sysv_shm.c revision 216104
1/*	$NetBSD: sysv_shm.c,v 1.23 1994/07/04 23:25:12 glass Exp $	*/
2/*-
3 * Copyright (c) 1994 Adam Glass and Charles Hannum.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 *    must display the following acknowledgement:
15 *	This product includes software developed by Adam Glass and Charles
16 *	Hannum.
17 * 4. The names of the authors may not be used to endorse or promote products
18 *    derived from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR
21 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
22 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
23 * IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT, INDIRECT,
24 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
25 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
26 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
27 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
28 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
29 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
30 */
31/*-
32 * Copyright (c) 2003-2005 McAfee, Inc.
33 * All rights reserved.
34 *
35 * This software was developed for the FreeBSD Project in part by McAfee
36 * Research, the Security Research Division of McAfee, Inc under DARPA/SPAWAR
37 * contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS research
38 * program.
39 *
40 * Redistribution and use in source and binary forms, with or without
41 * modification, are permitted provided that the following conditions
42 * are met:
43 * 1. Redistributions of source code must retain the above copyright
44 *    notice, this list of conditions and the following disclaimer.
45 * 2. Redistributions in binary form must reproduce the above copyright
46 *    notice, this list of conditions and the following disclaimer in the
47 *    documentation and/or other materials provided with the distribution.
48 *
49 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 * SUCH DAMAGE.
60 */
61
62#include <sys/cdefs.h>
63__FBSDID("$FreeBSD: head/sys/kern/sysv_shm.c 216104 2010-12-02 01:14:45Z trasz $");
64
65#include "opt_compat.h"
66#include "opt_sysvipc.h"
67
68#include <sys/param.h>
69#include <sys/systm.h>
70#include <sys/kernel.h>
71#include <sys/limits.h>
72#include <sys/lock.h>
73#include <sys/sysctl.h>
74#include <sys/shm.h>
75#include <sys/proc.h>
76#include <sys/malloc.h>
77#include <sys/mman.h>
78#include <sys/module.h>
79#include <sys/mutex.h>
80#include <sys/resourcevar.h>
81#include <sys/stat.h>
82#include <sys/syscall.h>
83#include <sys/syscallsubr.h>
84#include <sys/sysent.h>
85#include <sys/sysproto.h>
86#include <sys/jail.h>
87
88#include <security/mac/mac_framework.h>
89
90#include <vm/vm.h>
91#include <vm/vm_param.h>
92#include <vm/pmap.h>
93#include <vm/vm_object.h>
94#include <vm/vm_map.h>
95#include <vm/vm_page.h>
96#include <vm/vm_pager.h>
97
98static MALLOC_DEFINE(M_SHM, "shm", "SVID compatible shared memory segments");
99
100static int shmget_allocate_segment(struct thread *td,
101    struct shmget_args *uap, int mode);
102static int shmget_existing(struct thread *td, struct shmget_args *uap,
103    int mode, int segnum);
104
105#define	SHMSEG_FREE     	0x0200
106#define	SHMSEG_REMOVED  	0x0400
107#define	SHMSEG_ALLOCATED	0x0800
108#define	SHMSEG_WANTED		0x1000
109
110static int shm_last_free, shm_nused, shmalloced;
111vm_size_t shm_committed;
112static struct shmid_kernel	*shmsegs;
113
114struct shmmap_state {
115	vm_offset_t va;
116	int shmid;
117};
118
119static void shm_deallocate_segment(struct shmid_kernel *);
120static int shm_find_segment_by_key(key_t);
121static struct shmid_kernel *shm_find_segment_by_shmid(int);
122static struct shmid_kernel *shm_find_segment_by_shmidx(int);
123static int shm_delete_mapping(struct vmspace *vm, struct shmmap_state *);
124static void shmrealloc(void);
125static int shminit(void);
126static int sysvshm_modload(struct module *, int, void *);
127static int shmunload(void);
128static void shmexit_myhook(struct vmspace *vm);
129static void shmfork_myhook(struct proc *p1, struct proc *p2);
130static int sysctl_shmsegs(SYSCTL_HANDLER_ARGS);
131
132/*
133 * Tuneable values.
134 */
135#ifndef SHMMAXPGS
136#define	SHMMAXPGS	131072	/* Note: sysv shared memory is swap backed. */
137#endif
138#ifndef SHMMAX
139#define	SHMMAX	(SHMMAXPGS*PAGE_SIZE)
140#endif
141#ifndef SHMMIN
142#define	SHMMIN	1
143#endif
144#ifndef SHMMNI
145#define	SHMMNI	192
146#endif
147#ifndef SHMSEG
148#define	SHMSEG	128
149#endif
150#ifndef SHMALL
151#define	SHMALL	(SHMMAXPGS)
152#endif
153
154struct	shminfo shminfo = {
155	SHMMAX,
156	SHMMIN,
157	SHMMNI,
158	SHMSEG,
159	SHMALL
160};
161
162static int shm_use_phys;
163static int shm_allow_removed;
164
165SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmax, CTLFLAG_RW, &shminfo.shmmax, 0,
166    "Maximum shared memory segment size");
167SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmin, CTLFLAG_RW, &shminfo.shmmin, 0,
168    "Minimum shared memory segment size");
169SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmni, CTLFLAG_RDTUN, &shminfo.shmmni, 0,
170    "Number of shared memory identifiers");
171SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmseg, CTLFLAG_RDTUN, &shminfo.shmseg, 0,
172    "Number of segments per process");
173SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmall, CTLFLAG_RW, &shminfo.shmall, 0,
174    "Maximum number of pages available for shared memory");
175SYSCTL_INT(_kern_ipc, OID_AUTO, shm_use_phys, CTLFLAG_RW,
176    &shm_use_phys, 0, "Enable/Disable locking of shared memory pages in core");
177SYSCTL_INT(_kern_ipc, OID_AUTO, shm_allow_removed, CTLFLAG_RW,
178    &shm_allow_removed, 0,
