• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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/capability.h>
27 #include <linux/highuid.h>
28 #include <linux/security.h>
29 #include <linux/rcupdate.h>
30 #include <linux/workqueue.h>
31 #include <linux/seq_file.h>
32 #include <linux/proc_fs.h>
33 #include <linux/audit.h>
34 #include <linux/nsproxy.h>
35 #include <linux/rwsem.h>
36 #include <linux/memory.h>
37 #include <linux/ipc_namespace.h>
38 
39 #include <asm/unistd.h>
40 
41 #include "util.h"
42 
43 struct ipc_proc_iface {
44 	const char *path;
45 	const char *header;
46 	int ids;
47 	int (*show)(struct seq_file *, void *);
48 };
49 
50 struct ipc_namespace init_ipc_ns = {
51 	.kref = {
52 		.refcount	= ATOMIC_INIT(2),
53 	},
54 };
55 
56 atomic_t nr_ipc_ns = ATOMIC_INIT(1);
57 
58 
59 #ifdef CONFIG_MEMORY_HOTPLUG
60 
ipc_memory_notifier(struct work_struct * work)61 static void ipc_memory_notifier(struct work_struct *work)
62 {
63 	ipcns_notify(IPCNS_MEMCHANGED);
64 }
65 
66 static DECLARE_WORK(ipc_memory_wq, ipc_memory_notifier);
67 
68 
ipc_memory_callback(struct notifier_block * self,unsigned long action,void * arg)69 static int ipc_memory_callback(struct notifier_block *self,
70 				unsigned long action, void *arg)
71 {
72 	switch (action) {
73 	case MEM_ONLINE:    /* memory successfully brought online */
74 	case MEM_OFFLINE:   /* or offline: it's time to recompute msgmni */
75 		/*
76 		 * This is done by invoking the ipcns notifier chain with the
77 		 * IPC_MEMCHANGED event.
78 		 * In order not to keep the lock on the hotplug memory chain
79 		 * for too long, queue a work item that will, when waken up,
80 		 * activate the ipcns notification chain.
81 		 * No need to keep several ipc work items on the queue.
82 		 */
83 		if (!work_pending(&ipc_memory_wq))
84 			schedule_work(&ipc_memory_wq);
85 		break;
86 	case MEM_GOING_ONLINE:
87 	case MEM_GOING_OFFLINE:
88 	case MEM_CANCEL_ONLINE:
89 	case MEM_CANCEL_OFFLINE:
90 	default:
91 		break;
92 	}
93 
94 	return NOTIFY_OK;
95 }
96 
97 #endif /* CONFIG_MEMORY_HOTPLUG */
98 
99 /**
100  *	ipc_init	-	initialise IPC subsystem
101  *
102  *	The various system5 IPC resources (semaphores, messages and shared
103  *	memory) are initialised
104  *	A callback routine is registered into the memory hotplug notifier
105  *	chain: since msgmni scales to lowmem this callback routine will be
106  *	called upon successful memory add / remove to recompute msmgni.
107  */
108 
ipc_init(void)109 static int __init ipc_init(void)
110 {
111 	sem_init();
112 	msg_init();
113 	shm_init();
114 	hotplug_memory_notifier(ipc_memory_callback, IPC_CALLBACK_PRI);
115 	register_ipcns_notifier(&init_ipc_ns);
116 	return 0;
117 }
118 __initcall(ipc_init);
119 
120 /**
121  *	ipc_init_ids		-	initialise IPC identifiers
122  *	@ids: Identifier set
123  *
124  *	Set up the sequence range to use for the ipc identifier range (limited
125  *	below IPCMNI) then initialise the ids idr.
126  */
127 
ipc_init_ids(struct ipc_ids * ids)128 void ipc_init_ids(struct ipc_ids *ids)
129 {
130 	init_rwsem(&ids->rw_mutex);
131 
132 	ids->in_use = 0;
133 	ids->seq = 0;
134 	{
135 		int seq_limit = INT_MAX/SEQ_MULTIPLIER;
136 		if (seq_limit > USHORT_MAX)
137 			ids->seq_max = USHORT_MAX;
138 		 else
139 		 	ids->seq_max = seq_limit;
140 	}
141 
142 	idr_init(&ids->ipcs_idr);
143 }
144 
145 #ifdef CONFIG_PROC_FS
146 static const struct file_operations sysvipc_proc_fops;
147 /**
148  *	ipc_init_proc_interface	-  Create a proc interface for sysipc types using a seq_file interface.
