sysv_shm.c revision 273707
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 273707 2014-10-26 19:42:44Z mjg $");
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/racct.h>
81#include <sys/resourcevar.h>
82#include <sys/rwlock.h>
83#include <sys/stat.h>
84#include <sys/syscall.h>
85#include <sys/syscallsubr.h>
86#include <sys/sysent.h>
87#include <sys/sysproto.h>
88#include <sys/jail.h>
89
90#include <security/mac/mac_framework.h>
91
92#include <vm/vm.h>
93#include <vm/vm_param.h>
94#include <vm/pmap.h>
95#include <vm/vm_object.h>
96#include <vm/vm_map.h>
97#include <vm/vm_page.h>
98#include <vm/vm_pager.h>
99
100FEATURE(sysv_shm, "System V shared memory segments support");
101
102static MALLOC_DEFINE(M_SHM, "shm", "SVID compatible shared memory segments");
103
104static int shmget_allocate_segment(struct thread *td,
105    struct shmget_args *uap, int mode);
106static int shmget_existing(struct thread *td, struct shmget_args *uap,
107    int mode, int segnum);
108
109#define	SHMSEG_FREE     	0x0200
110#define	SHMSEG_REMOVED  	0x0400
111#define	SHMSEG_ALLOCATED	0x0800
112#define	SHMSEG_WANTED		0x1000
113
114static int shm_last_free, shm_nused, shmalloced;
115vm_size_t shm_committed;
116static struct shmid_kernel	*shmsegs;
117
118struct shmmap_state {
119	vm_offset_t va;
120	int shmid;
121};
122
123static void shm_deallocate_segment(struct shmid_kernel *);
124static int shm_find_segment_by_key(key_t);
125static struct shmid_kernel *shm_find_segment_by_shmid(int);
126static struct shmid_kernel *shm_find_segment_by_shmidx(int);
127static int shm_delete_mapping(struct vmspace *vm, struct shmmap_state *);
128static void shmrealloc(void);
129static int shminit(void);
130static int sysvshm_modload(struct module *, int, void *);
131static int shmunload(void);
132static void shmexit_myhook(struct vmspace *vm);
133static void shmfork_myhook(struct proc *p1, struct proc *p2);
134static int sysctl_shmsegs(SYSCTL_HANDLER_ARGS);
135
136/*
137 * Tuneable values.
138 */
139#ifndef SHMMAXPGS
140#define	SHMMAXPGS	131072	/* Note: sysv shared memory is swap backed. */
141#endif
142#ifndef SHMMAX
143#define	SHMMAX	(SHMMAXPGS*PAGE_SIZE)
144#endif
145#ifndef SHMMIN
146#define	SHMMIN	1
147#endif
148#ifndef SHMMNI
149#define	SHMMNI	192
150#endif
151#ifndef SHMSEG
152#define	SHMSEG	128
153#endif
154#ifndef SHMALL
155#define	SHMALL	(SHMMAXPGS)
156#endif
157
158struct	shminfo shminfo = {
159	.shmmax = SHMMAX,
160	.shmmin = SHMMIN,
161	.shmmni = SHMMNI,
162	.shmseg = SHMSEG,
163	.shmall = SHMALL
164};
165
166static int shm_use_phys;
167static int shm_allow_removed;
168
169SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmax, CTLFLAG_RWTUN, &shminfo.shmmax, 0,
170    "Maximum shared memory segment size");
171SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmin, CTLFLAG_RWTUN, &shminfo.shmmin, 0,
172    "Minimum shared memory segment size");
173SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmni, CTLFLAG_RDTUN, &shminfo.shmmni, 0,
174    "Number of shared memory identifiers");
175SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmseg, CTLFLAG_RDTUN, &shminfo.shmseg, 0,
176    "Number of segments per process");
177SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmall, CTLFLAG_RWTUN, &shminfo.shmall, 0,
178    "Maximum number of pages available for shared memory");
179SYSCTL_INT(_kern_ipc, OID_AUTO, shm_use_phys, CTLFLAG_RWTUN,
180    &shm_use_phys, 0, "Enable/Disable locking of shared memory pages in core");
181SYSCTL_INT(_kern_ipc, OID_AUTO, shm_allow_removed, CTLFLAG_RWTUN,
182    &shm_allow_removed, 0,
183    "Enable/Disable attachment to attached segments marked for removal");
184SYSCTL_PROC(_kern_ipc, OID_AUTO, shmsegs, CTLTYPE_OPAQUE | CTLFLAG_RD,
185    NULL, 0, sysctl_shmsegs, "",
186    "Current number of shared memory segments allocated");
187
188static int
189shm_find_segment_by_key(key)
190	key_t key;
191{
192	int i;
193
194	for (i = 0; i < shmalloced; i++)
195		if ((shmsegs[i].u.shm_perm.mode & SHMSEG_ALLOCATED) &&
196		    shmsegs[i].u.shm_perm.key == key)
197			return (i);
198	return (-1);
199}
200
201static struct shmid_kernel *
202shm_find_segment_by_shmid(int shmid)
203{
204	int segnum;
205	struct shmid_kernel *shmseg;
206
207	segnum = IPCID_TO_IX(shmid);
208	if (segnum < 0 || segnum >= shmalloced)
209		return (NULL);
210	shmseg = &shmsegs[segnum];
211	if ((shmseg->u.shm_perm.mode & SHMSEG_ALLOCATED) == 0 ||
212	    (!shm_allow_removed &&
213	     (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) != 0) ||
214	    shmseg->u.shm_perm.seq != IPCID_TO_SEQ(shmid))
215		return (NULL);
216	return (shmseg);
217}
218
219static struct shmid_kernel *
220shm_find_segment_by_shmidx(int segnum)