179    "Enable/Disable attachment to attached segments marked for removal");
180SYSCTL_PROC(_kern_ipc, OID_AUTO, shmsegs, CTLFLAG_RD,
181    NULL, 0, sysctl_shmsegs, "",
182    "Current number of shared memory segments allocated");
183
184static int
185shm_find_segment_by_key(key)
186	key_t key;
187{
188	int i;
189
190	for (i = 0; i < shmalloced; i++)
191		if ((shmsegs[i].u.shm_perm.mode & SHMSEG_ALLOCATED) &&
192		    shmsegs[i].u.shm_perm.key == key)
193			return (i);
194	return (-1);
195}
196
197static struct shmid_kernel *
198shm_find_segment_by_shmid(int shmid)
199{
200	int segnum;
201	struct shmid_kernel *shmseg;
202
203	segnum = IPCID_TO_IX(shmid);
204	if (segnum < 0 || segnum >= shmalloced)
205		return (NULL);
206	shmseg = &shmsegs[segnum];
207	if ((shmseg->u.shm_perm.mode & SHMSEG_ALLOCATED) == 0 ||
208	    (!shm_allow_removed &&
209	     (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) != 0) ||
210	    shmseg->u.shm_perm.seq != IPCID_TO_SEQ(shmid))
211		return (NULL);
212	return (shmseg);
213}
214
215static struct shmid_kernel *
216shm_find_segment_by_shmidx(int segnum)
217{
218	struct shmid_kernel *shmseg;
219
220	if (segnum < 0 || segnum >= shmalloced)
221		return (NULL);
222	shmseg = &shmsegs[segnum];
223	if ((shmseg->u.shm_perm.mode & SHMSEG_ALLOCATED) == 0 ||
224	    (!shm_allow_removed &&
225	     (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) != 0))
226		return (NULL);
227	return (shmseg);
228}
229
230static void
231shm_deallocate_segment(shmseg)
232	struct shmid_kernel *shmseg;
233{
234	vm_size_t size;
235
236	GIANT_REQUIRED;
237
238	vm_object_deallocate(shmseg->object);
239	shmseg->object = NULL;
240	size = round_page(shmseg->u.shm_segsz);
241	shm_committed -= btoc(size);
242	shm_nused--;
243	shmseg->u.shm_perm.mode = SHMSEG_FREE;
244#ifdef MAC
245	mac_sysvshm_cleanup(shmseg);
246#endif
247}
248
249static int
250shm_delete_mapping(struct vmspace *vm, struct shmmap_state *shmmap_s)
251{
252	struct shmid_kernel *shmseg;
253	int segnum, result;
254	vm_size_t size;
255
256	GIANT_REQUIRED;
257
258	segnum = IPCID_TO_IX(shmmap_s->shmid);
259	shmseg = &shmsegs[segnum];
260	size = round_page(shmseg->u.shm_segsz);
261	result = vm_map_remove(&vm->vm_map, shmmap_s->va, shmmap_s->va + size);
262	if (result != KERN_SUCCESS)
263		return (EINVAL);
264	shmmap_s->shmid = -1;
265	shmseg->u.shm_dtime = time_second;
266	if ((--shmseg->u.shm_nattch <= 0) &&
267	    (shmseg->u.shm_perm.mode & SHMSEG_REMOVED)) {
268		shm_deallocate_segment(shmseg);
269		shm_last_free = segnum;
270	}
271	return (0);
272}
273
274#ifndef _SYS_SYSPROTO_H_
275struct shmdt_args {
276	const void *shmaddr;
277};
278#endif
279int
280shmdt(td, uap)
281	struct thread *td;
282	struct shmdt_args *uap;
283{
284	struct proc *p = td->td_proc;
285	struct shmmap_state *shmmap_s;
286#ifdef MAC
287	struct shmid_kernel *shmsegptr;
288#endif
289	int i;
290	int error = 0;
291
292	if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
293		return (ENOSYS);
294	mtx_lock(&Giant);
295	shmmap_s = p->p_vmspace->vm_shm;
296 	if (shmmap_s == NULL) {
297		error = EINVAL;
298		goto done2;
299	}
300	for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) {
301		if (shmmap_s->shmid != -1 &&
302		    shmmap_s->va == (vm_offset_t)uap->shmaddr) {
303			break;
304		}
305	}
306	if (i == shminfo.shmseg) {
307		error = EINVAL;
308		goto done2;
309	}
310#ifdef MAC
311	shmsegptr = &shmsegs[IPCID_TO_IX(shmmap_s->shmid)];
312	error = mac_sysvshm_check_shmdt(td->td_ucred, shmsegptr);
313	if (error != 0)
314		goto done2;
315#endif
316	error = shm_delete_mapping(p->p_vmspace, shmmap_s);
317done2:
318	mtx_unlock(&Giant);
319	return (error);
320}
321
322#ifndef _SYS_SYSPROTO_H_
323struct shmat_args {
324	int shmid;
325	const void *shmaddr;
326	int shmflg;
327};
328#endif
329int
330kern_shmat(td, shmid, shmaddr, shmflg)
331	struct thread *td;
332	int shmid;
333	const void *shmaddr;
334	int shmflg;
335{
336	struct proc *p = td->td_proc;
337	int i, flags;
338	struct shmid_kernel *shmseg;
339	struct shmmap_state *shmmap_s = NULL;
340	vm_offset_t attach_va;
341	vm_prot_t prot;
342	vm_size_t size;
343	int rv;
344	int error = 0;
345
346	if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
347		return (ENOSYS);
348	mtx_lock(&Giant);
349	shmmap_s = p->p_vmspace->vm_shm;
350	if (shmmap_s == NULL) {
351		shmmap_s = malloc(shminfo.shmseg * sizeof(struct shmmap_state),
352		    M_SHM, M_WAITOK);
353		for (i = 0; i < shminfo.shmseg; i++)
354			shmmap_s[i].shmid = -1;
355		p->p_vmspace->vm_shm = shmmap_s;
356	}
357	shmseg = shm_find_segment_by_shmid(shmid);
358	if (shmseg == NULL) {
359		error = EINVAL;
360		goto done2;
361	}
362	error = ipcperm(td, &shmseg->u.shm_perm,
363	    (shmflg & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W);
364	if (error)
365		goto done2;
366#ifdef MAC
367	error = mac_sysvshm_check_shmat(td->td_ucred, shmseg, shmflg);
368	if (error != 0)
369		goto done2;
370#endif
371	for (i = 0; i < shminfo.shmseg; i++) {
372		if (shmmap_s->shmid == -1)
373			break;
374		shmmap_s++;
375	}
376	if (i >= shminfo.shmseg) {
377		error = EMFILE;
378		goto done2;
379	}
380	size = round_page(shmseg->u.shm_segsz);
381	prot = VM_PROT_READ;
382	if ((shmflg & SHM_RDONLY) == 0)
383		prot |= VM_PROT_WRITE;
384	flags = MAP_ANON | MAP_SHARED;
385	if (shmaddr) {
386		flags |= MAP_FIXED;
387		if (shmflg & SHM_RND) {
388			attach_va = (vm_offset_t)shmaddr & ~(SHMLBA-1);
389		} else if (((vm_offset_t)shmaddr & (SHMLBA-1)) == 0) {
390			attach_va = (vm_offset_t)shmaddr;
391		} else {
392			error = EINVAL;
393			goto done2;
394		}
395	} else {
396		/*
397		 * This is just a hint to vm_map_find() about where to
398		 * put it.