149  *	@path: Path in procfs
150  *	@header: Banner to be printed at the beginning of the file.
151  *	@ids: ipc id table to iterate.
152  *	@show: show routine.
153  */
ipc_init_proc_interface(const char * path,const char * header,int ids,int (* show)(struct seq_file *,void *))154 void __init ipc_init_proc_interface(const char *path, const char *header,
155 		int ids, int (*show)(struct seq_file *, void *))
156 {
157 	struct proc_dir_entry *pde;
158 	struct ipc_proc_iface *iface;
159 
160 	iface = kmalloc(sizeof(*iface), GFP_KERNEL);
161 	if (!iface)
162 		return;
163 	iface->path	= path;
164 	iface->header	= header;
165 	iface->ids	= ids;
166 	iface->show	= show;
167 
168 	pde = proc_create_data(path,
169 			       S_IRUGO,        /* world readable */
170 			       NULL,           /* parent dir */
171 			       &sysvipc_proc_fops,
172 			       iface);
173 	if (!pde) {
174 		kfree(iface);
175 	}
176 }
177 #endif
178 
179 /**
180  *	ipc_findkey	-	find a key in an ipc identifier set
181  *	@ids: Identifier set
182  *	@key: The key to find
183  *
184  *	Requires ipc_ids.rw_mutex locked.
185  *	Returns the LOCKED pointer to the ipc structure if found or NULL
186  *	if not.
187  *	If key is found ipc points to the owning ipc structure
188  */
189 
ipc_findkey(struct ipc_ids * ids,key_t key)190 static struct kern_ipc_perm *ipc_findkey(struct ipc_ids *ids, key_t key)
191 {
192 	struct kern_ipc_perm *ipc;
193 	int next_id;
194 	int total;
195 
196 	for (total = 0, next_id = 0; total < ids->in_use; next_id++) {
197 		ipc = idr_find(&ids->ipcs_idr, next_id);
198 
199 		if (ipc == NULL)
200 			continue;
201 
202 		if (ipc->key != key) {
203 			total++;
204 			continue;
205 		}
206 
207 		ipc_lock_by_ptr(ipc);
208 		return ipc;
209 	}
210 
211 	return NULL;
212 }
213 
214 /**
215  *	ipc_get_maxid 	-	get the last assigned id
216  *	@ids: IPC identifier set
217  *
218  *	Called with ipc_ids.rw_mutex held.
219  */
220 
ipc_get_maxid(struct ipc_ids * ids)221 int ipc_get_maxid(struct ipc_ids *ids)
222 {
223 	struct kern_ipc_perm *ipc;
224 	int max_id = -1;
225 	int total, id;
226 
227 	if (ids->in_use == 0)
228 		return -1;
229 
230 	if (ids->in_use == IPCMNI)
231 		return IPCMNI - 1;
232 
233 	/* Look for the last assigned id */
234 	total = 0;
235 	for (id = 0; id < IPCMNI && total < ids->in_use; id++) {
236 		ipc = idr_find(&ids->ipcs_idr, id);
237 		if (ipc != NULL) {
238 			max_id = id;
239 			total++;
240 		}
241 	}
242 	return max_id;
243 }
244 
245 /**
246  *	ipc_addid 	-	add an IPC identifier
247  *	@ids: IPC identifier set
248  *	@new: new IPC permission set
249  *	@size: limit for the number of used ids
250  *
251  *	Add an entry 'new' to the IPC ids idr. The permissions object is
252  *	initialised and the first free entry is set up and the id assigned
253  *	is returned. The 'new' entry is returned in a locked state on success.
254  *	On failure the entry is not locked and a negative err-code is returned.
255  *
256  *	Called with ipc_ids.rw_mutex held as a writer.
257  */
258 
ipc_addid(struct ipc_ids * ids,struct kern_ipc_perm * new,int size)259 int ipc_addid(struct ipc_ids* ids, struct kern_ipc_perm* new, int size)
260 {
261 	uid_t euid;
262 	gid_t egid;
263 	int id, err;
264 
265 	if (size > IPCMNI)
266 		size = IPCMNI;
267 
268 	if (ids->in_use >= size)
269 		return -ENOSPC;
270 
271 	spin_lock_init(&new->lock);
272 	new->deleted = 0;
273 	rcu_read_lock();
274 	spin_lock(&new->lock);
275 
276 	err = idr_get_new(&ids->ipcs_idr, new, &id);
277 	if (err) {
278 		spin_unlock(&new->lock);
279 		rcu_read_unlock();
280 		return err;
281 	}
282 
283 	ids->in_use++;
284 
285 	current_euid_egid(&euid, &egid);
286 	new->cuid = new->uid = euid;
287 	new->gid = new->cgid = egid;
288 
289 	new->seq = ids->seq++;
290 	if(ids->seq > ids->seq_max)
291 		ids->seq = 0;
292 
293 	new->id = ipc_buildid(id, new->seq);
294 	return id;
295 }
296 
297 /**
298  *	ipcget_new	-	create a new ipc object
299  *	@ns: namespace
300  *	@ids: IPC identifer set
301  *	@ops: the actual creation routine to call
302  *	@params: its parameters
303  *
304  *	This routine is called by sys_msgget, sys_semget() and sys_shmget()
305  *	when the key is IPC_PRIVATE.