221{
222	struct shmid_kernel *shmseg;
223
224	if (segnum < 0 || segnum >= shmalloced)
225		return (NULL);
226	shmseg = &shmsegs[segnum];
227	if ((shmseg->u.shm_perm.mode & SHMSEG_ALLOCATED) == 0 ||
228	    (!shm_allow_removed &&
229	     (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) != 0))
230		return (NULL);
231	return (shmseg);
232}
233
234static void
235shm_deallocate_segment(shmseg)
236	struct shmid_kernel *shmseg;
237{
238	vm_size_t size;
239
240	GIANT_REQUIRED;
241
242	vm_object_deallocate(shmseg->object);
243	shmseg->object = NULL;
244	size = round_page(shmseg->u.shm_segsz);
245	shm_committed -= btoc(size);
246	shm_nused--;
247	shmseg->u.shm_perm.mode = SHMSEG_FREE;
248#ifdef MAC
249	mac_sysvshm_cleanup(shmseg);
250#endif
251	racct_sub_cred(shmseg->cred, RACCT_NSHM, 1);
252	racct_sub_cred(shmseg->cred, RACCT_SHMSIZE, size);
253	crfree(shmseg->cred);
254	shmseg->cred = NULL;
255}
256
257static int
258shm_delete_mapping(struct vmspace *vm, struct shmmap_state *shmmap_s)
259{
260	struct shmid_kernel *shmseg;
261	int segnum, result;
262	vm_size_t size;
263
264	GIANT_REQUIRED;
265
266	segnum = IPCID_TO_IX(shmmap_s->shmid);
267	shmseg = &shmsegs[segnum];
268	size = round_page(shmseg->u.shm_segsz);
269	result = vm_map_remove(&vm->vm_map, shmmap_s->va, shmmap_s->va + size);
270	if (result != KERN_SUCCESS)
271		return (EINVAL);
272	shmmap_s->shmid = -1;
273	shmseg->u.shm_dtime = time_second;
274	if ((--shmseg->u.shm_nattch <= 0) &&
275	    (shmseg->u.shm_perm.mode & SHMSEG_REMOVED)) {
276		shm_deallocate_segment(shmseg);
277		shm_last_free = segnum;
278	}
279	return (0);
280}
281
282#ifndef _SYS_SYSPROTO_H_
283struct shmdt_args {
284	const void *shmaddr;
285};
286#endif
287int
288sys_shmdt(td, uap)
289	struct thread *td;
290	struct shmdt_args *uap;
291{
292	struct proc *p = td->td_proc;
293	struct shmmap_state *shmmap_s;
294#ifdef MAC
295	struct shmid_kernel *shmsegptr;
296#endif
297	int i;
298	int error = 0;
299
300	if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
301		return (ENOSYS);
302	mtx_lock(&Giant);
303	shmmap_s = p->p_vmspace->vm_shm;
304 	if (shmmap_s == NULL) {
305		error = EINVAL;
306		goto done2;
307	}
308	for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) {
309		if (shmmap_s->shmid != -1 &&
310		    shmmap_s->va == (vm_offset_t)uap->shmaddr) {
311			break;
312		}
313	}
314	if (i == shminfo.shmseg) {
315		error = EINVAL;
316		goto done2;
317	}
318#ifdef MAC
319	shmsegptr = &shmsegs[IPCID_TO_IX(shmmap_s->shmid)];
320	error = mac_sysvshm_check_shmdt(td->td_ucred, shmsegptr);
321	if (error != 0)
322		goto done2;
323#endif
324	error = shm_delete_mapping(p->p_vmspace, shmmap_s);
325done2:
326	mtx_unlock(&Giant);
327	return (error);
328}
329
330#ifndef _SYS_SYSPROTO_H_
331struct shmat_args {
332	int shmid;
333	const void *shmaddr;
334	int shmflg;
335};
336#endif
337int
338kern_shmat(td, shmid, shmaddr, shmflg)
339	struct thread *td;
340	int shmid;
341	const void *shmaddr;
342	int shmflg;
343{
344	struct proc *p = td->td_proc;
345	int i;
346	struct shmid_kernel *shmseg;
347	struct shmmap_state *shmmap_s = NULL;
348	vm_offset_t attach_va;
349	vm_prot_t prot;
350	vm_size_t size;
351	int rv;
352	int error = 0;
353
354	if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
355		return (ENOSYS);
356	mtx_lock(&Giant);
357	shmmap_s = p->p_vmspace->vm_shm;
358	if (shmmap_s == NULL) {
359		shmmap_s = malloc(shminfo.shmseg * sizeof(struct shmmap_state),
360		    M_SHM, M_WAITOK);
361		for (i = 0; i < shminfo.shmseg; i++)
362			shmmap_s[i].shmid = -1;
363		p->p_vmspace->vm_shm = shmmap_s;
364	}
365	shmseg = shm_find_segment_by_shmid(shmid);
366	if (shmseg == NULL) {
367		error = EINVAL;
368		goto done2;
369	}
370	error = ipcperm(td, &shmseg->u.shm_perm,
371	    (shmflg & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W);
372	if (error)
373		goto done2;
374#ifdef MAC
375	error = mac_sysvshm_check_shmat(td->td_ucred, shmseg, shmflg);
376	if (error != 0)
377		goto done2;
378#endif
379	for (i = 0; i < shminfo.shmseg; i++) {
380		if (shmmap_s->shmid == -1)
381			break;
382		shmmap_s++;
383	}
384	if (i >= shminfo.shmseg) {
385		error = EMFILE;
386		goto done2;
387	}
388	size = round_page(shmseg->u.shm_segsz);
389	prot = VM_PROT_READ;
390	if ((shmflg & SHM_RDONLY) == 0)
391		prot |= VM_PROT_WRITE;
392	if (shmaddr) {
393		if (shmflg & SHM_RND) {
394			attach_va = (vm_offset_t)shmaddr & ~(SHMLBA-1);
395		} else if (((vm_offset_t)shmaddr & (SHMLBA-1)) == 0) {
396			attach_va = (vm_offset_t)shmaddr;
397		} else {
398			error = EINVAL;
399			goto done2;
400		}
401	} else {
402		/*
403		 * This is just a hint to vm_map_find() about where to
404		 * put it.