399		 */
400		PROC_LOCK(p);
401		attach_va = round_page((vm_offset_t)p->p_vmspace->vm_daddr +
402		    lim_max(p, RLIMIT_DATA));
403		PROC_UNLOCK(p);
404	}
405
406	vm_object_reference(shmseg->object);
407	rv = vm_map_find(&p->p_vmspace->vm_map, shmseg->object,
408	    0, &attach_va, size, (flags & MAP_FIXED) ? VMFS_NO_SPACE :
409	    VMFS_ANY_SPACE, prot, prot, 0);
410	if (rv != KERN_SUCCESS) {
411		vm_object_deallocate(shmseg->object);
412		error = ENOMEM;
413		goto done2;
414	}
415	vm_map_inherit(&p->p_vmspace->vm_map,
416		attach_va, attach_va + size, VM_INHERIT_SHARE);
417
418	shmmap_s->va = attach_va;
419	shmmap_s->shmid = shmid;
420	shmseg->u.shm_lpid = p->p_pid;
421	shmseg->u.shm_atime = time_second;
422	shmseg->u.shm_nattch++;
423	td->td_retval[0] = attach_va;
424done2:
425	mtx_unlock(&Giant);
426	return (error);
427}
428
429int
430shmat(td, uap)
431	struct thread *td;
432	struct shmat_args *uap;
433{
434	return kern_shmat(td, uap->shmid, uap->shmaddr, uap->shmflg);
435}
436
437int
438kern_shmctl(td, shmid, cmd, buf, bufsz)
439	struct thread *td;
440	int shmid;
441	int cmd;
442	void *buf;
443	size_t *bufsz;
444{
445	int error = 0;
446	struct shmid_kernel *shmseg;
447
448	if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
449		return (ENOSYS);
450
451	mtx_lock(&Giant);
452	switch (cmd) {
453	/*
454	 * It is possible that kern_shmctl is being called from the Linux ABI
455	 * layer, in which case, we will need to implement IPC_INFO.  It should
456	 * be noted that other shmctl calls will be funneled through here for
457	 * Linix binaries as well.
458	 *
459	 * NB: The Linux ABI layer will convert this data to structure(s) more
460	 * consistent with the Linux ABI.
461	 */
462	case IPC_INFO:
463		memcpy(buf, &shminfo, sizeof(shminfo));
464		if (bufsz)
465			*bufsz = sizeof(shminfo);
466		td->td_retval[0] = shmalloced;
467		goto done2;
468	case SHM_INFO: {
469		struct shm_info shm_info;
470		shm_info.used_ids = shm_nused;
471		shm_info.shm_rss = 0;	/*XXX where to get from ? */
472		shm_info.shm_tot = 0;	/*XXX where to get from ? */
473		shm_info.shm_swp = 0;	/*XXX where to get from ? */
474		shm_info.swap_attempts = 0;	/*XXX where to get from ? */
475		shm_info.swap_successes = 0;	/*XXX where to get from ? */
476		memcpy(buf, &shm_info, sizeof(shm_info));
477		if (bufsz)
478			*bufsz = sizeof(shm_info);
479		td->td_retval[0] = shmalloced;
480		goto done2;
481	}
482	}
483	if (cmd == SHM_STAT)
484		shmseg = shm_find_segment_by_shmidx(shmid);
485	else
486		shmseg = shm_find_segment_by_shmid(shmid);
487	if (shmseg == NULL) {
488		error = EINVAL;
489		goto done2;
490	}
491#ifdef MAC
492	error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, cmd);
493	if (error != 0)
494		goto done2;
495#endif
496	switch (cmd) {
497	case SHM_STAT:
498	case IPC_STAT:
499		error = ipcperm(td, &shmseg->u.shm_perm, IPC_R);
500		if (error)
501			goto done2;
502		memcpy(buf, &shmseg->u, sizeof(struct shmid_ds));
503		if (bufsz)
504			*bufsz = sizeof(struct shmid_ds);
505		if (cmd == SHM_STAT)
506			td->td_retval[0] = IXSEQ_TO_IPCID(shmid, shmseg->u.shm_perm);
507		break;
508	case IPC_SET: {
509		struct shmid_ds *shmid;
510
511		shmid = (struct shmid_ds *)buf;
512		error = ipcperm(td, &shmseg->u.shm_perm, IPC_M);
513		if (error)
514			goto done2;
515		shmseg->u.shm_perm.uid = shmid->shm_perm.uid;
516		shmseg->u.shm_perm.gid = shmid->shm_perm.gid;
517		shmseg->u.shm_perm.mode =
518		    (shmseg->u.shm_perm.mode & ~ACCESSPERMS) |
519		    (shmid->shm_perm.mode & ACCESSPERMS);
520		shmseg->u.shm_ctime = time_second;
521		break;
522	}
523	case IPC_RMID:
524		error = ipcperm(td, &shmseg->u.shm_perm, IPC_M);
525		if (error)
526			goto done2;
527		shmseg->u.shm_perm.key = IPC_PRIVATE;
528		shmseg->u.shm_perm.mode |= SHMSEG_REMOVED;
529		if (shmseg->u.shm_nattch <= 0) {
530			shm_deallocate_segment(shmseg);
531			shm_last_free = IPCID_TO_IX(shmid);
532		}
533		break;
534#if 0
535	case SHM_LOCK:
536	case SHM_UNLOCK:
537#endif
538	default:
539		error = EINVAL;
540		break;
541	}
542done2:
543	mtx_unlock(&Giant);
544	return (error);
545}
546
547#ifndef _SYS_SYSPROTO_H_
548struct shmctl_args {
549	int shmid;
550	int cmd;
551	struct shmid_ds *buf;
552};
553#endif
554int
555shmctl(td, uap)
556	struct thread *td;
557	struct shmctl_args *uap;
558{
559	int error = 0;
560	struct shmid_ds buf;
561	size_t bufsz;
562
563	/*
564	 * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support
565	 * Linux binaries.  If we see the call come through the FreeBSD ABI,
566	 * return an error back to the user since we do not to support this.
