1/*
2 * Copyright (c) 2000-2007 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28/*
29 *	Copyright (c) 1990, 1996-1998 Apple Computer, Inc.
30 *	All Rights Reserved.
31 */
32/*
33 * posix_shm.c : Support for POSIX shared memory APIs
34 *
35 *	File:	posix_shm.c
36 *	Author:	Ananthakrishna Ramesh
37 *
38 * HISTORY
39 * 2-Sep-1999	A.Ramesh
40 *	Created for MacOSX
41 *
42 */
43/*
44 * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce
45 * support for mandatory and extensible security protections.  This notice
46 * is included in support of clause 2.2 (b) of the Apple Public License,
47 * Version 2.0.
48 */
49
50#include <sys/cdefs.h>
51#include <sys/param.h>
52#include <sys/systm.h>
53#include <sys/kernel.h>
54#include <sys/file_internal.h>
55#include <sys/filedesc.h>
56#include <sys/stat.h>
57#include <sys/proc_internal.h>
58#include <sys/kauth.h>
59#include <sys/mount.h>
60#include <sys/namei.h>
61#include <sys/vnode.h>
62#include <sys/vnode_internal.h>
63#include <sys/ioctl.h>
64#include <sys/tty.h>
65#include <sys/malloc.h>
66#include <sys/mman.h>
67#include <sys/stat.h>
68#include <sys/sysproto.h>
69#include <sys/proc_info.h>
70#include <security/audit/audit.h>
71
72#if CONFIG_MACF
73#include <security/mac_framework.h>
74#endif
75
76#include <mach/mach_types.h>
77#include <mach/mach_vm.h>
78#include <mach/vm_map.h>
79#include <mach/vm_prot.h>
80#include <mach/vm_inherit.h>
81#include <mach/kern_return.h>
82#include <mach/memory_object_control.h>
83
84#include <vm/vm_map.h>
85#include <vm/vm_protos.h>
86
87#define f_flag f_fglob->fg_flag
88#define f_type f_fglob->fg_ops->fo_type
89#define f_msgcount f_fglob->fg_msgcount
90#define f_cred f_fglob->fg_cred
91#define f_ops f_fglob->fg_ops
92#define f_offset f_fglob->fg_offset
93#define f_data f_fglob->fg_data
94#define	PSHMNAMLEN	31	/* maximum name segment length we bother with */
95
96struct pshmobj {
97	void *			pshmo_memobject;
98	memory_object_size_t	pshmo_size;
99	struct pshmobj *	pshmo_next;
100};
101
102struct pshminfo {
103	unsigned int	pshm_flags;
104	unsigned int	pshm_usecount;
105	off_t		pshm_length;
106	mode_t		pshm_mode;
107	uid_t		pshm_uid;
108	gid_t		pshm_gid;
109	char		pshm_name[PSHMNAMLEN + 1];	/* segment name */
110	struct pshmobj *pshm_memobjects;
111#if DIAGNOSTIC
112	unsigned int 	pshm_readcount;
113	unsigned int 	pshm_writecount;
114	proc_t		pshm_proc;
115#endif /* DIAGNOSTIC */
116	struct label*	pshm_label;
117};
118#define PSHMINFO_NULL (struct pshminfo *)0
119
120#define	PSHM_NONE	0x001
121#define	PSHM_DEFINED	0x002
122#define	PSHM_ALLOCATED	0x004
123#define	PSHM_MAPPED	0x008
124#define	PSHM_INUSE	0x010
125#define	PSHM_REMOVED	0x020
126#define	PSHM_INCREATE	0x040
127#define	PSHM_INDELETE	0x080
128#define PSHM_ALLOCATING	0x100
129
130struct	pshmcache {
131	LIST_ENTRY(pshmcache) pshm_hash;	/* hash chain */
132	struct	pshminfo *pshminfo;		/* vnode the name refers to */
133	int	pshm_nlen;		/* length of name */
134	char	pshm_name[PSHMNAMLEN + 1];	/* segment name */
135};
136#define PSHMCACHE_NULL (struct pshmcache *)0
137
138struct	pshmstats {
139	long	goodhits;		/* hits that we can really use */
140	long	neghits;		/* negative hits that we can use */
141	long	badhits;		/* hits we must drop */
142	long	falsehits;		/* hits with id mismatch */
143	long	miss;		/* misses */
144	long	longnames;		/* long names that ignore cache */
145};
146
147struct pshmname {
148	char	*pshm_nameptr;	/* pointer to looked up name */
149	long	pshm_namelen;	/* length of looked up component */
150	u_long	pshm_hash;	/* hash value of looked up name */
151};
152
153struct pshmnode {
154	off_t  		mapp_addr;
155	user_size_t	map_size;	/* XXX unused ? */
156	struct pshminfo *pinfo;
157	unsigned int	pshm_usecount;
158#if DIAGNOSTIC
159	unsigned int readcnt;
160	unsigned int writecnt;
161#endif
162};
163#define PSHMNODE_NULL (struct pshmnode *)0
164
165
166#define PSHMHASH(pnp) \
167	(&pshmhashtbl[(pnp)->pshm_hash & pshmhash])
168
169LIST_HEAD(pshmhashhead, pshmcache) *pshmhashtbl;	/* Hash Table */
170u_long	pshmhash;				/* size of hash table - 1 */
171long	pshmnument;			/* number of cache entries allocated */
172struct pshmstats pshmstats;		/* cache effectiveness statistics */
173
174static int pshm_read (struct fileproc *fp, struct uio *uio,
175		    int flags, vfs_context_t ctx);
176static int pshm_write (struct fileproc *fp, struct uio *uio,
177		    int flags, vfs_context_t ctx);
178static int pshm_ioctl (struct fileproc *fp, u_long com,