405		 */
406		PROC_LOCK(p);
407		attach_va = round_page((vm_offset_t)p->p_vmspace->vm_daddr +
408		    lim_max(p, RLIMIT_DATA));
409		PROC_UNLOCK(p);
410	}
411
412	vm_object_reference(shmseg->object);
413	rv = vm_map_find(&p->p_vmspace->vm_map, shmseg->object, 0, &attach_va,
414	    size, 0, shmaddr != NULL ? VMFS_NO_SPACE : VMFS_OPTIMAL_SPACE,
415	    prot, prot, MAP_INHERIT_SHARE | MAP_PREFAULT_PARTIAL);
416	if (rv != KERN_SUCCESS) {
417		vm_object_deallocate(shmseg->object);
418		error = ENOMEM;
419		goto done2;
420	}
421
422	shmmap_s->va = attach_va;
423	shmmap_s->shmid = shmid;
424	shmseg->u.shm_lpid = p->p_pid;
425	shmseg->u.shm_atime = time_second;
426	shmseg->u.shm_nattch++;
427	td->td_retval[0] = attach_va;
428done2:
429	mtx_unlock(&Giant);
430	return (error);
431}
432
433int
434sys_shmat(td, uap)
435	struct thread *td;
436	struct shmat_args *uap;
437{
438	return kern_shmat(td, uap->shmid, uap->shmaddr, uap->shmflg);
439}
440
441int
442kern_shmctl(td, shmid, cmd, buf, bufsz)
443	struct thread *td;
444	int shmid;
445	int cmd;
446	void *buf;
447	size_t *bufsz;
448{
449	int error = 0;
450	struct shmid_kernel *shmseg;
451
452	if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
453		return (ENOSYS);
454
455	mtx_lock(&Giant);
456	switch (cmd) {
457	/*
458	 * It is possible that kern_shmctl is being called from the Linux ABI
459	 * layer, in which case, we will need to implement IPC_INFO.  It should
460	 * be noted that other shmctl calls will be funneled through here for
461	 * Linix binaries as well.
462	 *
463	 * NB: The Linux ABI layer will convert this data to structure(s) more
464	 * consistent with the Linux ABI.
465	 */
466	case IPC_INFO:
467		memcpy(buf, &shminfo, sizeof(shminfo));
468		if (bufsz)
469			*bufsz = sizeof(shminfo);
470		td->td_retval[0] = shmalloced;
471		goto done2;
472	case SHM_INFO: {
473		struct shm_info shm_info;
474		shm_info.used_ids = shm_nused;
475		shm_info.shm_rss = 0;	/*XXX where to get from ? */
476		shm_info.shm_tot = 0;	/*XXX where to get from ? */
477		shm_info.shm_swp = 0;	/*XXX where to get from ? */
478		shm_info.swap_attempts = 0;	/*XXX where to get from ? */
479		shm_info.swap_successes = 0;	/*XXX where to get from ? */
480		memcpy(buf, &shm_info, sizeof(shm_info));
481		if (bufsz)
482			*bufsz = sizeof(shm_info);
483		td->td_retval[0] = shmalloced;
484		goto done2;
485	}
486	}
487	if (cmd == SHM_STAT)
488		shmseg = shm_find_segment_by_shmidx(shmid);
489	else
490		shmseg = shm_find_segment_by_shmid(shmid);
491	if (shmseg == NULL) {
492		error = EINVAL;
493		goto done2;
494	}
495#ifdef MAC
496	error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, cmd);
497	if (error != 0)
498		goto done2;
499#endif
500	switch (cmd) {
501	case SHM_STAT:
502	case IPC_STAT:
503		error = ipcperm(td, &shmseg->u.shm_perm, IPC_R);
504		if (error)
505			goto done2;
506		memcpy(buf, &shmseg->u, sizeof(struct shmid_ds));
507		if (bufsz)
508			*bufsz = sizeof(struct shmid_ds);
509		if (cmd == SHM_STAT)
510			td->td_retval[0] = IXSEQ_TO_IPCID(shmid, shmseg->u.shm_perm);
511		break;
512	case IPC_SET: {
513		struct shmid_ds *shmid;
514
515		shmid = (struct shmid_ds *)buf;
516		error = ipcperm(td, &shmseg->u.shm_perm, IPC_M);
517		if (error)
518			goto done2;
519		shmseg->u.shm_perm.uid = shmid->shm_perm.uid;
520		shmseg->u.shm_perm.gid = shmid->shm_perm.gid;
521		shmseg->u.shm_perm.mode =
522		    (shmseg->u.shm_perm.mode & ~ACCESSPERMS) |
523		    (shmid->shm_perm.mode & ACCESSPERMS);
524		shmseg->u.shm_ctime = time_second;
525		break;
526	}
527	case IPC_RMID:
528		error = ipcperm(td, &shmseg->u.shm_perm, IPC_M);
529		if (error)
530			goto done2;
531		shmseg->u.shm_perm.key = IPC_PRIVATE;
532		shmseg->u.shm_perm.mode |= SHMSEG_REMOVED;
533		if (shmseg->u.shm_nattch <= 0) {
534			shm_deallocate_segment(shmseg);
535			shm_last_free = IPCID_TO_IX(shmid);
536		}
537		break;
538#if 0
539	case SHM_LOCK:
540	case SHM_UNLOCK:
541#endif
542	default:
543		error = EINVAL;
544		break;
545	}
546done2:
547	mtx_unlock(&Giant);
548	return (error);
549}
550
551#ifndef _SYS_SYSPROTO_H_
552struct shmctl_args {
553	int shmid;
554	int cmd;
555	struct shmid_ds *buf;
556};
557#endif
558int
559sys_shmctl(td, uap)
560	struct thread *td;
561	struct shmctl_args *uap;
562{
563	int error = 0;
564	struct shmid_ds buf;
565	size_t bufsz;
566
567	/*
568	 * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support
569	 * Linux binaries.  If we see the call come through the FreeBSD ABI,
570	 * return an error back to the user since we do not to support this.
571	 */
572	if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO ||
573	    uap->cmd == SHM_STAT)
574		return (EINVAL);
575
576	/* IPC_SET needs to copyin the buffer before calling kern_shmctl */
577	if (uap->cmd == IPC_SET) {
578		if ((error = copyin(uap->buf, &buf, sizeof(struct shmid_ds))))
579			goto done;
580	}
581
582	error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz);
583	if (error)
584		goto done;
585
586	/* Cases in which we need to copyout */
587	switch (uap->cmd) {
588	case IPC_STAT:
589		error = copyout(&buf, uap->buf, bufsz);
590		break;
591	}
592
593done:
594	if (error) {
595		/* Invalidate the return value */
596		td->td_retval[0] = -1;
597	}
598	return (error);
599}
600
601
602static int
603shmget_existing(td, uap, mode, segnum)
604	struct thread *td;
605	struct shmget_args *uap;
606	int mode;
607	int segnum;
608{
609	struct shmid_kernel *shmseg;
610	int error;
611
612	shmseg = &shmsegs[segnum];
613	if (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) {
614		/*
615		 * This segment is in the process of being allocated.  Wait
616		 * until it's done, and look the key up again (in case the
617		 * allocation failed or it was freed).