567	 */
568	if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO ||
569	    uap->cmd == SHM_STAT)
570		return (EINVAL);
571
572	/* IPC_SET needs to copyin the buffer before calling kern_shmctl */
573	if (uap->cmd == IPC_SET) {
574		if ((error = copyin(uap->buf, &buf, sizeof(struct shmid_ds))))
575			goto done;
576	}
577
578	error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz);
579	if (error)
580		goto done;
581
582	/* Cases in which we need to copyout */
583	switch (uap->cmd) {
584	case IPC_STAT:
585		error = copyout(&buf, uap->buf, bufsz);
586		break;
587	}
588
589done:
590	if (error) {
591		/* Invalidate the return value */
592		td->td_retval[0] = -1;
593	}
594	return (error);
595}
596
597
598static int
599shmget_existing(td, uap, mode, segnum)
600	struct thread *td;
601	struct shmget_args *uap;
602	int mode;
603	int segnum;
604{
605	struct shmid_kernel *shmseg;
606	int error;
607
608	shmseg = &shmsegs[segnum];
609	if (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) {
610		/*
611		 * This segment is in the process of being allocated.  Wait
612		 * until it's done, and look the key up again (in case the
613		 * allocation failed or it was freed).
614		 */
615		shmseg->u.shm_perm.mode |= SHMSEG_WANTED;
616		error = tsleep(shmseg, PLOCK | PCATCH, "shmget", 0);
617		if (error)
618			return (error);
619		return (EAGAIN);
620	}
621	if ((uap->shmflg & (IPC_CREAT | IPC_EXCL)) == (IPC_CREAT | IPC_EXCL))
622		return (EEXIST);
623#ifdef MAC
624	error = mac_sysvshm_check_shmget(td->td_ucred, shmseg, uap->shmflg);
625	if (error != 0)
626		return (error);
627#endif
628	if (uap->size != 0 && uap->size > shmseg->u.shm_segsz)
629		return (EINVAL);
630	td->td_retval[0] = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm);
631	return (0);
632}
633
634static int
635shmget_allocate_segment(td, uap, mode)
636	struct thread *td;
637	struct shmget_args *uap;
638	int mode;
639{
640	int i, segnum, shmid;
641	size_t size;
642	struct ucred *cred = td->td_ucred;
643	struct shmid_kernel *shmseg;
644	vm_object_t shm_object;
645
646	GIANT_REQUIRED;
647
648	if (uap->size < shminfo.shmmin || uap->size > shminfo.shmmax)
649		return (EINVAL);
650	if (shm_nused >= shminfo.shmmni) /* Any shmids left? */
651		return (ENOSPC);
652	size = round_page(uap->size);
653	if (shm_committed + btoc(size) > shminfo.shmall)
654		return (ENOMEM);
655	if (shm_last_free < 0) {
656		shmrealloc();	/* Maybe expand the shmsegs[] array. */
657		for (i = 0; i < shmalloced; i++)
658			if (shmsegs[i].u.shm_perm.mode & SHMSEG_FREE)
659				break;
660		if (i == shmalloced)
661			return (ENOSPC);
662		segnum = i;
663	} else  {
664		segnum = shm_last_free;
665		shm_last_free = -1;
666	}
667	shmseg = &shmsegs[segnum];
668	/*
669	 * In case we sleep in malloc(), mark the segment present but deleted
670	 * so that noone else tries to create the same key.
671	 */
672	shmseg->u.shm_perm.mode = SHMSEG_ALLOCATED | SHMSEG_REMOVED;
673	shmseg->u.shm_perm.key = uap->key;
674	shmseg->u.shm_perm.seq = (shmseg->u.shm_perm.seq + 1) & 0x7fff;
675	shmid = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm);
676
677	/*
678	 * We make sure that we have allocated a pager before we need
679	 * to.
680	 */
681	shm_object = vm_pager_allocate(shm_use_phys ? OBJT_PHYS : OBJT_SWAP,
682	    0, size, VM_PROT_DEFAULT, 0, cred);
683	if (shm_object == NULL)
684		return (ENOMEM);
685	VM_OBJECT_LOCK(shm_object);
686	vm_object_clear_flag(shm_object, OBJ_ONEMAPPING);
687	vm_object_set_flag(shm_object, OBJ_NOSPLIT);
688	VM_OBJECT_UNLOCK(shm_object);
689
690	shmseg->object = shm_object;
691	shmseg->u.shm_perm.cuid = shmseg->u.shm_perm.uid = cred->cr_uid;
692	shmseg->u.shm_perm.cgid = shmseg->u.shm_perm.gid = cred->cr_gid;
693	shmseg->u.shm_perm.mode = (shmseg->u.shm_perm.mode & SHMSEG_WANTED) |
694	    (mode & ACCESSPERMS) | SHMSEG_ALLOCATED;
695	shmseg->u.shm_segsz = uap->size;
696	shmseg->u.shm_cpid = td->td_proc->p_pid;
697	shmseg->u.shm_lpid = shmseg->u.shm_nattch = 0;
698	shmseg->u.shm_atime = shmseg->u.shm_dtime = 0;
699#ifdef MAC
700	mac_sysvshm_create(cred, shmseg);
701#endif
702	shmseg->u.shm_ctime = time_second;
703	shm_committed += btoc(size);
704	shm_nused++;
705	if (shmseg->u.shm_perm.mode & SHMSEG_WANTED) {
706		/*
707		 * Somebody else wanted this key while we were asleep.  Wake
708		 * them up now.