306  */
ipcget_new(struct ipc_namespace * ns,struct ipc_ids * ids,struct ipc_ops * ops,struct ipc_params * params)307 static int ipcget_new(struct ipc_namespace *ns, struct ipc_ids *ids,
308 		struct ipc_ops *ops, struct ipc_params *params)
309 {
310 	int err;
311 retry:
312 	err = idr_pre_get(&ids->ipcs_idr, GFP_KERNEL);
313 
314 	if (!err)
315 		return -ENOMEM;
316 
317 	down_write(&ids->rw_mutex);
318 	err = ops->getnew(ns, params);
319 	up_write(&ids->rw_mutex);
320 
321 	if (err == -EAGAIN)
322 		goto retry;
323 
324 	return err;
325 }
326 
327 /**
328  *	ipc_check_perms	-	check security and permissions for an IPC
329  *	@ipcp: ipc permission set
330  *	@ops: the actual security routine to call
331  *	@params: its parameters
332  *
333  *	This routine is called by sys_msgget(), sys_semget() and sys_shmget()
334  *      when the key is not IPC_PRIVATE and that key already exists in the
335  *      ids IDR.
336  *
337  *	On success, the IPC id is returned.
338  *
339  *	It is called with ipc_ids.rw_mutex and ipcp->lock held.
340  */
ipc_check_perms(struct kern_ipc_perm * ipcp,struct ipc_ops * ops,struct ipc_params * params)341 static int ipc_check_perms(struct kern_ipc_perm *ipcp, struct ipc_ops *ops,
342 			struct ipc_params *params)
343 {
344 	int err;
345 
346 	if (ipcperms(ipcp, params->flg))
347 		err = -EACCES;
348 	else {
349 		err = ops->associate(ipcp, params->flg);
350 		if (!err)
351 			err = ipcp->id;
352 	}
353 
354 	return err;
355 }
356 
357 /**
358  *	ipcget_public	-	get an ipc object or create a new one
359  *	@ns: namespace
360  *	@ids: IPC identifer set
361  *	@ops: the actual creation routine to call
362  *	@params: its parameters
363  *
364  *	This routine is called by sys_msgget, sys_semget() and sys_shmget()
365  *	when the key is not IPC_PRIVATE.
366  *	It adds a new entry if the key is not found and does some permission
367  *      / security checkings if the key is found.
368  *
369  *	On success, the ipc id is returned.
370  */
ipcget_public(struct ipc_namespace * ns,struct ipc_ids * ids,struct ipc_ops * ops,struct ipc_params * params)371 static int ipcget_public(struct ipc_namespace *ns, struct ipc_ids *ids,
372 		struct ipc_ops *ops, struct ipc_params *params)
373 {
374 	struct kern_ipc_perm *ipcp;
375 	int flg = params->flg;
376 	int err;
377 retry:
378 	err = idr_pre_get(&ids->ipcs_idr, GFP_KERNEL);
379 
380 	/*
381 	 * Take the lock as a writer since we are potentially going to add
382 	 * a new entry + read locks are not "upgradable"
383 	 */
384 	down_write(&ids->rw_mutex);
385 	ipcp = ipc_findkey(ids, params->key);
386 	if (ipcp == NULL) {
387 		/* key not used */
388 		if (!(flg & IPC_CREAT))
389 			err = -ENOENT;
390 		else if (!err)
391 			err = -ENOMEM;
392 		else
393 			err = ops->getnew(ns, params);
394 	} else {
395 		/* ipc object has been locked by ipc_findkey() */
396 
397 		if (flg & IPC_CREAT && flg & IPC_EXCL)
398 			err = -EEXIST;
399 		else {
400 			err = 0;
401 			if (ops->more_checks)
402 				err = ops->more_checks(ipcp, params);
403 			if (!err)
404 				/*
405 				 * ipc_check_perms returns the IPC id on
406 				 * success
407 				 */
408 				err = ipc_check_perms(ipcp, ops, params);
409 		}
410 		ipc_unlock(ipcp);
411 	}
412 	up_write(&ids->rw_mutex);
413 
414 	if (err == -EAGAIN)
415 		goto retry;
416 
417 	return err;
418 }
419 
420 
421 /**
422  *	ipc_rmid	-	remove an IPC identifier
423  *	@ids: IPC identifier set
424  *	@ipcp: ipc perm structure containing the identifier to remove
425  *
426  *	ipc_ids.rw_mutex (as a writer) and the spinlock for this ID are held
427  *	before this function is called, and remain locked on the exit.