618		 */
619		shmseg->u.shm_perm.mode |= SHMSEG_WANTED;
620		error = tsleep(shmseg, PLOCK | PCATCH, "shmget", 0);
621		if (error)
622			return (error);
623		return (EAGAIN);
624	}
625	if ((uap->shmflg & (IPC_CREAT | IPC_EXCL)) == (IPC_CREAT | IPC_EXCL))
626		return (EEXIST);
627#ifdef MAC
628	error = mac_sysvshm_check_shmget(td->td_ucred, shmseg, uap->shmflg);
629	if (error != 0)
630		return (error);
631#endif
632	if (uap->size != 0 && uap->size > shmseg->u.shm_segsz)
633		return (EINVAL);
634	td->td_retval[0] = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm);
635	return (0);
636}
637
638static int
639shmget_allocate_segment(td, uap, mode)
640	struct thread *td;
641	struct shmget_args *uap;
642	int mode;
643{
644	int i, segnum, shmid;
645	size_t size;
646	struct ucred *cred = td->td_ucred;
647	struct shmid_kernel *shmseg;
648	vm_object_t shm_object;
649
650	GIANT_REQUIRED;
651
652	if (uap->size < shminfo.shmmin || uap->size > shminfo.shmmax)
653		return (EINVAL);
654	if (shm_nused >= shminfo.shmmni) /* Any shmids left? */
655		return (ENOSPC);
656	size = round_page(uap->size);
657	if (shm_committed + btoc(size) > shminfo.shmall)
658		return (ENOMEM);
659	if (shm_last_free < 0) {
660		shmrealloc();	/* Maybe expand the shmsegs[] array. */
661		for (i = 0; i < shmalloced; i++)
662			if (shmsegs[i].u.shm_perm.mode & SHMSEG_FREE)
663				break;
664		if (i == shmalloced)
665			return (ENOSPC);
666		segnum = i;
667	} else  {
668		segnum = shm_last_free;
669		shm_last_free = -1;
670	}
671	shmseg = &shmsegs[segnum];
672#ifdef RACCT
673	PROC_LOCK(td->td_proc);
674	if (racct_add(td->td_proc, RACCT_NSHM, 1)) {
675		PROC_UNLOCK(td->td_proc);
676		return (ENOSPC);
677	}
678	if (racct_add(td->td_proc, RACCT_SHMSIZE, size)) {
679		racct_sub(td->td_proc, RACCT_NSHM, 1);
680		PROC_UNLOCK(td->td_proc);
681		return (ENOMEM);
682	}
683	PROC_UNLOCK(td->td_proc);
684#endif
685	/*
686	 * In case we sleep in malloc(), mark the segment present but deleted
687	 * so that noone else tries to create the same key.
688	 */
689	shmseg->u.shm_perm.mode = SHMSEG_ALLOCATED | SHMSEG_REMOVED;
690	shmseg->u.shm_perm.key = uap->key;
691	shmseg->u.shm_perm.seq = (shmseg->u.shm_perm.seq + 1) & 0x7fff;
692	shmid = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm);
693
694	/*
695	 * We make sure that we have allocated a pager before we need
696	 * to.
697	 */
698	shm_object = vm_pager_allocate(shm_use_phys ? OBJT_PHYS : OBJT_SWAP,
699	    0, size, VM_PROT_DEFAULT, 0, cred);
700	if (shm_object == NULL) {
701#ifdef RACCT
702		PROC_LOCK(td->td_proc);
703		racct_sub(td->td_proc, RACCT_NSHM, 1);
704		racct_sub(td->td_proc, RACCT_SHMSIZE, size);
705		PROC_UNLOCK(td->td_proc);
706#endif
707		return (ENOMEM);
708	}
709	VM_OBJECT_WLOCK(shm_object);
710	vm_object_clear_flag(shm_object, OBJ_ONEMAPPING);
711	vm_object_set_flag(shm_object, OBJ_NOSPLIT);
712	VM_OBJECT_WUNLOCK(shm_object);
713
714	shmseg->object = shm_object;
715	shmseg->u.shm_perm.cuid = shmseg->u.shm_perm.uid = cred->cr_uid;
716	shmseg->u.shm_perm.cgid = shmseg->u.shm_perm.gid = cred->cr_gid;
717	shmseg->u.shm_perm.mode = (shmseg->u.shm_perm.mode & SHMSEG_WANTED) |
718	    (mode & ACCESSPERMS) | SHMSEG_ALLOCATED;
719	shmseg->cred = crhold(cred);
720	shmseg->u.shm_segsz = uap->size;
721	shmseg->u.shm_cpid = td->td_proc->p_pid;
722	shmseg->u.shm_lpid = shmseg->u.shm_nattch = 0;
723	shmseg->u.shm_atime = shmseg->u.shm_dtime = 0;
724#ifdef MAC
725	mac_sysvshm_create(cred, shmseg);
726#endif
727	shmseg->u.shm_ctime = time_second;
728	shm_committed += btoc(size);
729	shm_nused++;
730	if (shmseg->u.shm_perm.mode & SHMSEG_WANTED) {
731		/*
732		 * Somebody else wanted this key while we were asleep.  Wake
733		 * them up now.