709		 */
710		shmseg->u.shm_perm.mode &= ~SHMSEG_WANTED;
711		wakeup(shmseg);
712	}
713	td->td_retval[0] = shmid;
714	return (0);
715}
716
717#ifndef _SYS_SYSPROTO_H_
718struct shmget_args {
719	key_t key;
720	size_t size;
721	int shmflg;
722};
723#endif
724int
725shmget(td, uap)
726	struct thread *td;
727	struct shmget_args *uap;
728{
729	int segnum, mode;
730	int error;
731
732	if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
733		return (ENOSYS);
734	mtx_lock(&Giant);
735	mode = uap->shmflg & ACCESSPERMS;
736	if (uap->key != IPC_PRIVATE) {
737	again:
738		segnum = shm_find_segment_by_key(uap->key);
739		if (segnum >= 0) {
740			error = shmget_existing(td, uap, mode, segnum);
741			if (error == EAGAIN)
742				goto again;
743			goto done2;
744		}
745		if ((uap->shmflg & IPC_CREAT) == 0) {
746			error = ENOENT;
747			goto done2;
748		}
749	}
750	error = shmget_allocate_segment(td, uap, mode);
751done2:
752	mtx_unlock(&Giant);
753	return (error);
754}
755
756static void
757shmfork_myhook(p1, p2)
758	struct proc *p1, *p2;
759{
760	struct shmmap_state *shmmap_s;
761	size_t size;
762	int i;
763
764	mtx_lock(&Giant);
765	size = shminfo.shmseg * sizeof(struct shmmap_state);
766	shmmap_s = malloc(size, M_SHM, M_WAITOK);
767	bcopy(p1->p_vmspace->vm_shm, shmmap_s, size);
768	p2->p_vmspace->vm_shm = shmmap_s;
769	for (i = 0; i < shminfo.shmseg; i++, shmmap_s++)
770		if (shmmap_s->shmid != -1)
771			shmsegs[IPCID_TO_IX(shmmap_s->shmid)].u.shm_nattch++;
772	mtx_unlock(&Giant);
773}
774
775static void
776shmexit_myhook(struct vmspace *vm)
777{
778	struct shmmap_state *base, *shm;
779	int i;
780
781	if ((base = vm->vm_shm) != NULL) {
782		vm->vm_shm = NULL;
783		mtx_lock(&Giant);
784		for (i = 0, shm = base; i < shminfo.shmseg; i++, shm++) {
785			if (shm->shmid != -1)
786				shm_delete_mapping(vm, shm);
787		}
788		mtx_unlock(&Giant);
789		free(base, M_SHM);
790	}
791}
792
793static void
794shmrealloc(void)
795{
796	int i;
797	struct shmid_kernel *newsegs;
798
799	if (shmalloced >= shminfo.shmmni)
800		return;
801
802	newsegs = malloc(shminfo.shmmni * sizeof(*newsegs), M_SHM, M_WAITOK);
803	if (newsegs == NULL)
804		return;
805	for (i = 0; i < shmalloced; i++)
806		bcopy(&shmsegs[i], &newsegs[i], sizeof(newsegs[0]));
807	for (; i < shminfo.shmmni; i++) {
808		shmsegs[i].u.shm_perm.mode = SHMSEG_FREE;
809		shmsegs[i].u.shm_perm.seq = 0;
810#ifdef MAC
811		mac_sysvshm_init(&shmsegs[i]);
812#endif
813	}
814	free(shmsegs, M_SHM);
815	shmsegs = newsegs;
816	shmalloced = shminfo.shmmni;
817}
818
819static struct syscall_helper_data shm_syscalls[] = {
820	SYSCALL_INIT_HELPER(shmat),
821	SYSCALL_INIT_HELPER(shmctl),
822	SYSCALL_INIT_HELPER(shmdt),
823	SYSCALL_INIT_HELPER(shmget),
824#if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
825    defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
826	SYSCALL_INIT_HELPER(freebsd7_shmctl),
827#endif
828#if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43))
829	SYSCALL_INIT_HELPER(shmsys),
830#endif
831	SYSCALL_INIT_LAST
832};
833
834#ifdef COMPAT_FREEBSD32
835#include <compat/freebsd32/freebsd32.h>
836#include <compat/freebsd32/freebsd32_ipc.h>
837#include <compat/freebsd32/freebsd32_proto.h>
838#include <compat/freebsd32/freebsd32_signal.h>
839#include <compat/freebsd32/freebsd32_syscall.h>
840#include <compat/freebsd32/freebsd32_util.h>
841
842static struct syscall_helper_data shm32_syscalls[] = {
843	SYSCALL32_INIT_HELPER(shmat),
844	SYSCALL32_INIT_HELPER(shmdt),
845	SYSCALL32_INIT_HELPER(shmget),
846	SYSCALL32_INIT_HELPER(freebsd32_shmsys),
847	SYSCALL32_INIT_HELPER(freebsd32_shmctl),
848#if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
849    defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
850	SYSCALL32_INIT_HELPER(freebsd7_freebsd32_shmctl),
851#endif
852	SYSCALL_INIT_LAST
853};
854#endif
855
856static int
857shminit()
858{
859	int i, error;
860
861#ifndef BURN_BRIDGES
862	if (TUNABLE_ULONG_FETCH("kern.ipc.shmmaxpgs", &shminfo.shmall) != 0)
863		printf("kern.ipc.shmmaxpgs is now called kern.ipc.shmall!\n");
864#endif
865	TUNABLE_ULONG_FETCH("kern.ipc.shmall", &shminfo.shmall);
866
867	/* Initialize shmmax dealing with possible overflow. */
868	for (i = PAGE_SIZE; i > 0; i--) {
869		shminfo.shmmax = shminfo.shmall * i;
870		if (shminfo.shmmax >= shminfo.shmall)
871			break;
872	}
873
874	TUNABLE_ULONG_FETCH("kern.ipc.shmmin", &shminfo.shmmin);
875	TUNABLE_ULONG_FETCH("kern.ipc.shmmni", &shminfo.shmmni);
876	TUNABLE_ULONG_FETCH("kern.ipc.shmseg", &shminfo.shmseg);
877	TUNABLE_INT_FETCH("kern.ipc.shm_use_phys", &shm_use_phys);
878
879	shmalloced = shminfo.shmmni;
880	shmsegs = malloc(shmalloced * sizeof(shmsegs[0]), M_SHM, M_WAITOK);
881	for (i = 0; i < shmalloced; i++) {
882		shmsegs[i].u.shm_perm.mode = SHMSEG_FREE;
883		shmsegs[i].u.shm_perm.seq = 0;
884#ifdef MAC
885		mac_sysvshm_init(&shmsegs[i]);
886#endif
887	}
888	shm_last_free = 0;
889	shm_nused = 0;
890	shm_committed = 0;
891	shmexit_hook = &shmexit_myhook;
892	shmfork_hook = &shmfork_myhook;
893
894	error = syscall_helper_register(shm_syscalls);
895	if (error != 0)
896		return (error);
897#ifdef COMPAT_FREEBSD32
898	error = syscall32_helper_register(shm32_syscalls);
899	if (error != 0)
900		return (error);
901#endif
902	return (0);
903}
904
905static int
906shmunload()
907{
908	int i;
909
910	if (shm_nused > 0)
911		return (EBUSY);
912
913#ifdef COMPAT_FREEBSD32
914	syscall32_helper_unregister(shm32_syscalls);
915#endif
916	syscall_helper_unregister(shm_syscalls);
917
918	for (i = 0; i < shmalloced; i++) {
919#ifdef MAC
920		mac_sysvshm_destroy(&shmsegs[i]);
921#endif
922		/*
923		 * Objects might be still mapped into the processes
924		 * address spaces.  Actual free would happen on the
925		 * last mapping destruction.