428  */
429 
ipc_rmid(struct ipc_ids * ids,struct kern_ipc_perm * ipcp)430 void ipc_rmid(struct ipc_ids *ids, struct kern_ipc_perm *ipcp)
431 {
432 	int lid = ipcid_to_idx(ipcp->id);
433 
434 	idr_remove(&ids->ipcs_idr, lid);
435 
436 	ids->in_use--;
437 
438 	ipcp->deleted = 1;
439 
440 	return;
441 }
442 
443 /**
444  *	ipc_alloc	-	allocate ipc space
445  *	@size: size desired
446  *
447  *	Allocate memory from the appropriate pools and return a pointer to it.
448  *	NULL is returned if the allocation fails
449  */
450 
ipc_alloc(int size)451 void* ipc_alloc(int size)
452 {
453 	void* out;
454 	if(size > PAGE_SIZE)
455 		out = vmalloc(size);
456 	else
457 		out = kmalloc(size, GFP_KERNEL);
458 	return out;
459 }
460 
461 /**
462  *	ipc_free        -       free ipc space
463  *	@ptr: pointer returned by ipc_alloc
464  *	@size: size of block
465  *
466  *	Free a block created with ipc_alloc(). The caller must know the size
467  *	used in the allocation call.
468  */
469 
ipc_free(void * ptr,int size)470 void ipc_free(void* ptr, int size)
471 {
472 	if(size > PAGE_SIZE)
473 		vfree(ptr);
474 	else
475 		kfree(ptr);
476 }
477 
478 /*
479  * rcu allocations:
480  * There are three headers that are prepended to the actual allocation:
481  * - during use: ipc_rcu_hdr.
482  * - during the rcu grace period: ipc_rcu_grace.
483  * - [only if vmalloc]: ipc_rcu_sched.
484  * Their lifetime doesn't overlap, thus the headers share the same memory.
485  * Unlike a normal union, they are right-aligned, thus some container_of
486  * forward/backward casting is necessary:
487  */
488 struct ipc_rcu_hdr
489 {
490 	int refcount;
491 	int is_vmalloc;
492 	void *data[0];
493 };
494 
495 
496 struct ipc_rcu_grace
497 {
498 	struct rcu_head rcu;
499 	/* "void *" makes sure alignment of following data is sane. */
500 	void *data[0];
501 };
502 
503 struct ipc_rcu_sched
504 {
505 	struct work_struct work;
506 	/* "void *" makes sure alignment of following data is sane. */
507 	void *data[0];
508 };
509 
510 #define HDRLEN_KMALLOC		(sizeof(struct ipc_rcu_grace) > sizeof(struct ipc_rcu_hdr) ? \
511 					sizeof(struct ipc_rcu_grace) : sizeof(struct ipc_rcu_hdr))
512 #define HDRLEN_VMALLOC		(sizeof(struct ipc_rcu_sched) > HDRLEN_KMALLOC ? \
513 					sizeof(struct ipc_rcu_sched) : HDRLEN_KMALLOC)
514 
rcu_use_vmalloc(int size)515 static inline int rcu_use_vmalloc(int size)
516 {
517 	/* Too big for a single page? */
518 	if (HDRLEN_KMALLOC + size > PAGE_SIZE)
519 		return 1;
520 	return 0;
521 }
522 
523 /**
524  *	ipc_rcu_alloc	-	allocate ipc and rcu space
525  *	@size: size desired
526  *
527  *	Allocate memory for the rcu header structure +  the object.
528  *	Returns the pointer to the object.
529  *	NULL is returned if the allocation fails.
530  */
531 
ipc_rcu_alloc(int size)532 void* ipc_rcu_alloc(int size)
533 {
534 	void* out;
535 	/*
536 	 * We prepend the allocation with the rcu struct, and
537 	 * workqueue if necessary (for vmalloc).