734		 */
735		shmseg->u.shm_perm.mode &= ~SHMSEG_WANTED;
736		wakeup(shmseg);
737	}
738	td->td_retval[0] = shmid;
739	return (0);
740}
741
742#ifndef _SYS_SYSPROTO_H_
743struct shmget_args {
744	key_t key;
745	size_t size;
746	int shmflg;
747};
748#endif
749int
750sys_shmget(td, uap)
751	struct thread *td;
752	struct shmget_args *uap;
753{
754	int segnum, mode;
755	int error;
756
757	if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
758		return (ENOSYS);
759	mtx_lock(&Giant);
760	mode = uap->shmflg & ACCESSPERMS;
761	if (uap->key != IPC_PRIVATE) {
762	again:
763		segnum = shm_find_segment_by_key(uap->key);
764		if (segnum >= 0) {
765			error = shmget_existing(td, uap, mode, segnum);
766			if (error == EAGAIN)
767				goto again;
768			goto done2;
769		}
770		if ((uap->shmflg & IPC_CREAT) == 0) {
771			error = ENOENT;
772			goto done2;
773		}
774	}
775	error = shmget_allocate_segment(td, uap, mode);
776done2:
777	mtx_unlock(&Giant);
778	return (error);
779}
780
781static void
782shmfork_myhook(p1, p2)
783	struct proc *p1, *p2;
784{
785	struct shmmap_state *shmmap_s;
786	size_t size;
787	int i;
788
789	mtx_lock(&Giant);
790	size = shminfo.shmseg * sizeof(struct shmmap_state);
791	shmmap_s = malloc(size, M_SHM, M_WAITOK);
792	bcopy(p1->p_vmspace->vm_shm, shmmap_s, size);
793	p2->p_vmspace->vm_shm = shmmap_s;
794	for (i = 0; i < shminfo.shmseg; i++, shmmap_s++)
795		if (shmmap_s->shmid != -1)
796			shmsegs[IPCID_TO_IX(shmmap_s->shmid)].u.shm_nattch++;
797	mtx_unlock(&Giant);
798}
799
800static void
801shmexit_myhook(struct vmspace *vm)
802{
803	struct shmmap_state *base, *shm;
804	int i;
805
806	if ((base = vm->vm_shm) != NULL) {
807		vm->vm_shm = NULL;
808		mtx_lock(&Giant);
809		for (i = 0, shm = base; i < shminfo.shmseg; i++, shm++) {
810			if (shm->shmid != -1)
811				shm_delete_mapping(vm, shm);
812		}
813		mtx_unlock(&Giant);
814		free(base, M_SHM);
815	}
816}
817
818static void
819shmrealloc(void)
820{
821	int i;
822	struct shmid_kernel *newsegs;
823
824	if (shmalloced >= shminfo.shmmni)
825		return;
826
827	newsegs = malloc(shminfo.shmmni * sizeof(*newsegs), M_SHM, M_WAITOK);
828	if (newsegs == NULL)
829		return;
830	for (i = 0; i < shmalloced; i++)
831		bcopy(&shmsegs[i], &newsegs[i], sizeof(newsegs[0]));
832	for (; i < shminfo.shmmni; i++) {
833		shmsegs[i].u.shm_perm.mode = SHMSEG_FREE;
834		shmsegs[i].u.shm_perm.seq = 0;
835#ifdef MAC
836		mac_sysvshm_init(&shmsegs[i]);
837#endif
838	}
839	free(shmsegs, M_SHM);
840	shmsegs = newsegs;
841	shmalloced = shminfo.shmmni;
842}
843
844static struct syscall_helper_data shm_syscalls[] = {
845	SYSCALL_INIT_HELPER(shmat),
846	SYSCALL_INIT_HELPER(shmctl),
847	SYSCALL_INIT_HELPER(shmdt),
848	SYSCALL_INIT_HELPER(shmget),
849#if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
850    defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
851	SYSCALL_INIT_HELPER_COMPAT(freebsd7_shmctl),
852#endif
853#if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43))
854	SYSCALL_INIT_HELPER(shmsys),
855#endif
856	SYSCALL_INIT_LAST
857};
858
859#ifdef COMPAT_FREEBSD32
860#include <compat/freebsd32/freebsd32.h>
861#include <compat/freebsd32/freebsd32_ipc.h>
862#include <compat/freebsd32/freebsd32_proto.h>
863#include <compat/freebsd32/freebsd32_signal.h>
864#include <compat/freebsd32/freebsd32_syscall.h>
865#include <compat/freebsd32/freebsd32_util.h>
866
867static struct syscall_helper_data shm32_syscalls[] = {
868	SYSCALL32_INIT_HELPER_COMPAT(shmat),
869	SYSCALL32_INIT_HELPER_COMPAT(shmdt),
870	SYSCALL32_INIT_HELPER_COMPAT(shmget),
871	SYSCALL32_INIT_HELPER(freebsd32_shmsys),
872	SYSCALL32_INIT_HELPER(freebsd32_shmctl),
873#if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
874    defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
875	SYSCALL32_INIT_HELPER(freebsd7_freebsd32_shmctl),
876#endif
877	SYSCALL_INIT_LAST
878};
879#endif
880
881static int
882shminit()
883{
884	int i, error;
885
886#ifndef BURN_BRIDGES
887	if (TUNABLE_ULONG_FETCH("kern.ipc.shmmaxpgs", &shminfo.shmall) != 0)
888		printf("kern.ipc.shmmaxpgs is now called kern.ipc.shmall!\n");
889#endif
890	if (shminfo.shmmax == SHMMAX) {
891		/* Initialize shmmax dealing with possible overflow. */
892		for (i = PAGE_SIZE; i != 0; i--) {
893			shminfo.shmmax = shminfo.shmall * i;
894			if ((shminfo.shmmax / shminfo.shmall) == (u_long)i)
895				break;
896		}
897	}
898	shmalloced = shminfo.shmmni;
899	shmsegs = malloc(shmalloced * sizeof(shmsegs[0]), M_SHM, M_WAITOK);
900	for (i = 0; i < shmalloced; i++) {
901		shmsegs[i].u.shm_perm.mode = SHMSEG_FREE;
902		shmsegs[i].u.shm_perm.seq = 0;
903#ifdef MAC
904		mac_sysvshm_init(&shmsegs[i]);
905#endif
906	}
907	shm_last_free = 0;
908	shm_nused = 0;
909	shm_committed = 0;
910	shmexit_hook = &shmexit_myhook;
911	shmfork_hook = &shmfork_myhook;
912
913	error = syscall_helper_register(shm_syscalls, SY_THR_STATIC_KLD);
914	if (error != 0)
915		return (error);
916#ifdef COMPAT_FREEBSD32
917	error = syscall32_helper_register(shm32_syscalls, SY_THR_STATIC_KLD);
918	if (error != 0)
919		return (error);
920#endif
921	return (0);
922}
923
924static int
925shmunload()
926{
927	int i;
928
929	if (shm_nused > 0)
930		return (EBUSY);
931
932#ifdef COMPAT_FREEBSD32
933	syscall32_helper_unregister(shm32_syscalls);
934#endif
935	syscall_helper_unregister(shm_syscalls);
936
937	for (i = 0; i < shmalloced; i++) {
938#ifdef MAC
939		mac_sysvshm_destroy(&shmsegs[i]);
940#endif
941		/*
942		 * Objects might be still mapped into the processes
943		 * address spaces.  Actual free would happen on the
944		 * last mapping destruction.