179		    caddr_t data, vfs_context_t ctx);
180static int pshm_select (struct fileproc *fp, int which, void *wql, vfs_context_t ctx);
181static int pshm_close(struct pshminfo *pinfo, int dropref);
182static int pshm_closefile (struct fileglob *fg, vfs_context_t ctx);
183
184static int pshm_kqfilter(struct fileproc *fp, struct knote *kn, vfs_context_t ctx);
185
186int pshm_access(struct pshminfo *pinfo, int mode, kauth_cred_t cred, proc_t p);
187static int pshm_cache_add(struct pshminfo *pshmp, struct pshmname *pnp, struct pshmcache *pcp);
188static void pshm_cache_delete(struct pshmcache *pcp);
189#if NOT_USED
190static void pshm_cache_purge(void);
191#endif	/* NOT_USED */
192static int pshm_cache_search(struct pshminfo **pshmp, struct pshmname *pnp,
193	struct pshmcache **pcache, int addref);
194
195static const struct fileops pshmops = {
196	DTYPE_PSXSHM,
197	pshm_read,
198	pshm_write,
199	pshm_ioctl,
200	pshm_select,
201	pshm_closefile,
202	pshm_kqfilter,
203	0
204};
205
206static lck_grp_t       *psx_shm_subsys_lck_grp;
207static lck_grp_attr_t  *psx_shm_subsys_lck_grp_attr;
208static lck_attr_t      *psx_shm_subsys_lck_attr;
209static lck_mtx_t        psx_shm_subsys_mutex;
210
211#define PSHM_SUBSYS_LOCK() lck_mtx_lock(& psx_shm_subsys_mutex)
212#define PSHM_SUBSYS_UNLOCK() lck_mtx_unlock(& psx_shm_subsys_mutex)
213
214
215/* Initialize the mutex governing access to the posix shm subsystem */
216__private_extern__ void
217pshm_lock_init( void )
218{
219
220    psx_shm_subsys_lck_grp_attr = lck_grp_attr_alloc_init();
221
222    psx_shm_subsys_lck_grp = lck_grp_alloc_init("posix shared memory", psx_shm_subsys_lck_grp_attr);
223
224    psx_shm_subsys_lck_attr = lck_attr_alloc_init();
225    lck_mtx_init(& psx_shm_subsys_mutex, psx_shm_subsys_lck_grp, psx_shm_subsys_lck_attr);
226}
227
228/*
229 * Lookup an entry in the cache
230 *
231 *
232 * status of -1 is returned if matches
233 * If the lookup determines that the name does not exist
234 * (negative cacheing), a status of ENOENT is returned. If the lookup
235 * fails, a status of zero is returned.
236 */
237
238static int
239pshm_cache_search(struct pshminfo **pshmp, struct pshmname *pnp,
240	struct pshmcache **pcache, int addref)
241{
242	struct pshmcache *pcp, *nnp;
243	struct pshmhashhead *pcpp;
244
245	if (pnp->pshm_namelen > PSHMNAMLEN) {
246		pshmstats.longnames++;
247		return (0);
248	}
249
250	pcpp = PSHMHASH(pnp);
251	for (pcp = pcpp->lh_first; pcp != 0; pcp = nnp) {
252		nnp = pcp->pshm_hash.le_next;
253		if (pcp->pshm_nlen == pnp->pshm_namelen &&
254		    !bcmp(pcp->pshm_name, pnp->pshm_nameptr, 						(u_int)pcp-> pshm_nlen))
255			break;
256	}
257
258	if (pcp == 0) {
259		pshmstats.miss++;
260		return (0);
261	}
262
263	/* We found a "positive" match, return the vnode */
264        if (pcp->pshminfo) {
265		pshmstats.goodhits++;
266		/* TOUCH(ncp); */
267		*pshmp = pcp->pshminfo;
268		*pcache = pcp;
269		if (addref)
270			pcp->pshminfo->pshm_usecount++;
271		return (-1);
272	}
273
274	/*
275	 * We found a "negative" match, ENOENT notifies client of this match.
276	 */
277	pshmstats.neghits++;
278	return (ENOENT);
279}
280
281/*
282 * Add an entry to the cache.
283 * XXX should be static?
284 */
285static int
286pshm_cache_add(struct pshminfo *pshmp, struct pshmname *pnp, struct pshmcache *pcp)
287{
288	struct pshmhashhead *pcpp;
289	struct pshminfo *dpinfo;
290	struct pshmcache *dpcp;
291
292#if DIAGNOSTIC
293	if (pnp->pshm_namelen > PSHMNAMLEN)
294		panic("cache_enter: name too long");
295#endif
296
297
298	/*  if the entry has already been added by some one else return */
299	if (pshm_cache_search(&dpinfo, pnp, &dpcp, 0) == -1) {
300		return(EEXIST);
301	}
302	pshmnument++;
303
304	/*
305	 * Fill in cache info, if vp is NULL this is a "negative" cache entry.
306	 */
307	pcp->pshminfo = pshmp;
308	pcp->pshm_nlen = pnp->pshm_namelen;
309	bcopy(pnp->pshm_nameptr, pcp->pshm_name, (unsigned)pcp->pshm_nlen);
310	pcpp = PSHMHASH(pnp);
311#if DIAGNOSTIC
312	{
313		struct pshmcache *p;
314
315		for (p = pcpp->lh_first; p != 0; p = p->pshm_hash.le_next)
316			if (p == pcp)
317				panic("cache_enter: duplicate");
318	}
319#endif
320	LIST_INSERT_HEAD(pcpp, pcp, pshm_hash);
321	return(0);
322}
323
324/*
325 * Name cache initialization, from vfs_init() when we are booting
326 */
327void
328pshm_cache_init(void)
329{
330	pshmhashtbl = hashinit(desiredvnodes / 8, M_SHM, &pshmhash);
331}
332
333#if NOT_USED
334/*
335 * Invalidate a all entries to particular vnode.