538 	 */
539 	if (rcu_use_vmalloc(size)) {
540 		out = vmalloc(HDRLEN_VMALLOC + size);
541 		if (out) {
542 			out += HDRLEN_VMALLOC;
543 			container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 1;
544 			container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
545 		}
546 	} else {
547 		out = kmalloc(HDRLEN_KMALLOC + size, GFP_KERNEL);
548 		if (out) {
549 			out += HDRLEN_KMALLOC;
550 			container_of(out, struct ipc_rcu_hdr, data)->is_vmalloc = 0;
551 			container_of(out, struct ipc_rcu_hdr, data)->refcount = 1;
552 		}
553 	}
554 
555 	return out;
556 }
557 
ipc_rcu_getref(void * ptr)558 void ipc_rcu_getref(void *ptr)
559 {
560 	container_of(ptr, struct ipc_rcu_hdr, data)->refcount++;
561 }
562 
ipc_do_vfree(struct work_struct * work)563 static void ipc_do_vfree(struct work_struct *work)
564 {
565 	vfree(container_of(work, struct ipc_rcu_sched, work));
566 }
567 
568 /**
569  * ipc_schedule_free - free ipc + rcu space
570  * @head: RCU callback structure for queued work
571  *
572  * Since RCU callback function is called in bh,
573  * we need to defer the vfree to schedule_work().
574  */
ipc_schedule_free(struct rcu_head * head)575 static void ipc_schedule_free(struct rcu_head *head)
576 {
577 	struct ipc_rcu_grace *grace;
578 	struct ipc_rcu_sched *sched;
579 
580 	grace = container_of(head, struct ipc_rcu_grace, rcu);
581 	sched = container_of(&(grace->data[0]), struct ipc_rcu_sched,
582 				data[0]);
583 
584 	INIT_WORK(&sched->work, ipc_do_vfree);
585 	schedule_work(&sched->work);
586 }
587 
588 /**
589  * ipc_immediate_free - free ipc + rcu space
590  * @head: RCU callback structure that contains pointer to be freed
591  *
592  * Free from the RCU callback context.
593  */
ipc_immediate_free(struct rcu_head * head)594 static void ipc_immediate_free(struct rcu_head *head)
595 {
596 	struct ipc_rcu_grace *free =
597 		container_of(head, struct ipc_rcu_grace, rcu);
598 	kfree(free);
599 }
600 
ipc_rcu_putref(void * ptr)601 void ipc_rcu_putref(void *ptr)
602 {
603 	if (--container_of(ptr, struct ipc_rcu_hdr, data)->refcount > 0)
604 		return;
605 
606 	if (container_of(ptr, struct ipc_rcu_hdr, data)->is_vmalloc) {
607 		call_rcu(&container_of(ptr, struct ipc_rcu_grace, data)->rcu,
608 				ipc_schedule_free);
609 	} else {
610 		call_rcu(&container_of(ptr, struct ipc_rcu_grace, data)->rcu,
611 				ipc_immediate_free);
612 	}
613 }
614 
615 /**
616  *	ipcperms	-	check IPC permissions
617  *	@ipcp: IPC permission set
618  *	@flag: desired permission set.
619  *
620  *	Check user, group, other permissions for access
621  *	to ipc resources. return 0 if allowed
622  */
623 
ipcperms(struct kern_ipc_perm * ipcp,short flag)624 int ipcperms (struct kern_ipc_perm *ipcp, short flag)
625 {	/* flag will most probably be 0 or S_...UGO from <linux/stat.h> */
626 	uid_t euid = current_euid();
627 	int requested_mode, granted_mode;
628 
629 	audit_ipc_obj(ipcp);
630 	requested_mode = (flag >> 6) | (flag >> 3) | flag;
631 	granted_mode = ipcp->mode;
632 	if (euid == ipcp->cuid ||
633 	    euid == ipcp->uid)
634 		granted_mode >>= 6;
635 	else if (in_group_p(ipcp->cgid) || in_group_p(ipcp->gid))
636 		granted_mode >>= 3;
637 	/* is there some bit set in requested_mode but not in granted_mode? */
638 	if ((requested_mode & ~granted_mode & 0007) &&
639 	    !capable(CAP_IPC_OWNER))
640 		return -1;
641 
642 	return security_ipc_permission(ipcp, flag);
643 }
644 
645 /*
646  * Functions to convert between the kern_ipc_perm structure and the
647  * old/new ipc_perm structures
648  */
649 
650 /**
651  *	kernel_to_ipc64_perm	-	convert kernel ipc permissions to user
652  *	@in: kernel permissions
653  *	@out: new style IPC permissions
654  *
655  *	Turn the kernel object @in into a set of permissions descriptions
656  *	for returning to userspace (@out).