926		 */
927		if (shmsegs[i].u.shm_perm.mode != SHMSEG_FREE)
928			vm_object_deallocate(shmsegs[i].object);
929	}
930	free(shmsegs, M_SHM);
931	shmexit_hook = NULL;
932	shmfork_hook = NULL;
933	return (0);
934}
935
936static int
937sysctl_shmsegs(SYSCTL_HANDLER_ARGS)
938{
939
940	return (SYSCTL_OUT(req, shmsegs, shmalloced * sizeof(shmsegs[0])));
941}
942
943#if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43))
944struct oshmid_ds {
945	struct	ipc_perm_old shm_perm;	/* operation perms */
946	int	shm_segsz;		/* size of segment (bytes) */
947	u_short	shm_cpid;		/* pid, creator */
948	u_short	shm_lpid;		/* pid, last operation */
949	short	shm_nattch;		/* no. of current attaches */
950	time_t	shm_atime;		/* last attach time */
951	time_t	shm_dtime;		/* last detach time */
952	time_t	shm_ctime;		/* last change time */
953	void	*shm_handle;		/* internal handle for shm segment */
954};
955
956struct oshmctl_args {
957	int shmid;
958	int cmd;
959	struct oshmid_ds *ubuf;
960};
961
962static int
963oshmctl(struct thread *td, struct oshmctl_args *uap)
964{
965#ifdef COMPAT_43
966	int error = 0;
967	struct shmid_kernel *shmseg;
968	struct oshmid_ds outbuf;
969
970	if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
971		return (ENOSYS);
972	mtx_lock(&Giant);
973	shmseg = shm_find_segment_by_shmid(uap->shmid);
974	if (shmseg == NULL) {
975		error = EINVAL;
976		goto done2;
977	}
978	switch (uap->cmd) {
979	case IPC_STAT:
980		error = ipcperm(td, &shmseg->u.shm_perm, IPC_R);
981		if (error)
982			goto done2;
983#ifdef MAC
984		error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, uap->cmd);
985		if (error != 0)
986			goto done2;
987#endif
988		ipcperm_new2old(&shmseg->u.shm_perm, &outbuf.shm_perm);
989		outbuf.shm_segsz = shmseg->u.shm_segsz;
990		outbuf.shm_cpid = shmseg->u.shm_cpid;
991		outbuf.shm_lpid = shmseg->u.shm_lpid;
992		outbuf.shm_nattch = shmseg->u.shm_nattch;
993		outbuf.shm_atime = shmseg->u.shm_atime;
994		outbuf.shm_dtime = shmseg->u.shm_dtime;
995		outbuf.shm_ctime = shmseg->u.shm_ctime;
996		outbuf.shm_handle = shmseg->object;
997		error = copyout(&outbuf, uap->ubuf, sizeof(outbuf));
998		if (error)
999			goto done2;
1000		break;
1001	default:
1002		error = freebsd7_shmctl(td, (struct freebsd7_shmctl_args *)uap);
1003		break;
1004	}
1005done2:
1006	mtx_unlock(&Giant);
1007	return (error);
1008#else
1009	return (EINVAL);
1010#endif
1011}
1012
1013/* XXX casting to (sy_call_t *) is bogus, as usual. */
1014static sy_call_t *shmcalls[] = {
1015	(sy_call_t *)shmat, (sy_call_t *)oshmctl,
1016	(sy_call_t *)shmdt, (sy_call_t *)shmget,
1017	(sy_call_t *)freebsd7_shmctl
1018};
1019
1020int
1021shmsys(td, uap)
1022	struct thread *td;
1023	/* XXX actually varargs. */
1024	struct shmsys_args /* {
1025		int	which;
1026		int	a2;
1027		int	a3;
1028		int	a4;
1029	} */ *uap;
1030{
1031	int error;
1032
1033	if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
1034		return (ENOSYS);
1035	if (uap->which < 0 ||
1036	    uap->which >= sizeof(shmcalls)/sizeof(shmcalls[0]))
1037		return (EINVAL);
1038	mtx_lock(&Giant);
1039	error = (*shmcalls[uap->which])(td, &uap->a2);
1040	mtx_unlock(&Giant);
1041	return (error);
1042}
1043
1044#endif	/* i386 && (COMPAT_FREEBSD4 || COMPAT_43) */
1045
1046#ifdef COMPAT_FREEBSD32
1047
1048int
1049freebsd32_shmsys(struct thread *td, struct freebsd32_shmsys_args *uap)
1050{
1051
1052#if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1053    defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1054	switch (uap->which) {
1055	case 0:	{	/* shmat */
1056		struct shmat_args ap;
1057
1058		ap.shmid = uap->a2;
1059		ap.shmaddr = PTRIN(uap->a3);
1060		ap.shmflg = uap->a4;
1061		return (sysent[SYS_shmat].sy_call(td, &ap));
1062	}
1063	case 2: {	/* shmdt */
1064		struct shmdt_args ap;
1065
1066		ap.shmaddr = PTRIN(uap->a2);
1067		return (sysent[SYS_shmdt].sy_call(td, &ap));
1068	}
1069	case 3: {	/* shmget */
1070		struct shmget_args ap;
1071
1072		ap.key = uap->a2;
1073		ap.size = uap->a3;
1074		ap.shmflg = uap->a4;
1075		return (sysent[SYS_shmget].sy_call(td, &ap));
1076	}
1077	case 4: {	/* shmctl */
1078		struct freebsd7_freebsd32_shmctl_args ap;
1079
1080		ap.shmid = uap->a2;
1081		ap.cmd = uap->a3;
1082		ap.buf = PTRIN(uap->a4);
1083		return (freebsd7_freebsd32_shmctl(td, &ap));
1084	}
1085	case 1:		/* oshmctl */
1086	default:
1087		return (EINVAL);
1088	}
1089#else
1090	return (nosys(td, NULL));
1091#endif
1092}
1093
1094#if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1095    defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1096int
1097freebsd7_freebsd32_shmctl(struct thread *td,
1098    struct freebsd7_freebsd32_shmctl_args *uap)
1099{
1100	int error = 0;
1101	union {
1102		struct shmid_ds shmid_ds;
1103		struct shm_info shm_info;
1104		struct shminfo shminfo;
1105	} u;
1106	union {
1107		struct shmid_ds32_old shmid_ds32;
1108		struct shm_info32 shm_info32;
1109		struct shminfo32 shminfo32;
1110	} u32;
1111	size_t sz;
1112
1113	if (uap->cmd == IPC_SET) {
1114		if ((error = copyin(uap->buf, &u32.shmid_ds32,
1115		    sizeof(u32.shmid_ds32))))
1116			goto done;
1117		freebsd32_ipcperm_old_in(&u32.shmid_ds32.shm_perm,
1118		    &u.shmid_ds.shm_perm);
1119		CP(u32.shmid_ds32, u.shmid_ds, shm_segsz);
1120		CP(u32.shmid_ds32, u.shmid_ds, shm_lpid);
1121		CP(u32.shmid_ds32, u.shmid_ds, shm_cpid);
1122		CP(u32.shmid_ds32, u.shmid_ds, shm_nattch);
1123		CP(u32.shmid_ds32, u.shmid_ds, shm_atime);
1124		CP(u32.shmid_ds32, u.shmid_ds, shm_dtime);
1125		CP(u32.shmid_ds32, u.shmid_ds, shm_ctime);
1126	}
1127
1128	error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz);
1129	if (error)
1130		goto done;
1131
1132	/* Cases in which we need to copyout */
1133	switch (uap->cmd) {
1134	case IPC_INFO:
1135		CP(u.shminfo, u32.shminfo32, shmmax);
1136		CP(u.shminfo, u32.shminfo32, shmmin);
1137		CP(u.shminfo, u32.shminfo32, shmmni);
1138		CP(u.shminfo, u32.shminfo32, shmseg);
1139		CP(u.shminfo, u32.shminfo32, shmall);
1140		error = copyout(&u32.shminfo32, uap->buf,
1141		    sizeof(u32.shminfo32));
1142		break;
1143	case SHM_INFO:
1144		CP(u.shm_info, u32.shm_info32, used_ids);
1145		CP(u.shm_info, u32.shm_info32, shm_rss);
1146		CP(u.shm_info, u32.shm_info32, shm_tot);
1147		CP(u.shm_info, u32.shm_info32, shm_swp);
1148		CP(u.shm_info, u32.shm_info32, swap_attempts);
1149		CP(u.shm_info, u32.shm_info32, swap_successes);
1150		error = copyout(&u32.shm_info32, uap->buf,
1151		    sizeof(u32.shm_info32));
1152		break;
1153	case SHM_STAT:
1154	case IPC_STAT:
1155		freebsd32_ipcperm_old_out(&u.shmid_ds.shm_perm,
1156		    &u32.shmid_ds32.shm_perm);
1157		if (u.shmid_ds.shm_segsz > INT32_MAX)
1158			u32.shmid_ds32.shm_segsz = INT32_MAX;
1159		else
1160			CP(u.shmid_ds, u32.shmid_ds32, shm_segsz);
1161		CP(u.shmid_ds, u32.shmid_ds32, shm_lpid);
1162		CP(u.shmid_ds, u32.shmid_ds32, shm_cpid);
1163		CP(u.shmid_ds, u32.shmid_ds32, shm_nattch);
1164		CP(u.shmid_ds, u32.shmid_ds32, shm_atime);
1165		CP(u.shmid_ds, u32.shmid_ds32, shm_dtime);
1166		CP(u.shmid_ds, u32.shmid_ds32, shm_ctime);
1167		u32.shmid_ds32.shm_internal = 0;
1168		error = copyout(&u32.shmid_ds32, uap->buf,
1169		    sizeof(u32.shmid_ds32));
1170		break;
1171	}
1172
1173done:
1174	if (error) {
1175		/* Invalidate the return value */
1176		td->td_retval[0] = -1;
1177	}
1178	return (error);
1179}
1180#endif
1181
1182int
1183freebsd32_shmctl(struct thread *td, struct freebsd32_shmctl_args *uap)
1184{
1185	int error = 0;
1186	union {
1187		struct shmid_ds shmid_ds;
1188		struct shm_info shm_info;
1189		struct shminfo shminfo;
1190	} u;
1191	union {
1192		struct shmid_ds32 shmid_ds32;
1193		struct shm_info32 shm_info32;
1194		struct shminfo32 shminfo32;
1195	} u32;
1196	size_t sz;
1197
1198	if (uap->cmd == IPC_SET) {
1199		if ((error = copyin(uap->buf, &u32.shmid_ds32,
1200		    sizeof(u32.shmid_ds32))))
1201			goto done;
1202		freebsd32_ipcperm_in(&u32.shmid_ds32.shm_perm,
1203		    &u.shmid_ds.shm_perm);
1204		CP(u32.shmid_ds32, u.shmid_ds, shm_segsz);
1205		CP(u32.shmid_ds32, u.shmid_ds, shm_lpid);
1206		CP(u32.shmid_ds32, u.shmid_ds, shm_cpid);
1207		CP(u32.shmid_ds32, u.shmid_ds, shm_nattch);
1208		CP(u32.shmid_ds32, u.shmid_ds, shm_atime);
1209		CP(u32.shmid_ds32, u.shmid_ds, shm_dtime);
1210		CP(u32.shmid_ds32, u.shmid_ds, shm_ctime);
1211	}
1212
1213	error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz);
1214	if (error)
1215		goto done;
1216
1217	/* Cases in which we need to copyout */
1218	switch (uap->cmd) {
1219	case IPC_INFO:
1220		CP(u.shminfo, u32.shminfo32, shmmax);
1221		CP(u.shminfo, u32.shminfo32, shmmin);
1222		CP(u.shminfo, u32.shminfo32, shmmni);
1223		CP(u.shminfo, u32.shminfo32, shmseg);
1224		CP(u.shminfo, u32.shminfo32, shmall);