336 *
337 * We actually just increment the v_id, that will do it. The entries will
338 * be purged by lookup as they get found. If the v_id wraps around, we
339 * need to ditch the entire cache, to avoid confusion. No valid vnode will
340 * ever have (v_id == 0).
341 */
342static void
343pshm_cache_purge(void)
344{
345	struct pshmcache *pcp;
346	struct pshmhashhead *pcpp;
347
348	for (pcpp = &pshmhashtbl[pshmhash]; pcpp >= pshmhashtbl; pcpp--) {
349		while ( (pcp = pcpp->lh_first) )
350			pshm_cache_delete(pcp);
351	}
352}
353#endif	/* NOT_USED */
354
355static void
356pshm_cache_delete(struct pshmcache *pcp)
357{
358#if DIAGNOSTIC
359	if (pcp->pshm_hash.le_prev == 0)
360		panic("namecache purge le_prev");
361	if (pcp->pshm_hash.le_next == pcp)
362		panic("namecache purge le_next");
363#endif /* DIAGNOSTIC */
364	LIST_REMOVE(pcp, pshm_hash);
365	pcp->pshm_hash.le_prev = 0;
366	pshmnument--;
367}
368
369
370int
371shm_open(proc_t p, struct shm_open_args *uap, int32_t *retval)
372{
373	size_t  i;
374	int indx, error;
375	struct pshmname nd;
376	struct pshminfo *pinfo;
377	struct fileproc *fp = NULL;
378	char *pnbuf = NULL;
379	struct pshminfo *new_pinfo = PSHMINFO_NULL;
380	struct pshmnode *new_pnode = PSHMNODE_NULL;
381	struct pshmcache *pcache = PSHMCACHE_NULL;	/* ignored on return */
382	char * nameptr;
383	char * cp;
384	size_t pathlen, plen;
385	int fmode ;
386	int cmode = uap->mode;
387	int incache = 0;
388	struct pshmcache *pcp = NULL;
389
390	AUDIT_ARG(fflags, uap->oflag);
391	AUDIT_ARG(mode, uap->mode);
392
393	pinfo = PSHMINFO_NULL;
394
395	/*
396	 * Preallocate everything we might need up front to avoid taking
397	 * and dropping the lock, opening us up to race conditions.
398	 */
399	MALLOC_ZONE(pnbuf, caddr_t, MAXPATHLEN, M_NAMEI, M_WAITOK);
400	if (pnbuf == NULL) {
401		error = ENOSPC;
402		goto bad;
403	}
404
405	pathlen = MAXPATHLEN;
406	error = copyinstr(uap->name, (void *)pnbuf, MAXPATHLEN, &pathlen);
407	if (error) {
408		goto bad;
409	}
410	AUDIT_ARG(text, pnbuf);
411	if (pathlen > PSHMNAMLEN) {
412		error = ENAMETOOLONG;
413		goto bad;
414	}
415#ifdef PSXSHM_NAME_RESTRICT
416	nameptr = pnbuf;
417	if (*nameptr == '/') {
418		while (*(nameptr++) == '/') {
419			plen--;
420			error = EINVAL;
421			goto bad;
422		}
423	} else {
424		error = EINVAL;
425		goto bad;
426	}
427#endif /* PSXSHM_NAME_RESTRICT */
428
429	plen = pathlen;
430	nameptr = pnbuf;
431	nd.pshm_nameptr = nameptr;
432	nd.pshm_namelen = plen;
433	nd. pshm_hash =0;
434
435	for (cp = nameptr, i=1; *cp != 0 && i <= plen; i++, cp++) {
436		nd.pshm_hash += (unsigned char)*cp * i;
437	}
438
439	/*
440	 * attempt to allocate a new fp; if unsuccessful, the fp will be
441	 * left unmodified (NULL).
442	 */
443	error = falloc(p, &fp, &indx, vfs_context_current());
444	if (error)
445		goto bad;
446
447	cmode &=  ALLPERMS;
448
449	fmode = FFLAGS(uap->oflag);
450	if ((fmode & (FREAD | FWRITE)) == 0) {
451		error = EINVAL;
452		goto bad;
453	}
454
455	/*
456	 * We allocate a new entry if we are less than the maximum
457	 * allowed and the one at the front of the LRU list is in use.
458	 * Otherwise we use the one at the front of the LRU list.
459	 */
460	MALLOC(pcp, struct pshmcache *, sizeof(struct pshmcache), M_SHM, M_WAITOK|M_ZERO);
461	if (pcp == NULL) {
462		error = ENOSPC;
463		goto bad;
464	}
465
466	MALLOC(new_pinfo, struct pshminfo *, sizeof(struct pshminfo), M_SHM, M_WAITOK|M_ZERO);
467	if (new_pinfo == PSHMINFO_NULL) {
468		error = ENOSPC;
469		goto bad;
470	}
471#if CONFIG_MACF
472	mac_posixshm_label_init(new_pinfo);
473#endif
474
475	MALLOC(new_pnode, struct pshmnode *, sizeof(struct pshmnode), M_SHM, M_WAITOK|M_ZERO);
476	if (new_pnode == PSHMNODE_NULL) {
477		error = ENOSPC;
478		goto bad;
479	}
480
481	PSHM_SUBSYS_LOCK();
482
483	/*
484	 * If we find the entry in the cache, this will take a reference,
485	 * allowing us to unlock it for the permissions check.