657  */
658 
659 
kernel_to_ipc64_perm(struct kern_ipc_perm * in,struct ipc64_perm * out)660 void kernel_to_ipc64_perm (struct kern_ipc_perm *in, struct ipc64_perm *out)
661 {
662 	out->key	= in->key;
663 	out->uid	= in->uid;
664 	out->gid	= in->gid;
665 	out->cuid	= in->cuid;
666 	out->cgid	= in->cgid;
667 	out->mode	= in->mode;
668 	out->seq	= in->seq;
669 }
670 
671 /**
672  *	ipc64_perm_to_ipc_perm	-	convert new ipc permissions to old
673  *	@in: new style IPC permissions
674  *	@out: old style IPC permissions
675  *
676  *	Turn the new style permissions object @in into a compatibility
677  *	object and store it into the @out pointer.
678  */
679 
ipc64_perm_to_ipc_perm(struct ipc64_perm * in,struct ipc_perm * out)680 void ipc64_perm_to_ipc_perm (struct ipc64_perm *in, struct ipc_perm *out)
681 {
682 	out->key	= in->key;
683 	SET_UID(out->uid, in->uid);
684 	SET_GID(out->gid, in->gid);
685 	SET_UID(out->cuid, in->cuid);
686 	SET_GID(out->cgid, in->cgid);
687 	out->mode	= in->mode;
688 	out->seq	= in->seq;
689 }
690 
691 /**
692  * ipc_lock - Lock an ipc structure without rw_mutex held
693  * @ids: IPC identifier set
694  * @id: ipc id to look for
695  *
696  * Look for an id in the ipc ids idr and lock the associated ipc object.
697  *
698  * The ipc object is locked on exit.
699  */
700 
ipc_lock(struct ipc_ids * ids,int id)701 struct kern_ipc_perm *ipc_lock(struct ipc_ids *ids, int id)
702 {
703 	struct kern_ipc_perm *out;
704 	int lid = ipcid_to_idx(id);
705 
706 	rcu_read_lock();
707 	out = idr_find(&ids->ipcs_idr, lid);
708 	if (out == NULL) {
709 		rcu_read_unlock();
710 		return ERR_PTR(-EINVAL);
711 	}
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 		spin_unlock(&out->lock);
720 		rcu_read_unlock();
721 		return ERR_PTR(-EINVAL);
722 	}
723 
724 	return out;
725 }
726 
ipc_lock_check(struct ipc_ids * ids,int id)727 struct kern_ipc_perm *ipc_lock_check(struct ipc_ids *ids, int id)
728 {
729 	struct kern_ipc_perm *out;
730 
731 	out = ipc_lock(ids, id);
732 	if (IS_ERR(out))
733 		return out;
734 
735 	if (ipc_checkid(out, id)) {
736 		ipc_unlock(out);
737 		return ERR_PTR(-EIDRM);
738 	}
739 
740 	return out;
741 }
742 
743 /**
744  * ipcget - Common sys_*get() code
745  * @ns : namsepace
746  * @ids : IPC identifier set
747  * @ops : operations to be called on ipc object creation, permission checks
748  *        and further checks
749  * @params : the parameters needed by the previous operations.
750  *
751  * Common routine called by sys_msgget(), sys_semget() and sys_shmget().
752  */
ipcget(struct ipc_namespace * ns,struct ipc_ids * ids,struct ipc_ops * ops,struct ipc_params * params)753 int ipcget(struct ipc_namespace *ns, struct ipc_ids *ids,
754 			struct ipc_ops *ops, struct ipc_params *params)
755 {
756 	if (params->key == IPC_PRIVATE)
757 		return ipcget_new(ns, ids, ops, params);
758 	else
759 		return ipcget_public(ns, ids, ops, params);
760 }
761 
762 /**
763  * ipc_update_perm - update the permissions of an IPC.
764  * @in:  the permission given as input.
765  * @out: the permission of the ipc to set.
