos.
843 */
844static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
845					      loff_t *new_pos)
846{
847	struct kern_ipc_perm *ipc;
848	int total, id;
849
850	total = 0;
851	for (id = 0; id < pos && total < ids->in_use; id++) {
852		ipc = idr_find(&ids->ipcs_idr, id);
853		if (ipc != NULL)
854			total++;
855	}
856
857	if (total >= ids->in_use)
858		return NULL;
859
860	for ( ; pos < IPCMNI; pos++) {
861		ipc = idr_find(&ids->ipcs_idr, pos);
862		if (ipc != NULL) {
863			*new_pos = pos + 1;
864			ipc_lock_by_ptr(ipc);
865			return ipc;
866		}
867	}
868
869	/* Out of range - return NULL to terminate iteration */
870	return NULL;
871}
872
873static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
874{
875	struct ipc_proc_iter *iter = s->private;
876	struct ipc_proc_iface *iface = iter->iface;
877	struct kern_ipc_perm *ipc = it;
878
879	/* If we had an ipc id locked before, unlock it */
880	if (ipc && ipc != SEQ_START_TOKEN)
881		ipc_unlock(ipc);
882
883	return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
884}
885
886/*
887 * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
888 * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
889 */
890static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
891{
892	struct ipc_proc_iter *iter = s->private;
893	struct ipc_proc_iface *iface = iter->iface;
894	struct ipc_ids *ids;
895
896	ids = &iter->ns->ids[iface->ids];
897
898	/*
899	 * Take the lock - this will be released by the corresponding
900	 * call to stop().
901	 */
902	down_read(&ids->rw_mutex);
903
904	/* pos < 0 is invalid */
905	if (*pos < 0)
906		return NULL;
907
908	/* pos == 0 means header */
909	if (*pos == 0)
910		return SEQ_START_TOKEN;
911
912	/* Find the (pos-1)th ipc */
913	return sysvipc_find_ipc(ids, *pos - 1, pos);
914}
915
916static void sysvipc_proc_stop(struct seq_file *s, void *it)
917{
918	struct kern_ipc_perm *ipc = it;
919	struct ipc_proc_iter *iter = s->private;
920	struct ipc_proc_iface *iface = iter->iface;
921	struct ipc_ids *ids;
922
923	/* If we had a locked structure, release it */
924	if (ipc && ipc != SEQ_START_TOKEN)
925		ipc_unlock(ipc);
926
927	ids = &iter->ns->ids[iface->ids];
928	/* Release the lock we took in start() */
929	up_read(&ids->rw_mutex);
930}
931
932static int sysvipc_proc_show(struct seq_file *s, void *it)
933{
934	struct ipc_proc_iter *iter = s->private;
935	struct ipc_proc_iface *iface = iter->iface;
936
937	if (it == SEQ_START_TOKEN)
938		return seq_puts(s, iface->header);
939
940	return iface->show(s, it);
941}
942
943static const struct seq_operations sysvipc_proc_seqops = {
944	.start = sysvipc_proc_start,
945	.stop  = sysvipc_proc_stop,
946	.next  = sysvipc_proc_next,
947	.show  = sysvipc_proc_show,
948};
949
950static int sysvipc_proc_open(struct inode *inode, struct file *file)
951{
952	int ret;
953	struct seq_file *seq;
954	struct ipc_proc_iter *iter;
955
956	ret = -ENOMEM;
957	iter = kmalloc(sizeof(*iter), GFP_KERNEL);
958	if (!iter)
959		goto out;
960
961	ret = seq_open(file, &sysvipc_proc_seqops);
962	if (ret)
963		goto out_kfree;
964
965	seq = file->private_data;
966	seq->private = iter;
967
968	iter->iface = PDE(inode)->data;
969	iter->ns    = get_ipc_ns(current->nsproxy->ipc_ns);
970out:
971	return ret;
972out_kfree:
973	kfree(iter);
974	goto out;
975}
976
977static int sysvipc_proc_release(struct inode *inode, struct file *file)
978{
979	struct seq_file *seq = file->private_data;
980	struct ipc_proc_iter *iter = seq->private;
981	put_ipc_ns(iter->ns);
982	return seq_release_private(inode, file);
983}
984
985static const struct file_operations sysvipc_proc_fops = {
986	.open    = sysvipc_proc_open,
987	.read    = seq_read,
988	.llseek  = seq_lseek,
989	.release = sysvipc_proc_release,
990};
991#endif /* CONFIG_PROC_FS */
992