sysv_shm.c revision 82607 
1/* $FreeBSD: head/sys/kern/sysv_shm.c 82607 2001-08-31 00:02:18Z dillon $ */
2/*	$NetBSD: sysv_shm.c,v 1.23 1994/07/04 23:25:12 glass Exp $	*/
3
4/*
5 * Copyright (c) 1994 Adam Glass and Charles Hannum.  All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 * 3. All advertising materials mentioning features or use of this software
16 *    must display the following acknowledgement:
17 *	This product includes software developed by Adam Glass and Charles
18 *	Hannum.
19 * 4. The names of the authors may not be used to endorse or promote products
20 *    derived from this software without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR
23 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
24 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
25 * IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT, INDIRECT,
26 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
27 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
31 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32 */
33
34#include "opt_compat.h"
35#include "opt_rlimit.h"
36#include "opt_sysvipc.h"
37
38#include <sys/param.h>
39#include <sys/systm.h>
40#include <sys/kernel.h>
41#include <sys/lock.h>
42#include <sys/sysctl.h>
43#include <sys/shm.h>
44#include <sys/proc.h>
45#include <sys/malloc.h>
46#include <sys/mman.h>
47#include <sys/mutex.h>
48#include <sys/stat.h>
49#include <sys/syscall.h>
50#include <sys/sysent.h>
51#include <sys/sysproto.h>
52#include <sys/jail.h>
53
54#include <vm/vm.h>
55#include <vm/vm_param.h>
56#include <vm/pmap.h>
57#include <vm/vm_object.h>
58#include <vm/vm_map.h>
59#include <vm/vm_page.h>
60#include <vm/vm_pager.h>
61
62static MALLOC_DEFINE(M_SHM, "shm", "SVID compatible shared memory segments");
63
64struct oshmctl_args;
65static int oshmctl __P((struct proc *p, struct oshmctl_args *uap));
66
67static int shmget_allocate_segment __P((struct proc *p,
68    struct shmget_args *uap, int mode));
69static int shmget_existing __P((struct proc *p, struct shmget_args *uap,
70    int mode, int segnum));
71
72/* XXX casting to (sy_call_t *) is bogus, as usual. */
73static sy_call_t *shmcalls[] = {
74	(sy_call_t *)shmat, (sy_call_t *)oshmctl,
75	(sy_call_t *)shmdt, (sy_call_t *)shmget,
76	(sy_call_t *)shmctl
77};
78
79#define	SHMSEG_FREE     	0x0200
80#define	SHMSEG_REMOVED  	0x0400
81#define	SHMSEG_ALLOCATED	0x0800
82#define	SHMSEG_WANTED		0x1000
83
84static int shm_last_free, shm_nused, shm_committed, shmalloced;
85static struct shmid_ds	*shmsegs;
86
87struct shm_handle {
88	/* vm_offset_t kva; */
89	vm_object_t shm_object;
90};
91
92struct shmmap_state {
93	vm_offset_t va;
94	int shmid;
95};
96
97static void shm_deallocate_segment __P((struct shmid_ds *));
98static int shm_find_segment_by_key __P((key_t));
99static struct shmid_ds *shm_find_segment_by_shmid __P((int));
100static int shm_delete_mapping __P((struct proc *, struct shmmap_state *));
101static void shmrealloc __P((void));
102static void shminit __P((void));
103static int sysvshm_modload __P((struct module *, int, void *));
104static int shmunload __P((void));
105static void shmexit_myhook __P((struct proc *p));
106static void shmfork_myhook __P((struct proc *p1, struct proc *p2));
107static int sysctl_shmsegs __P((SYSCTL_HANDLER_ARGS));
108
109/*
110 * Tuneable values.
111 */
112#ifndef SHMMAXPGS
113#define	SHMMAXPGS	8192	/* Note: sysv shared memory is swap backed. */
114#endif
115#ifndef SHMMAX
116#define	SHMMAX	(SHMMAXPGS*PAGE_SIZE)
117#endif
118#ifndef SHMMIN
119#define	SHMMIN	1
120#endif
121#ifndef SHMMNI
122#define	SHMMNI	192
123#endif
124#ifndef SHMSEG
125#define	SHMSEG	128
126#endif
127#ifndef SHMALL
128#define	SHMALL	(SHMMAXPGS)
129#endif
130
131struct	shminfo shminfo = {
132	SHMMAX,
133	SHMMIN,
134	SHMMNI,
135	SHMSEG,
136	SHMALL
137};
138
139static int shm_use_phys;
140
141SYSCTL_DECL(_kern_ipc);
142SYSCTL_INT(_kern_ipc, OID_AUTO, shmmax, CTLFLAG_RW, &shminfo.shmmax, 0, "");
143SYSCTL_INT(_kern_ipc, OID_AUTO, shmmin, CTLFLAG_RW, &shminfo.shmmin, 0, "");
144SYSCTL_INT(_kern_ipc, OID_AUTO, shmmni, CTLFLAG_RD, &shminfo.shmmni, 0, "");
145SYSCTL_INT(_kern_ipc, OID_AUTO, shmseg, CTLFLAG_RD, &shminfo.shmseg, 0, "");