486	 */
487	error = pshm_cache_search(&pinfo, &nd, &pcache, 1);
488
489	PSHM_SUBSYS_UNLOCK();
490
491	if (error == ENOENT) {
492		error = EINVAL;
493		goto bad;
494	}
495
496	if (!error) {
497		incache = 0;
498		if (fmode & O_CREAT) {
499			/*  create a new one (commit the allocation) */
500			pinfo = new_pinfo;
501			pinfo->pshm_flags = PSHM_DEFINED | PSHM_INCREATE;
502			pinfo->pshm_usecount = 1; /* existence reference */
503			pinfo->pshm_mode = cmode;
504			pinfo->pshm_uid = kauth_getuid();
505			pinfo->pshm_gid = kauth_getgid();
506			bcopy(pnbuf, &pinfo->pshm_name[0], pathlen);
507			pinfo->pshm_name[pathlen]=0;
508#if CONFIG_MACF
509			error = mac_posixshm_check_create(kauth_cred_get(), nameptr);
510			if (error) {
511				goto bad;
512			}
513			mac_posixshm_label_associate(kauth_cred_get(), pinfo, nameptr);
514#endif
515		}
516	} else {
517		incache = 1;
518		if (fmode & O_CREAT) {
519			/*  already exists */
520			if ((fmode & O_EXCL)) {
521				AUDIT_ARG(posix_ipc_perm, pinfo->pshm_uid,
522						pinfo->pshm_gid,
523						pinfo->pshm_mode);
524
525				/* shm obj exists and opened O_EXCL */
526				error = EEXIST;
527				goto bad;
528			}
529
530			if( pinfo->pshm_flags & PSHM_INDELETE) {
531				error = ENOENT;
532				goto bad;
533			}
534			AUDIT_ARG(posix_ipc_perm, pinfo->pshm_uid,
535					pinfo->pshm_gid, pinfo->pshm_mode);
536#if CONFIG_MACF
537			if ((error = mac_posixshm_check_open(kauth_cred_get(), pinfo, fmode))) {
538				goto bad;
539			}
540#endif
541			if ( (error = pshm_access(pinfo, fmode, kauth_cred_get(), p)) ) {
542				goto bad;
543			}
544		}
545	}
546	if (!(fmode & O_CREAT)) {
547		if (!incache) {
548			/* O_CREAT is not set and the object does not exist */
549			error = ENOENT;
550			goto bad;
551		}
552		if( pinfo->pshm_flags & PSHM_INDELETE) {
553			error = ENOENT;
554			goto bad;
555		}
556#if CONFIG_MACF
557		if ((error = mac_posixshm_check_open(kauth_cred_get(), pinfo, fmode))) {
558			goto bad;
559		}
560#endif
561
562		if ((error = pshm_access(pinfo, fmode, kauth_cred_get(), p))) {
563			goto bad;
564		}
565	}
566	if (fmode & O_TRUNC) {
567		error = EINVAL;
568		goto bad;
569	}
570
571
572	PSHM_SUBSYS_LOCK();
573
574#if DIAGNOSTIC
575	if (fmode & FWRITE)
576		pinfo->pshm_writecount++;
577	if (fmode & FREAD)
578		pinfo->pshm_readcount++;
579#endif
580	if (!incache) {
581		/* if successful, this will consume the pcp */
582		if ( (error = pshm_cache_add(pinfo, &nd, pcp)) ) {
583			goto bad_locked;
584		}
585		/*
586		 * add reference for the new entry; otherwise, we obtained
587		 * one from the cache hit earlier.
588		 */
589		pinfo->pshm_usecount++;
590	}
591	pinfo->pshm_flags &= ~PSHM_INCREATE;
592	new_pnode->pinfo = pinfo;
593
594	PSHM_SUBSYS_UNLOCK();
595
596	/*
597	 * if incache, we did not use the new pcp or new_pinfo and must
598	 * free them
599	 */
600	if (incache) {
601		FREE(pcp, M_SHM);
602
603		if (new_pinfo != PSHMINFO_NULL) {
604#if CONFIG_MACF
605			mac_posixshm_label_destroy(new_pinfo);
606#endif
607			FREE(new_pinfo, M_SHM);
608		}
609	}
610
611	proc_fdlock(p);
612	fp->f_flag = fmode & FMASK;
613	fp->f_ops = &pshmops;
614	fp->f_data = (caddr_t)new_pnode;
615	*fdflags(p, indx) |= UF_EXCLOSE;
616	procfdtbl_releasefd(p, indx, NULL);
617	fp_drop(p, indx, fp, 1);
618	proc_fdunlock(p);
619
620	*retval = indx;
621	FREE_ZONE(pnbuf, MAXPATHLEN, M_NAMEI);
622	return (0);
623
624bad_locked:
625	PSHM_SUBSYS_UNLOCK();
626bad:
627	/*
628	 * If we obtained the entry from the cache, we need to drop the
629	 * reference; holding the reference may have prevented unlinking,
630	 * so we need to call pshm_close() to get the full effect.