766  */
ipc_update_perm(struct ipc64_perm * in,struct kern_ipc_perm * out)767 void ipc_update_perm(struct ipc64_perm *in, struct kern_ipc_perm *out)
768 {
769 	out->uid = in->uid;
770 	out->gid = in->gid;
771 	out->mode = (out->mode & ~S_IRWXUGO)
772 		| (in->mode & S_IRWXUGO);
773 }
774 
775 /**
776  * ipcctl_pre_down - retrieve an ipc and check permissions for some IPC_XXX cmd
777  * @ids:  the table of ids where to look for the ipc
778  * @id:   the id of the ipc to retrieve
779  * @cmd:  the cmd to check
780  * @perm: the permission to set
781  * @extra_perm: one extra permission parameter used by msq
782  *
783  * This function does some common audit and permissions check for some IPC_XXX
784  * cmd and is called from semctl_down, shmctl_down and msgctl_down.
785  * It must be called without any lock held and
786  *  - retrieves the ipc with the given id in the given table.
787  *  - performs some audit and permission check, depending on the given cmd
788  *  - returns the ipc with both ipc and rw_mutex locks held in case of success
789  *    or an err-code without any lock held otherwise.
790  */
ipcctl_pre_down(struct ipc_ids * ids,int id,int cmd,struct ipc64_perm * perm,int extra_perm)791 struct kern_ipc_perm *ipcctl_pre_down(struct ipc_ids *ids, int id, int cmd,
792 				      struct ipc64_perm *perm, int extra_perm)
793 {
794 	struct kern_ipc_perm *ipcp;
795 	uid_t euid;
796 	int err;
797 
798 	down_write(&ids->rw_mutex);
799 	ipcp = ipc_lock_check(ids, id);
800 	if (IS_ERR(ipcp)) {
801 		err = PTR_ERR(ipcp);
802 		goto out_up;
803 	}
804 
805 	audit_ipc_obj(ipcp);
806 	if (cmd == IPC_SET)
807 		audit_ipc_set_perm(extra_perm, perm->uid,
808 					 perm->gid, perm->mode);
809 
810 	euid = current_euid();
811 	if (euid == ipcp->cuid ||
812 	    euid == ipcp->uid  || capable(CAP_SYS_ADMIN))
813 		return ipcp;
814 
815 	err = -EPERM;
816 	ipc_unlock(ipcp);
817 out_up:
818 	up_write(&ids->rw_mutex);
819 	return ERR_PTR(err);
820 }
821 
822 #ifdef __ARCH_WANT_IPC_PARSE_VERSION
823 
824 
825 /**
826  *	ipc_parse_version	-	IPC call version
827  *	@cmd: pointer to command
828  *
829  *	Return IPC_64 for new style IPC and IPC_OLD for old style IPC.
830  *	The @cmd value is turned from an encoding command and version into
831  *	just the command code.
832  */
833 
ipc_parse_version(int * cmd)834 int ipc_parse_version (int *cmd)
835 {
836 	if (*cmd & IPC_64) {
837 		*cmd ^= IPC_64;
838 		return IPC_64;
839 	} else {
840 		return IPC_OLD;
841 	}
842 }
843 
844 #endif /* __ARCH_WANT_IPC_PARSE_VERSION */
845 
846 #ifdef CONFIG_PROC_FS
847 struct ipc_proc_iter {
848 	struct ipc_namespace *ns;
849 	struct ipc_proc_iface *iface;
850 };
851 
852 /*
853  * This routine locks the ipc structure found at least at position pos.
854  */
sysvipc_find_ipc(struct ipc_ids * ids,loff_t pos,loff_t * new_pos)855 static struct kern_ipc_perm *sysvipc_find_ipc(struct ipc_ids *ids, loff_t pos,
856 					      loff_t *new_pos)
857 {
858 	struct kern_ipc_perm *ipc;
859 	int total, id;
860 
861 	total = 0;
862 	for (id = 0; id < pos && total < ids->in_use; id++) {
863 		ipc = idr_find(&ids->ipcs_idr, id);
864 		if (ipc != NULL)
865 			total++;
866 	}
867 
868 	if (total >= ids->in_use)
869 		return NULL;
870 
871 	for ( ; pos < IPCMNI; pos++) {
872 		ipc = idr_find(&ids->ipcs_idr, pos);
873 		if (ipc != NULL) {
874 			*new_pos = pos + 1;
875 			ipc_lock_by_ptr(ipc);
876 			return ipc;
877 		}
878 	}
879 
880 	/* Out of range - return NULL to terminate iteration */
881 	return NULL;
882 }
883 
sysvipc_proc_next(struct seq_file * s,void * it,loff_t * pos)884 static void *sysvipc_proc_next(struct seq_file *s, void *it, loff_t *pos)
885 {
886 	struct ipc_proc_iter *iter = s->private;
887 	struct ipc_proc_iface *iface = iter->iface;
888 	struct kern_ipc_perm *ipc = it;
889 
890 	/* If we had an ipc id locked before, unlock it */
891 	if (ipc && ipc != SEQ_START_TOKEN)
892 		ipc_unlock(ipc);
893 
894 	return sysvipc_find_ipc(&iter->ns->ids[iface->ids], *pos, pos);
895 }
896 
897 /*
898  * File positions: pos 0 -> header, pos n -> ipc id = n - 1.