945		 */
946		if (shmsegs[i].u.shm_perm.mode != SHMSEG_FREE)
947			vm_object_deallocate(shmsegs[i].object);
948	}
949	free(shmsegs, M_SHM);
950	shmexit_hook = NULL;
951	shmfork_hook = NULL;
952	return (0);
953}
954
955static int
956sysctl_shmsegs(SYSCTL_HANDLER_ARGS)
957{
958
959	return (SYSCTL_OUT(req, shmsegs, shmalloced * sizeof(shmsegs[0])));
960}
961
962#if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43))
963struct oshmid_ds {
964	struct	ipc_perm_old shm_perm;	/* operation perms */
965	int	shm_segsz;		/* size of segment (bytes) */
966	u_short	shm_cpid;		/* pid, creator */
967	u_short	shm_lpid;		/* pid, last operation */
968	short	shm_nattch;		/* no. of current attaches */
969	time_t	shm_atime;		/* last attach time */
970	time_t	shm_dtime;		/* last detach time */
971	time_t	shm_ctime;		/* last change time */
972	void	*shm_handle;		/* internal handle for shm segment */
973};
974
975struct oshmctl_args {
976	int shmid;
977	int cmd;
978	struct oshmid_ds *ubuf;
979};
980
981static int
982oshmctl(struct thread *td, struct oshmctl_args *uap)
983{
984#ifdef COMPAT_43
985	int error = 0;
986	struct shmid_kernel *shmseg;
987	struct oshmid_ds outbuf;
988
989	if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
990		return (ENOSYS);
991	mtx_lock(&Giant);
992	shmseg = shm_find_segment_by_shmid(uap->shmid);
993	if (shmseg == NULL) {
994		error = EINVAL;
995		goto done2;
996	}
997	switch (uap->cmd) {
998	case IPC_STAT:
999		error = ipcperm(td, &shmseg->u.shm_perm, IPC_R);
1000		if (error)
1001			goto done2;
1002#ifdef MAC
1003		error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, uap->cmd);
1004		if (error != 0)
1005			goto done2;
1006#endif
1007		ipcperm_new2old(&shmseg->u.shm_perm, &outbuf.shm_perm);
1008		outbuf.shm_segsz = shmseg->u.shm_segsz;
1009		outbuf.shm_cpid = shmseg->u.shm_cpid;
1010		outbuf.shm_lpid = shmseg->u.shm_lpid;
1011		outbuf.shm_nattch = shmseg->u.shm_nattch;
1012		outbuf.shm_atime = shmseg->u.shm_atime;
1013		outbuf.shm_dtime = shmseg->u.shm_dtime;
1014		outbuf.shm_ctime = shmseg->u.shm_ctime;
1015		outbuf.shm_handle = shmseg->object;
1016		error = copyout(&outbuf, uap->ubuf, sizeof(outbuf));
1017		if (error)
1018			goto done2;
1019		break;
1020	default:
1021		error = freebsd7_shmctl(td, (struct freebsd7_shmctl_args *)uap);
1022		break;
1023	}
1024done2:
1025	mtx_unlock(&Giant);
1026	return (error);
1027#else
1028	return (EINVAL);
1029#endif
1030}
1031
1032/* XXX casting to (sy_call_t *) is bogus, as usual. */
1033static sy_call_t *shmcalls[] = {
1034	(sy_call_t *)sys_shmat, (sy_call_t *)oshmctl,
1035	(sy_call_t *)sys_shmdt, (sy_call_t *)sys_shmget,
1036	(sy_call_t *)freebsd7_shmctl
1037};
1038
1039int
1040sys_shmsys(td, uap)
1041	struct thread *td;
1042	/* XXX actually varargs. */
1043	struct shmsys_args /* {
1044		int	which;
1045		int	a2;
1046		int	a3;
1047		int	a4;
1048	} */ *uap;
1049{
1050	int error;
1051
1052	if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
1053		return (ENOSYS);
1054	if (uap->which < 0 ||
1055	    uap->which >= sizeof(shmcalls)/sizeof(shmcalls[0]))
1056		return (EINVAL);
1057	mtx_lock(&Giant);
1058	error = (*shmcalls[uap->which])(td, &uap->a2);
1059	mtx_unlock(&Giant);
1060	return (error);
1061}
1062
1063#endif	/* i386 && (COMPAT_FREEBSD4 || COMPAT_43) */
1064
1065#ifdef COMPAT_FREEBSD32
1066
1067int
1068freebsd32_shmsys(struct thread *td, struct freebsd32_shmsys_args *uap)
1069{
1070
1071#if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1072    defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1073	switch (uap->which) {
1074	case 0:	{	/* shmat */
1075		struct shmat_args ap;
1076
1077		ap.shmid = uap->a2;
1078		ap.shmaddr = PTRIN(uap->a3);
1079		ap.shmflg = uap->a4;
1080		return (sysent[SYS_shmat].sy_call(td, &ap));
1081	}
1082	case 2: {	/* shmdt */
1083		struct shmdt_args ap;
1084
1085		ap.shmaddr = PTRIN(uap->a2);
1086		return (sysent[SYS_shmdt].sy_call(td, &ap));
1087	}
1088	case 3: {	/* shmget */
1089		struct shmget_args ap;
1090
1091		ap.key = uap->a2;
1092		ap.size = uap->a3;
1093		ap.shmflg = uap->a4;
1094		return (sysent[SYS_shmget].sy_call(td, &ap));
1095	}
1096	case 4: {	/* shmctl */
1097		struct freebsd7_freebsd32_shmctl_args ap;
1098
1099		ap.shmid = uap->a2;
1100		ap.cmd = uap->a3;
1101		ap.buf = PTRIN(uap->a4);
1102		return (freebsd7_freebsd32_shmctl(td, &ap));
1103	}
1104	case 1:		/* oshmctl */