631	 */
632	if (incache) {
633		PSHM_SUBSYS_LOCK();
634		pshm_close(pinfo, 1);
635		PSHM_SUBSYS_UNLOCK();
636	}
637
638	if (pcp != NULL)
639		FREE(pcp, M_SHM);
640
641	if (new_pnode != PSHMNODE_NULL)
642		FREE(new_pnode, M_SHM);
643
644	if (fp != NULL)
645		fp_free(p, indx, fp);
646
647	if (new_pinfo != PSHMINFO_NULL) {
648#if CONFIG_MACF
649		mac_posixshm_label_destroy(new_pinfo);
650#endif
651		FREE(new_pinfo, M_SHM);
652	}
653	if (pnbuf != NULL)
654		FREE_ZONE(pnbuf, MAXPATHLEN, M_NAMEI);
655	return (error);
656}
657
658
659int
660pshm_truncate(__unused proc_t p, struct fileproc *fp, __unused int fd,
661				off_t length, __unused int32_t *retval)
662{
663	struct pshminfo * pinfo;
664	struct pshmnode * pnode ;
665	kern_return_t kret;
666	mem_entry_name_port_t mem_object;
667	mach_vm_size_t total_size, alloc_size;
668	memory_object_size_t mosize;
669	struct pshmobj *pshmobj, *pshmobj_next, **pshmobj_next_p;
670	vm_map_t	user_map;
671#if CONFIG_MACF
672	int error;
673#endif
674
675	user_map = current_map();
676
677	if (fp->f_type != DTYPE_PSXSHM) {
678		return(EINVAL);
679	}
680
681
682	if (((pnode = (struct pshmnode *)fp->f_data)) == PSHMNODE_NULL )
683		return(EINVAL);
684
685	PSHM_SUBSYS_LOCK();
686	if ((pinfo = pnode->pinfo) == PSHMINFO_NULL) {
687		PSHM_SUBSYS_UNLOCK();
688		return(EINVAL);
689	}
690	if ((pinfo->pshm_flags & (PSHM_DEFINED|PSHM_ALLOCATING|PSHM_ALLOCATED))
691			!= PSHM_DEFINED) {
692		PSHM_SUBSYS_UNLOCK();
693		return(EINVAL);
694	}
695#if CONFIG_MACF
696	error = mac_posixshm_check_truncate(kauth_cred_get(), pinfo, length);
697	if (error) {
698		PSHM_SUBSYS_UNLOCK();
699		return(error);
700	}
701#endif
702
703	pinfo->pshm_flags |= PSHM_ALLOCATING;
704	total_size = vm_map_round_page(length,
705				       vm_map_page_mask(user_map));
706	pshmobj_next_p = &pinfo->pshm_memobjects;
707
708	for (alloc_size = 0;
709	     alloc_size < total_size;
710	     alloc_size += mosize) {
711
712		PSHM_SUBSYS_UNLOCK();
713
714		mosize = MIN(total_size - alloc_size, ANON_MAX_SIZE);
715		kret = mach_make_memory_entry_64(
716			VM_MAP_NULL,
717			&mosize,
718			0,
719			MAP_MEM_NAMED_CREATE | VM_PROT_DEFAULT,
720			&mem_object,
721			0);
722
723		if (kret != KERN_SUCCESS)
724			goto out;
725
726		MALLOC(pshmobj, struct pshmobj *, sizeof (struct pshmobj),
727		       M_SHM, M_WAITOK);
728		if (pshmobj == NULL) {
729			kret = KERN_NO_SPACE;
730			mach_memory_entry_port_release(mem_object);
731			mem_object = NULL;
732			goto out;
733		}
734
735		PSHM_SUBSYS_LOCK();
736
737		pshmobj->pshmo_memobject = (void *) mem_object;
738		pshmobj->pshmo_size = mosize;
739		pshmobj->pshmo_next = NULL;
740
741		*pshmobj_next_p = pshmobj;
742		pshmobj_next_p = &pshmobj->pshmo_next;
743	}
744
745	pinfo->pshm_flags = PSHM_ALLOCATED;
746	pinfo->pshm_length = total_size;
747	PSHM_SUBSYS_UNLOCK();
748	return(0);
749
750out:
751	PSHM_SUBSYS_LOCK();
752	for (pshmobj = pinfo->pshm_memobjects;
753	     pshmobj != NULL;
754	     pshmobj = pshmobj_next) {
755		pshmobj_next = pshmobj->pshmo_next;
756		mach_memory_entry_port_release(pshmobj->pshmo_memobject);
757		FREE(pshmobj, M_SHM);
758	}
759	pinfo->pshm_memobjects = NULL;
760	pinfo->pshm_flags &= ~PSHM_ALLOCATING;
761	PSHM_SUBSYS_UNLOCK();
762
763	switch (kret) {
764	case KERN_INVALID_ADDRESS:
765	case KERN_NO_SPACE:
766		return (ENOMEM);
767	case KERN_PROTECTION_FAILURE:
768		return (EACCES);
769	default:
770		return (EINVAL);
771
772	}
773}
774
775int
776pshm_stat(struct pshmnode *pnode, void *ub, int isstat64)
777{
778	struct stat *sb = (struct stat *)0;	/* warning avoidance ; protected by isstat64 */
779	struct stat64 * sb64 = (struct stat64 *)0;  /* warning avoidance ; protected by isstat64 */
780	struct pshminfo *pinfo;
781#if CONFIG_MACF
782	int error;
783#endif
784
785	PSHM_SUBSYS_LOCK();
786	if ((pinfo = pnode->pinfo) == PSHMINFO_NULL){
787		PSHM_SUBSYS_UNLOCK();
788		return(EINVAL);
789	}
790
791#if CONFIG_MACF
792	error = mac_posixshm_check_stat(kauth_cred_get(), pinfo);
793	if (error) {
794		PSHM_SUBSYS_UNLOCK();
795		return(error);
796	}
797#endif
798
799	if (isstat64 != 0) {
800		sb64 = (struct stat64 *)ub;
801		bzero(sb64, sizeof(struct stat64));
802		sb64->st_mode = pinfo->pshm_mode;
803		sb64->st_uid = pinfo->pshm_uid;
804		sb64->st_gid = pinfo->pshm_gid;
805		sb64->st_size = pinfo->pshm_length;