899  * SeqFile iterator: iterator value locked ipc pointer or SEQ_TOKEN_START.
900  */
sysvipc_proc_start(struct seq_file * s,loff_t * pos)901 static void *sysvipc_proc_start(struct seq_file *s, loff_t *pos)
902 {
903 	struct ipc_proc_iter *iter = s->private;
904 	struct ipc_proc_iface *iface = iter->iface;
905 	struct ipc_ids *ids;
906 
907 	ids = &iter->ns->ids[iface->ids];
908 
909 	/*
910 	 * Take the lock - this will be released by the corresponding
911 	 * call to stop().
912 	 */
913 	down_read(&ids->rw_mutex);
914 
915 	/* pos < 0 is invalid */
916 	if (*pos < 0)
917 		return NULL;
918 
919 	/* pos == 0 means header */
920 	if (*pos == 0)
921 		return SEQ_START_TOKEN;
922 
923 	/* Find the (pos-1)th ipc */
924 	return sysvipc_find_ipc(ids, *pos - 1, pos);
925 }
926 
sysvipc_proc_stop(struct seq_file * s,void * it)927 static void sysvipc_proc_stop(struct seq_file *s, void *it)
928 {
929 	struct kern_ipc_perm *ipc = it;
930 	struct ipc_proc_iter *iter = s->private;
931 	struct ipc_proc_iface *iface = iter->iface;
932 	struct ipc_ids *ids;
933 
934 	/* If we had a locked structure, release it */
935 	if (ipc && ipc != SEQ_START_TOKEN)
936 		ipc_unlock(ipc);
937 
938 	ids = &iter->ns->ids[iface->ids];
939 	/* Release the lock we took in start() */
940 	up_read(&ids->rw_mutex);
941 }
942 
sysvipc_proc_show(struct seq_file * s,void * it)943 static int sysvipc_proc_show(struct seq_file *s, void *it)
944 {
945 	struct ipc_proc_iter *iter = s->private;
946 	struct ipc_proc_iface *iface = iter->iface;
947 
948 	if (it == SEQ_START_TOKEN)
949 		return seq_puts(s, iface->header);
950 
951 	return iface->show(s, it);
952 }
953 
954 static struct seq_operations sysvipc_proc_seqops = {
955 	.start = sysvipc_proc_start,
956 	.stop  = sysvipc_proc_stop,
957 	.next  = sysvipc_proc_next,
958 	.show  = sysvipc_proc_show,
959 };
960 
sysvipc_proc_open(struct inode * inode,struct file * file)961 static int sysvipc_proc_open(struct inode *inode, struct file *file)
962 {
963 	int ret;
964 	struct seq_file *seq;
965 	struct ipc_proc_iter *iter;
966 
967 	ret = -ENOMEM;
968 	iter = kmalloc(sizeof(*iter), GFP_KERNEL);
969 	if (!iter)
970 		goto out;
971 
972 	ret = seq_open(file, &sysvipc_proc_seqops);
973 	if (ret)
974 		goto out_kfree;
975 
976 	seq = file->private_data;
977 	seq->private = iter;
978 
979 	iter->iface = PDE(inode)->data;
980 	iter->ns    = get_ipc_ns(current->nsproxy->ipc_ns);
981 out:
982 	return ret;
983 out_kfree:
984 	kfree(iter);
985 	goto out;
986 }
987 
sysvipc_proc_release(struct inode * inode,struct file * file)988 static int sysvipc_proc_release(struct inode *inode, struct file *file)
989 {
990 	struct seq_file *seq = file->private_data;
991 	struct ipc_proc_iter *iter = seq->private;
992 	put_ipc_ns(iter->ns);
993 	return seq_release_private(inode, file);
994 }
995 
996 static const struct file_operations sysvipc_proc_fops = {
997 	.open    = sysvipc_proc_open,
998 	.read    = seq_read,
999 	.llseek  = seq_lseek,
1000 	.release = sysvipc_proc_release,
1001 };
1002 #endif /* CONFIG_PROC_FS */
1003