1105	default:
1106		return (EINVAL);
1107	}
1108#else
1109	return (nosys(td, NULL));
1110#endif
1111}
1112
1113#if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1114    defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1115int
1116freebsd7_freebsd32_shmctl(struct thread *td,
1117    struct freebsd7_freebsd32_shmctl_args *uap)
1118{
1119	int error = 0;
1120	union {
1121		struct shmid_ds shmid_ds;
1122		struct shm_info shm_info;
1123		struct shminfo shminfo;
1124	} u;
1125	union {
1126		struct shmid_ds32_old shmid_ds32;
1127		struct shm_info32 shm_info32;
1128		struct shminfo32 shminfo32;
1129	} u32;
1130	size_t sz;
1131
1132	if (uap->cmd == IPC_SET) {
1133		if ((error = copyin(uap->buf, &u32.shmid_ds32,
1134		    sizeof(u32.shmid_ds32))))
1135			goto done;
1136		freebsd32_ipcperm_old_in(&u32.shmid_ds32.shm_perm,
1137		    &u.shmid_ds.shm_perm);
1138		CP(u32.shmid_ds32, u.shmid_ds, shm_segsz);
1139		CP(u32.shmid_ds32, u.shmid_ds, shm_lpid);
1140		CP(u32.shmid_ds32, u.shmid_ds, shm_cpid);
1141		CP(u32.shmid_ds32, u.shmid_ds, shm_nattch);
1142		CP(u32.shmid_ds32, u.shmid_ds, shm_atime);
1143		CP(u32.shmid_ds32, u.shmid_ds, shm_dtime);
1144		CP(u32.shmid_ds32, u.shmid_ds, shm_ctime);
1145	}
1146
1147	error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz);
1148	if (error)
1149		goto done;
1150
1151	/* Cases in which we need to copyout */
1152	switch (uap->cmd) {
1153	case IPC_INFO:
1154		CP(u.shminfo, u32.shminfo32, shmmax);
1155		CP(u.shminfo, u32.shminfo32, shmmin);
1156		CP(u.shminfo, u32.shminfo32, shmmni);
1157		CP(u.shminfo, u32.shminfo32, shmseg);
1158		CP(u.shminfo, u32.shminfo32, shmall);
1159		error = copyout(&u32.shminfo32, uap->buf,
1160		    sizeof(u32.shminfo32));
1161		break;
1162	case SHM_INFO:
1163		CP(u.shm_info, u32.shm_info32, used_ids);
1164		CP(u.shm_info, u32.shm_info32, shm_rss);
1165		CP(u.shm_info, u32.shm_info32, shm_tot);
1166		CP(u.shm_info, u32.shm_info32, shm_swp);
1167		CP(u.shm_info, u32.shm_info32, swap_attempts);
1168		CP(u.shm_info, u32.shm_info32, swap_successes);
1169		error = copyout(&u32.shm_info32, uap->buf,
1170		    sizeof(u32.shm_info32));
1171		break;
1172	case SHM_STAT:
1173	case IPC_STAT:
1174		freebsd32_ipcperm_old_out(&u.shmid_ds.shm_perm,
1175		    &u32.shmid_ds32.shm_perm);
1176		if (u.shmid_ds.shm_segsz > INT32_MAX)
1177			u32.shmid_ds32.shm_segsz = INT32_MAX;
1178		else
1179			CP(u.shmid_ds, u32.shmid_ds32, shm_segsz);
1180		CP(u.shmid_ds, u32.shmid_ds32, shm_lpid);
1181		CP(u.shmid_ds, u32.shmid_ds32, shm_cpid);
1182		CP(u.shmid_ds, u32.shmid_ds32, shm_nattch);
1183		CP(u.shmid_ds, u32.shmid_ds32, shm_atime);
1184		CP(u.shmid_ds, u32.shmid_ds32, shm_dtime);
1185		CP(u.shmid_ds, u32.shmid_ds32, shm_ctime);
1186		u32.shmid_ds32.shm_internal = 0;
1187		error = copyout(&u32.shmid_ds32, uap->buf,
1188		    sizeof(u32.shmid_ds32));
1189		break;
1190	}
1191
1192done:
1193	if (error) {
1194		/* Invalidate the return value */
1195		td->td_retval[0] = -1;
1196	}
1197	return (error);
1198}
1199#endif
1200
1201int
1202freebsd32_shmctl(struct thread *td, struct freebsd32_shmctl_args *uap)
1203{
1204	int error = 0;
1205	union {
1206		struct shmid_ds shmid_ds;
1207		struct shm_info shm_info;
1208		struct shminfo shminfo;
1209	} u;
1210	union {
1211		struct shmid_ds32 shmid_ds32;
1212		struct shm_info32 shm_info32;
1213		struct shminfo32 shminfo32;
1214	} u32;
1215	size_t sz;
1216
1217	if (uap->cmd == IPC_SET) {
1218		if ((error = copyin(uap->buf, &u32.shmid_ds32,
1219		    sizeof(u32.shmid_ds32))))
1220			goto done;
1221		freebsd32_ipcperm_in(&u32.shmid_ds32.shm_perm,
1222		    &u.shmid_ds.shm_perm);
1223		CP(u32.shmid_ds32, u.shmid_ds, shm_segsz);
1224		CP(u32.shmid_ds32, u.shmid_ds, shm_lpid);
1225		CP(u32.shmid_ds32, u.shmid_ds, shm_cpid);
1226		CP(u32.shmid_ds32, u.shmid_ds, shm_nattch);
1227		CP(u32.shmid_ds32, u.shmid_ds, shm_atime);
1228		CP(u32.shmid_ds32, u.shmid_ds, shm_dtime);
1229		CP(u32.shmid_ds32, u.shmid_ds, shm_ctime);
1230	}
1231
1232	error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz);
1233	if (error)
1234		goto done;
1235
1236	/* Cases in which we need to copyout */
1237	switch (uap->cmd) {
1238	case IPC_INFO:
1239		CP(u.shminfo, u32.shminfo32, shmmax);
1240		CP(u.shminfo, u32.shminfo32, shmmin);
1241		CP(u.shminfo, u32.shminfo32, shmmni);
1242		CP(u.shminfo, u32.shminfo32, shmseg);