806	} else {
807		sb = (struct stat *)ub;
808		bzero(sb, sizeof(struct stat));
809		sb->st_mode = pinfo->pshm_mode;
810		sb->st_uid = pinfo->pshm_uid;
811		sb->st_gid = pinfo->pshm_gid;
812		sb->st_size = pinfo->pshm_length;
813	}
814	PSHM_SUBSYS_UNLOCK();
815
816	return(0);
817}
818
819/*
820 * This is called only from shm_open which holds pshm_lock();
821 * XXX This code is repeated many times
822 */
823int
824pshm_access(struct pshminfo *pinfo, int mode, kauth_cred_t cred, __unused proc_t p)
825{
826	int mode_req = ((mode & FREAD) ? S_IRUSR : 0) |
827		       ((mode & FWRITE) ? S_IWUSR : 0);
828
829	/* Otherwise, user id 0 always gets access. */
830	if (!suser(cred, NULL))
831		return (0);
832
833	return(posix_cred_access(cred, pinfo->pshm_uid, pinfo->pshm_gid, pinfo->pshm_mode, mode_req));
834}
835
836int
837pshm_mmap(__unused proc_t p, struct mmap_args *uap, user_addr_t *retval, struct fileproc *fp, off_t pageoff)
838{
839	vm_map_offset_t	user_addr = (vm_map_offset_t)uap->addr;
840	vm_map_size_t	user_size = (vm_map_size_t)uap->len ;
841	vm_map_offset_t	user_start_addr;
842	vm_map_size_t	map_size, mapped_size;
843	int prot = uap->prot;
844	int flags = uap->flags;
845	vm_object_offset_t file_pos = (vm_object_offset_t)uap->pos;
846	vm_object_offset_t map_pos;
847	vm_map_t	user_map;
848	int		alloc_flags;
849	boolean_t 	docow;
850	kern_return_t	kret;
851	struct pshminfo * pinfo;
852	struct pshmnode * pnode;
853	struct pshmobj * pshmobj;
854#if CONFIG_MACF
855	int error;
856#endif
857
858	if (user_size == 0)
859		return(0);
860
861	if ((flags & MAP_SHARED) == 0)
862		return(EINVAL);
863
864
865	if ((prot & PROT_WRITE) && ((fp->f_flag & FWRITE) == 0)) {
866		return(EPERM);
867	}
868
869	if (((pnode = (struct pshmnode *)fp->f_data)) == PSHMNODE_NULL )
870		return(EINVAL);
871
872	PSHM_SUBSYS_LOCK();
873	if ((pinfo = pnode->pinfo) == PSHMINFO_NULL) {
874		PSHM_SUBSYS_UNLOCK();
875		return(EINVAL);
876	}
877
878	if ((pinfo->pshm_flags & PSHM_ALLOCATED) != PSHM_ALLOCATED) {
879		PSHM_SUBSYS_UNLOCK();
880		return(EINVAL);
881	}
882	if ((off_t)user_size > pinfo->pshm_length) {
883		PSHM_SUBSYS_UNLOCK();
884		return(EINVAL);
885	}
886	if ((off_t)(user_size + file_pos) > pinfo->pshm_length) {
887		PSHM_SUBSYS_UNLOCK();
888		return(EINVAL);
889	}
890	if ((pshmobj = pinfo->pshm_memobjects) == NULL) {
891		PSHM_SUBSYS_UNLOCK();
892		return(EINVAL);
893	}
894
895#if CONFIG_MACF
896	error = mac_posixshm_check_mmap(kauth_cred_get(), pinfo, prot, flags);
897	if (error) {
898		PSHM_SUBSYS_UNLOCK();
899		return(error);
900	}
901#endif
902
903	PSHM_SUBSYS_UNLOCK();
904	user_map = current_map();
905
906	if ((flags & MAP_FIXED) == 0) {
907		alloc_flags = VM_FLAGS_ANYWHERE;
908		user_addr = vm_map_round_page(user_addr,
909					      vm_map_page_mask(user_map));
910	} else {
911		if (user_addr != vm_map_round_page(user_addr,
912						   vm_map_page_mask(user_map)))
913			return (EINVAL);
914		/*
915		 * We do not get rid of the existing mappings here because
916		 * it wouldn't be atomic (see comment in mmap()).  We let
917		 * Mach VM know that we want it to replace any existing
918		 * mapping with the new one.
919		 */
920		alloc_flags = VM_FLAGS_FIXED | VM_FLAGS_OVERWRITE;
921	}
922	docow = FALSE;
923
924	mapped_size = 0;
925
926	/* reserver the entire space first... */
927	kret = vm_map_enter_mem_object(user_map,
928				       &user_addr,
929				       user_size,
930				       0,
931				       alloc_flags,
932				       IPC_PORT_NULL,
933				       0,
934				       FALSE,
935				       VM_PROT_NONE,
936				       VM_PROT_NONE,
937				       VM_INHERIT_NONE);
938	user_start_addr = user_addr;
939	if (kret != KERN_SUCCESS) {
940		goto out;
941	}
942
943	/* ... and overwrite with the real mappings */
944	for (map_pos = 0, pshmobj = pinfo->pshm_memobjects;
945	     user_size != 0;
946	     map_pos += pshmobj->pshmo_size, pshmobj = pshmobj->pshmo_next) {
947		if (pshmobj == NULL) {
948			/* nothing there to map !? */
949			goto out;
950		}
951		if (file_pos >= map_pos + pshmobj->pshmo_size) {
952			continue;
953		}
954		map_size = pshmobj->pshmo_size - (file_pos - map_pos);
955		if (map_size > user_size) {
956			map_size = user_size;
957		}
958		kret = vm_map_enter_mem_object(
959			user_map,
960			&user_addr,
961			map_size,
962			0,
963			VM_FLAGS_FIXED | VM_FLAGS_OVERWRITE,