1225		error = copyout(&u32.shminfo32, uap->buf,
1226		    sizeof(u32.shminfo32));
1227		break;
1228	case SHM_INFO:
1229		CP(u.shm_info, u32.shm_info32, used_ids);
1230		CP(u.shm_info, u32.shm_info32, shm_rss);
1231		CP(u.shm_info, u32.shm_info32, shm_tot);
1232		CP(u.shm_info, u32.shm_info32, shm_swp);
1233		CP(u.shm_info, u32.shm_info32, swap_attempts);
1234		CP(u.shm_info, u32.shm_info32, swap_successes);
1235		error = copyout(&u32.shm_info32, uap->buf,
1236		    sizeof(u32.shm_info32));
1237		break;
1238	case SHM_STAT:
1239	case IPC_STAT:
1240		freebsd32_ipcperm_out(&u.shmid_ds.shm_perm,
1241		    &u32.shmid_ds32.shm_perm);
1242		if (u.shmid_ds.shm_segsz > INT32_MAX)
1243			u32.shmid_ds32.shm_segsz = INT32_MAX;
1244		else
1245			CP(u.shmid_ds, u32.shmid_ds32, shm_segsz);
1246		CP(u.shmid_ds, u32.shmid_ds32, shm_lpid);
1247		CP(u.shmid_ds, u32.shmid_ds32, shm_cpid);
1248		CP(u.shmid_ds, u32.shmid_ds32, shm_nattch);
1249		CP(u.shmid_ds, u32.shmid_ds32, shm_atime);
1250		CP(u.shmid_ds, u32.shmid_ds32, shm_dtime);
1251		CP(u.shmid_ds, u32.shmid_ds32, shm_ctime);
1252		error = copyout(&u32.shmid_ds32, uap->buf,
1253		    sizeof(u32.shmid_ds32));
1254		break;
1255	}
1256
1257done:
1258	if (error) {
1259		/* Invalidate the return value */
1260		td->td_retval[0] = -1;
1261	}
1262	return (error);
1263}
1264#endif
1265
1266#if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1267    defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1268
1269#ifndef CP
1270#define CP(src, dst, fld)	do { (dst).fld = (src).fld; } while (0)
1271#endif
1272
1273#ifndef _SYS_SYSPROTO_H_
1274struct freebsd7_shmctl_args {
1275	int shmid;
1276	int cmd;
1277	struct shmid_ds_old *buf;
1278};
1279#endif
1280int
1281freebsd7_shmctl(td, uap)
1282	struct thread *td;
1283	struct freebsd7_shmctl_args *uap;
1284{
1285	int error = 0;
1286	struct shmid_ds_old old;
1287	struct shmid_ds buf;
1288	size_t bufsz;
1289
1290	/*
1291	 * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support
1292	 * Linux binaries.  If we see the call come through the FreeBSD ABI,
1293	 * return an error back to the user since we do not to support this.
1294	 */
1295	if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO ||
1296	    uap->cmd == SHM_STAT)
1297		return (EINVAL);
1298
1299	/* IPC_SET needs to copyin the buffer before calling kern_shmctl */
1300	if (uap->cmd == IPC_SET) {
1301		if ((error = copyin(uap->buf, &old, sizeof(old))))
1302			goto done;
1303		ipcperm_old2new(&old.shm_perm, &buf.shm_perm);
1304		CP(old, buf, shm_segsz);
1305		CP(old, buf, shm_lpid);
1306		CP(old, buf, shm_cpid);
1307		CP(old, buf, shm_nattch);
1308		CP(old, buf, shm_atime);
1309		CP(old, buf, shm_dtime);
1310		CP(old, buf, shm_ctime);
1311	}
1312
1313	error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz);
1314	if (error)
1315		goto done;
1316
1317	/* Cases in which we need to copyout */
1318	switch (uap->cmd) {
1319	case IPC_STAT:
1320		ipcperm_new2old(&buf.shm_perm, &old.shm_perm);
1321		if (buf.shm_segsz > INT_MAX)
1322			old.shm_segsz = INT_MAX;
1323		else
1324			CP(buf, old, shm_segsz);
1325		CP(buf, old, shm_lpid);
1326		CP(buf, old, shm_cpid);
1327		if (buf.shm_nattch > SHRT_MAX)
1328			old.shm_nattch = SHRT_MAX;
1329		else
1330			CP(buf, old, shm_nattch);
1331		CP(buf, old, shm_atime);
1332		CP(buf, old, shm_dtime);
1333		CP(buf, old, shm_ctime);
1334		old.shm_internal = NULL;
1335		error = copyout(&old, uap->buf, sizeof(old));
1336		break;
1337	}
1338
1339done:
1340	if (error) {
1341		/* Invalidate the return value */
1342		td->td_retval[0] = -1;
1343	}
1344	return (error);
1345}
1346
1347#endif	/* COMPAT_FREEBSD4 || COMPAT_FREEBSD5 || COMPAT_FREEBSD6 ||
1348	   COMPAT_FREEBSD7 */
1349
1350static int
1351sysvshm_modload(struct module *module, int cmd, void *arg)
1352{
1353	int error = 0;
1354
1355	switch (cmd) {
1356	case MOD_LOAD:
1357		error = shminit();
1358		if (error != 0)
1359			shmunload();
1360		break;
1361	case MOD_UNLOAD:
1362		error = shmunload();
1363		break;
1364	case MOD_SHUTDOWN:
1365		break;
1366	default:
1367		error = EINVAL;
1368		break;
1369	}
1370	return (error);
1371}
1372
1373static moduledata_t sysvshm_mod = {
1374	"sysvshm",
1375	&sysvshm_modload,
1376	NULL
1377};
1378
1379DECLARE_MODULE(sysvshm, sysvshm_mod, SI_SUB_SYSV_SHM, SI_ORDER_FIRST);
1380MODULE_VERSION(sysvshm, 1);
1381