1243		CP(u.shminfo, u32.shminfo32, shmall);
1244		error = copyout(&u32.shminfo32, uap->buf,
1245		    sizeof(u32.shminfo32));
1246		break;
1247	case SHM_INFO:
1248		CP(u.shm_info, u32.shm_info32, used_ids);
1249		CP(u.shm_info, u32.shm_info32, shm_rss);
1250		CP(u.shm_info, u32.shm_info32, shm_tot);
1251		CP(u.shm_info, u32.shm_info32, shm_swp);
1252		CP(u.shm_info, u32.shm_info32, swap_attempts);
1253		CP(u.shm_info, u32.shm_info32, swap_successes);
1254		error = copyout(&u32.shm_info32, uap->buf,
1255		    sizeof(u32.shm_info32));
1256		break;
1257	case SHM_STAT:
1258	case IPC_STAT:
1259		freebsd32_ipcperm_out(&u.shmid_ds.shm_perm,
1260		    &u32.shmid_ds32.shm_perm);
1261		if (u.shmid_ds.shm_segsz > INT32_MAX)
1262			u32.shmid_ds32.shm_segsz = INT32_MAX;
1263		else
1264			CP(u.shmid_ds, u32.shmid_ds32, shm_segsz);
1265		CP(u.shmid_ds, u32.shmid_ds32, shm_lpid);
1266		CP(u.shmid_ds, u32.shmid_ds32, shm_cpid);
1267		CP(u.shmid_ds, u32.shmid_ds32, shm_nattch);
1268		CP(u.shmid_ds, u32.shmid_ds32, shm_atime);
1269		CP(u.shmid_ds, u32.shmid_ds32, shm_dtime);
1270		CP(u.shmid_ds, u32.shmid_ds32, shm_ctime);
1271		error = copyout(&u32.shmid_ds32, uap->buf,
1272		    sizeof(u32.shmid_ds32));
1273		break;
1274	}
1275
1276done:
1277	if (error) {
1278		/* Invalidate the return value */
1279		td->td_retval[0] = -1;
1280	}
1281	return (error);
1282}
1283#endif
1284
1285#if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1286    defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1287
1288#ifndef CP
1289#define CP(src, dst, fld)	do { (dst).fld = (src).fld; } while (0)
1290#endif
1291
1292#ifndef _SYS_SYSPROTO_H_
1293struct freebsd7_shmctl_args {
1294	int shmid;
1295	int cmd;
1296	struct shmid_ds_old *buf;
1297};
1298#endif
1299int
1300freebsd7_shmctl(td, uap)
1301	struct thread *td;
1302	struct freebsd7_shmctl_args *uap;
1303{
1304	int error = 0;
1305	struct shmid_ds_old old;
1306	struct shmid_ds buf;
1307	size_t bufsz;
1308
1309	/*
1310	 * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support
1311	 * Linux binaries.  If we see the call come through the FreeBSD ABI,
1312	 * return an error back to the user since we do not to support this.
1313	 */
1314	if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO ||
1315	    uap->cmd == SHM_STAT)
1316		return (EINVAL);
1317
1318	/* IPC_SET needs to copyin the buffer before calling kern_shmctl */
1319	if (uap->cmd == IPC_SET) {
1320		if ((error = copyin(uap->buf, &old, sizeof(old))))
1321			goto done;
1322		ipcperm_old2new(&old.shm_perm, &buf.shm_perm);
1323		CP(old, buf, shm_segsz);
1324		CP(old, buf, shm_lpid);
1325		CP(old, buf, shm_cpid);
1326		CP(old, buf, shm_nattch);
1327		CP(old, buf, shm_atime);
1328		CP(old, buf, shm_dtime);
1329		CP(old, buf, shm_ctime);
1330	}
1331
1332	error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz);
1333	if (error)
1334		goto done;
1335
1336	/* Cases in which we need to copyout */
1337	switch (uap->cmd) {
1338	case IPC_STAT:
1339		ipcperm_new2old(&buf.shm_perm, &old.shm_perm);
1340		if (buf.shm_segsz > INT_MAX)
1341			old.shm_segsz = INT_MAX;
1342		else
1343			CP(buf, old, shm_segsz);
1344		CP(buf, old, shm_lpid);
1345		CP(buf, old, shm_cpid);
1346		if (buf.shm_nattch > SHRT_MAX)
1347			old.shm_nattch = SHRT_MAX;
1348		else
1349			CP(buf, old, shm_nattch);
1350		CP(buf, old, shm_atime);
1351		CP(buf, old, shm_dtime);
1352		CP(buf, old, shm_ctime);
1353		old.shm_internal = NULL;
1354		error = copyout(&old, uap->buf, sizeof(old));
1355		break;
1356	}
1357
1358done:
1359	if (error) {
1360		/* Invalidate the return value */
1361		td->td_retval[0] = -1;
1362	}
1363	return (error);
1364}
1365
1366#endif	/* COMPAT_FREEBSD4 || COMPAT_FREEBSD5 || COMPAT_FREEBSD6 ||
1367	   COMPAT_FREEBSD7 */
1368
1369static int
1370sysvshm_modload(struct module *module, int cmd, void *arg)
1371{
1372	int error = 0;
1373
1374	switch (cmd) {
1375	case MOD_LOAD:
1376		error = shminit();
1377		if (error != 0)
1378			shmunload();
1379		break;
1380	case MOD_UNLOAD:
1381		error = shmunload();
1382		break;
1383	case MOD_SHUTDOWN:
1384		break;
1385	default:
1386		error = EINVAL;
1387		break;
1388	}
1389	return (error);
1390}
1391
1392static moduledata_t sysvshm_mod = {
1393	"sysvshm",
1394	&sysvshm_modload,
1395	NULL
1396};
1397
1398DECLARE_MODULE(sysvshm, sysvshm_mod, SI_SUB_SYSV_SHM, SI_ORDER_FIRST);
1399MODULE_VERSION(sysvshm, 1);
1400