964			pshmobj->pshmo_memobject,
965			file_pos - map_pos,
966			docow,
967			prot,
968			VM_PROT_DEFAULT,
969			VM_INHERIT_SHARE);
970		if (kret != KERN_SUCCESS)
971			goto out;
972
973		user_addr += map_size;
974		user_size -= map_size;
975		mapped_size += map_size;
976		file_pos += map_size;
977	}
978
979	PSHM_SUBSYS_LOCK();
980	pnode->mapp_addr = user_start_addr;
981	pnode->map_size = mapped_size;
982	pinfo->pshm_flags |= (PSHM_MAPPED | PSHM_INUSE);
983	PSHM_SUBSYS_UNLOCK();
984out:
985	if (kret != KERN_SUCCESS) {
986		if (mapped_size != 0) {
987			(void) mach_vm_deallocate(current_map(),
988						  user_start_addr,
989						  mapped_size);
990		}
991	}
992
993	switch (kret) {
994	case KERN_SUCCESS:
995		*retval = (user_start_addr + pageoff);
996		return (0);
997	case KERN_INVALID_ADDRESS:
998	case KERN_NO_SPACE:
999		return (ENOMEM);
1000	case KERN_PROTECTION_FAILURE:
1001		return (EACCES);
1002	default:
1003		return (EINVAL);
1004	}
1005
1006}
1007
1008int
1009shm_unlink(__unused proc_t p, struct shm_unlink_args *uap,
1010			__unused int32_t *retval)
1011{
1012	size_t i;
1013	int error=0;
1014	struct pshmname nd;
1015	struct pshminfo *pinfo;
1016	char * pnbuf;
1017	char * nameptr;
1018	char * cp;
1019	size_t pathlen, plen;
1020	int incache = 0;
1021	struct pshmcache *pcache = PSHMCACHE_NULL;
1022	struct pshmobj *pshmobj, *pshmobj_next;
1023
1024	pinfo = PSHMINFO_NULL;
1025
1026	MALLOC_ZONE(pnbuf, caddr_t, MAXPATHLEN, M_NAMEI, M_WAITOK);
1027	if (pnbuf == NULL) {
1028		return(ENOSPC);		/* XXX non-standard */
1029	}
1030	pathlen = MAXPATHLEN;
1031	error = copyinstr(uap->name, (void *)pnbuf, MAXPATHLEN, &pathlen);
1032	if (error) {
1033		goto bad;
1034	}
1035	AUDIT_ARG(text, pnbuf);
1036	if (pathlen > PSHMNAMLEN) {
1037		error = ENAMETOOLONG;
1038		goto bad;
1039	}
1040
1041
1042#ifdef PSXSHM_NAME_RESTRICT
1043	nameptr = pnbuf;
1044	if (*nameptr == '/') {
1045		while (*(nameptr++) == '/') {
1046			plen--;
1047			error = EINVAL;
1048			goto bad;
1049		}
1050        } else {
1051		error = EINVAL;
1052		goto bad;
1053	}
1054#endif /* PSXSHM_NAME_RESTRICT */
1055
1056	plen = pathlen;
1057	nameptr = pnbuf;
1058	nd.pshm_nameptr = nameptr;
1059	nd.pshm_namelen = plen;
1060	nd. pshm_hash =0;
1061
1062        for (cp = nameptr, i=1; *cp != 0 && i <= plen; i++, cp++) {
1063               nd.pshm_hash += (unsigned char)*cp * i;
1064	}
1065
1066	PSHM_SUBSYS_LOCK();
1067	error = pshm_cache_search(&pinfo, &nd, &pcache, 0);
1068
1069	if (error == ENOENT) {
1070		PSHM_SUBSYS_UNLOCK();
1071		goto bad;
1072
1073	}
1074	/* During unlink lookup failure also implies ENOENT */
1075	if (!error) {
1076		PSHM_SUBSYS_UNLOCK();
1077		error = ENOENT;
1078		goto bad;
1079	} else
1080		incache = 1;
1081
1082	if ((pinfo->pshm_flags & (PSHM_DEFINED | PSHM_ALLOCATED))==0) {
1083		PSHM_SUBSYS_UNLOCK();
1084		error = EINVAL;
1085		goto bad;
1086	}
1087
1088	if (pinfo->pshm_flags & PSHM_ALLOCATING) {
1089		/* XXX should we wait for flag to clear and then proceed ? */
1090		PSHM_SUBSYS_UNLOCK();
1091		error = EAGAIN;
1092		goto bad;
1093	}
1094
1095	if (pinfo->pshm_flags & PSHM_INDELETE) {
1096		PSHM_SUBSYS_UNLOCK();
1097		error = 0;
1098		goto bad;
1099	}
1100#if CONFIG_MACF
1101	error = mac_posixshm_check_unlink(kauth_cred_get(), pinfo, nameptr);
1102	if (error) {
1103		PSHM_SUBSYS_UNLOCK();
1104		goto bad;
1105	}
1106#endif
1107
1108	AUDIT_ARG(posix_ipc_perm, pinfo->pshm_uid, pinfo->pshm_gid,
1109		  pinfo->pshm_mode);
1110
1111	/*
1112	 * following file semantics, unlink should be allowed
1113	 * for users with write permission only.
1114	 */
1115	if ( (error = pshm_access(pinfo, FWRITE, kauth_cred_get(), p)) ) {
1116		PSHM_SUBSYS_UNLOCK();
1117		goto bad;
1118	}
1119
1120	pinfo->pshm_flags |= PSHM_INDELETE;
1121	pshm_cache_delete(pcache);
1122	pinfo->pshm_flags |= PSHM_REMOVED;
1123	/* release the existence reference */
1124 	if (!--pinfo->pshm_usecount) {
1125#if CONFIG_MACF
1126		mac_posixshm_label_destroy(pinfo);
1127#endif
1128		PSHM_SUBSYS_UNLOCK();
1129		/*
1130		 * If this is the last reference going away on the object,
1131		 * then we need to destroy the backing object.  The name
1132		 * has an implied but uncounted reference on the object,
1133		 * once it's created, since it's used as a rendezvous, and
1134		 * therefore may be subsequently reopened.
1135		 */
1136		for (pshmobj = pinfo->pshm_memobjects;
1137		     pshmobj != NULL;
1138		     pshmobj = pshmobj_next) {
1139			mach_memory_entry_port_release(pshmobj->pshmo_memobject);
1140			pshmobj_next = pshmobj->pshmo_next;
1141			FREE(pshmobj, M_SHM);
1142		}
1143		FREE(pinfo,M_SHM);
1144	} else {
1145		PSHM_SUBSYS_UNLOCK();
1146	}
1147	FREE(pcache, M_SHM);
1148	error = 0;
1149bad:
1150	FREE_ZONE(pnbuf, MAXPATHLEN, M_NAMEI);
1151	return (error);
1152}
1153
1154/* already called locked */
1155static int
1156pshm_close(struct pshminfo *pinfo, int dropref)
1157{
1158	int error = 0;
1159	struct pshmobj *pshmobj, *pshmobj_next;
1160
1161	/*
1162	 * If we are dropping the reference we took on the cache object, don't
1163	 * enforce the allocation requirement.
1164	 */
1165	if ( !dropref && ((pinfo->pshm_flags & PSHM_ALLOCATED) != PSHM_ALLOCATED)) {
1166		return(EINVAL);
1167	}
1168#if DIAGNOSTIC
1169	if(!pinfo->pshm_usecount) {
1170		kprintf("negative usecount in pshm_close\n");
1171	}
1172#endif /* DIAGNOSTIC */
1173	pinfo->pshm_usecount--; /* release this fd's reference */
1174
1175 	if ((pinfo->pshm_flags & PSHM_REMOVED) && !pinfo->pshm_usecount) {
1176#if CONFIG_MACF
1177		mac_posixshm_label_destroy(pinfo);
1178#endif
1179		PSHM_SUBSYS_UNLOCK();
1180		/*
1181		 * If this is the last reference going away on the object,
1182		 * then we need to destroy the backing object.
1183		 */
1184		for (pshmobj = pinfo->pshm_memobjects;
1185		     pshmobj != NULL;
1186		     pshmobj = pshmobj_next) {
1187			mach_memory_entry_port_release(pshmobj->pshmo_memobject);
1188			pshmobj_next = pshmobj->pshmo_next;
1189			FREE(pshmobj, M_SHM);
1190		}
1191		PSHM_SUBSYS_LOCK();
1192		FREE(pinfo,M_SHM);
1193	}
1194	return (error);
1195}
1196
1197/* vfs_context_t passed to match prototype for struct fileops */
1198static int
1199pshm_closefile(struct fileglob *fg, __unused vfs_context_t ctx)
1200{
1201	int error = EINVAL;
1202	struct pshmnode *pnode;
1203
1204	PSHM_SUBSYS_LOCK();
1205
1206	if ((pnode = (struct pshmnode *)fg->fg_data) != NULL) {
1207		if (pnode->pinfo != PSHMINFO_NULL) {
1208			error =  pshm_close(pnode->pinfo, 0);
1209		}
1210		FREE(pnode, M_SHM);
1211	}
1212
1213	PSHM_SUBSYS_UNLOCK();
1214
1215	return(error);
1216}
1217
1218static int
1219pshm_read(__unused struct fileproc *fp, __unused struct uio *uio,
1220			__unused int flags, __unused vfs_context_t ctx)
1221{
1222	return(ENOTSUP);
1223}
1224
1225static int
1226pshm_write(__unused struct fileproc *fp, __unused struct uio *uio,
1227			__unused int flags, __unused vfs_context_t ctx)
1228{
1229	return(ENOTSUP);
1230}
1231
1232static int
1233pshm_ioctl(__unused struct fileproc *fp, __unused u_long com,
1234			__unused caddr_t data, __unused vfs_context_t ctx)
1235{
1236	return(ENOTSUP);
1237}
1238
1239static int
1240pshm_select(__unused struct fileproc *fp, __unused int which, __unused void *wql,
1241			__unused vfs_context_t ctx)
1242{
1243	return(ENOTSUP);
1244}
1245
1246static int
1247pshm_kqfilter(__unused struct fileproc *fp, __unused struct knote *kn,
1248				__unused vfs_context_t ctx)
1249{
1250	return(ENOTSUP);
1251}
1252
1253int
1254fill_pshminfo(struct pshmnode * pshm, struct pshm_info * info)
1255{
1256	struct pshminfo *pinfo;
1257	struct vinfo_stat *sb;
1258
1259	PSHM_SUBSYS_LOCK();
1260	if ((pinfo = pshm->pinfo) == PSHMINFO_NULL){
1261		PSHM_SUBSYS_UNLOCK();
1262		return(EINVAL);
1263	}
1264
1265	sb = &info->pshm_stat;
1266
1267	bzero(sb, sizeof(struct vinfo_stat));
1268	sb->vst_mode = pinfo->pshm_mode;
1269	sb->vst_uid = pinfo->pshm_uid;
1270	sb->vst_gid = pinfo->pshm_gid;
1271	sb->vst_size = pinfo->pshm_length;
1272
1273	info->pshm_mappaddr = pshm->mapp_addr;
1274	bcopy(&pinfo->pshm_name[0], &info->pshm_name[0], PSHMNAMLEN+1);
1275
1276	PSHM_SUBSYS_UNLOCK();
1277	return(0);
1278}
1279
1280#if CONFIG_MACF
1281void
1282pshm_label_associate(struct fileproc *fp, struct vnode *vp, vfs_context_t ctx)
1283{
1284	struct pshmnode *pnode;
1285	struct pshminfo *pshm;
1286
1287	PSHM_SUBSYS_LOCK();
1288	pnode = (struct pshmnode *)fp->f_fglob->fg_data;
1289	if (pnode != NULL) {
1290		pshm = pnode->pinfo;
1291		if (pshm != NULL)
1292			mac_posixshm_vnode_label_associate(
1293				vfs_context_ucred(ctx), pshm, pshm->pshm_label,
1294				vp, vp->v_label);
1295	}
1296	PSHM_SUBSYS_UNLOCK();
1297}
1298#endif
1299