146SYSCTL_INT(_kern_ipc, OID_AUTO, shmall, CTLFLAG_RW, &shminfo.shmall, 0, "");
147SYSCTL_INT(_kern_ipc, OID_AUTO, shm_use_phys, CTLFLAG_RW,
148    &shm_use_phys, 0, "");
149SYSCTL_PROC(_kern_ipc, OID_AUTO, shmsegs, CTLFLAG_RD,
150    NULL, 0, sysctl_shmsegs, "", "");
151
152static int
153shm_find_segment_by_key(key)
154	key_t key;
155{
156	int i;
157
158	for (i = 0; i < shmalloced; i++)
159		if ((shmsegs[i].shm_perm.mode & SHMSEG_ALLOCATED) &&
160		    shmsegs[i].shm_perm.key == key)
161			return i;
162	return -1;
163}
164
165static struct shmid_ds *
166shm_find_segment_by_shmid(shmid)
167	int shmid;
168{
169	int segnum;
170	struct shmid_ds *shmseg;
171
172	segnum = IPCID_TO_IX(shmid);
173	if (segnum < 0 || segnum >= shmalloced)
174		return NULL;
175	shmseg = &shmsegs[segnum];
176	if ((shmseg->shm_perm.mode & (SHMSEG_ALLOCATED | SHMSEG_REMOVED))
177	    != SHMSEG_ALLOCATED ||
178	    shmseg->shm_perm.seq != IPCID_TO_SEQ(shmid))
179		return NULL;
180	return shmseg;
181}
182
183static void
184shm_deallocate_segment(shmseg)
185	struct shmid_ds *shmseg;
186{
187	struct shm_handle *shm_handle;
188	size_t size;
189
190	GIANT_REQUIRED;
191
192	shm_handle = shmseg->shm_internal;
193	vm_object_deallocate(shm_handle->shm_object);
194	free((caddr_t)shm_handle, M_SHM);
195	shmseg->shm_internal = NULL;
196	size = round_page(shmseg->shm_segsz);
197	shm_committed -= btoc(size);
198	shm_nused--;
199	shmseg->shm_perm.mode = SHMSEG_FREE;
200}
201
202static int
203shm_delete_mapping(p, shmmap_s)
204	struct proc *p;
205	struct shmmap_state *shmmap_s;
206{
207	struct shmid_ds *shmseg;
208	int segnum, result;
209	size_t size;
210
211	GIANT_REQUIRED;
212
213	segnum = IPCID_TO_IX(shmmap_s->shmid);
214	shmseg = &shmsegs[segnum];
215	size = round_page(shmseg->shm_segsz);
216	result = vm_map_remove(&p->p_vmspace->vm_map, shmmap_s->va,
217	    shmmap_s->va + size);
218	if (result != KERN_SUCCESS)
219		return EINVAL;
220	shmmap_s->shmid = -1;
221	shmseg->shm_dtime = time_second;
222	if ((--shmseg->shm_nattch <= 0) &&
223	    (shmseg->shm_perm.mode & SHMSEG_REMOVED)) {
224		shm_deallocate_segment(shmseg);
225		shm_last_free = segnum;
226	}
227	return 0;
228}
229
230#ifndef _SYS_SYSPROTO_H_
231struct shmdt_args {
232	void *shmaddr;
233};
234#endif
235
236/*
237 * MPSAFE
238 */
239int
240shmdt(p, uap)
241	struct proc *p;
242	struct shmdt_args *uap;
243{
244	struct shmmap_state *shmmap_s;
245	int i;
246	int error = 0;
247
248	mtx_lock(&Giant);
249
250	if (!jail_sysvipc_allowed && jailed(p->p_ucred)) {
251		error = ENOSYS;
252		goto done2;
253	}
254
255	shmmap_s = (struct shmmap_state *)p->p_vmspace->vm_shm;
256 	if (shmmap_s == NULL) {
257		error = EINVAL;
258		goto done2;
259	}
260	for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) {
261		if (shmmap_s->shmid != -1 &&
262		    shmmap_s->va == (vm_offset_t)uap->shmaddr) {
263			break;
264		}
265	}
266	if (i == shminfo.shmseg) {
267		error = EINVAL;
268		goto done2;
269	}
270	error = shm_delete_mapping(p, shmmap_s);
271done2:
272	mtx_unlock(&Giant);
273	return (error);
274}
275
276#ifndef _SYS_SYSPROTO_H_
277struct shmat_args {
278	int shmid;
279	void *shmaddr;
280	int shmflg;
281};
282#endif
283
284/*
285 * MPSAFE
286 */
287int
288shmat(p, uap)
289	struct proc *p;
290	struct shmat_args *uap;
291{
292	int i, flags;
293	struct shmid_ds *shmseg;
294	struct shmmap_state *shmmap_s = NULL;
295	struct shm_handle *shm_handle;
296	vm_offset_t attach_va;
297	vm_prot_t prot;
298	vm_size_t size;
299	int rv;
300	int error = 0;
301
302	mtx_lock(&Giant);
303
304	if (!jail_sysvipc_allowed && jailed(p->p_ucred)) {
305		error = ENOSYS;
306		goto done2;
307	}
308
309	shmmap_s = (struct shmmap_state *)p->p_vmspace->vm_shm;
310	if (shmmap_s == NULL) {
311		size = shminfo.shmseg * sizeof(struct shmmap_state);
312		shmmap_s = malloc(size, M_SHM, M_WAITOK);
313		for (i = 0; i < shminfo.shmseg; i++)
314			shmmap_s[i].shmid = -1;
315		p->p_vmspace->vm_shm = (caddr_t)shmmap_s;
316	}
317	shmseg = shm_find_segment_by_shmid(uap->shmid);
318	if (shmseg == NULL) {
319		error = EINVAL;
320		goto done2;
321	}
322	error = ipcperm(p, &shmseg->shm_perm,
323	    (uap->shmflg & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W);
324	if (error)
325		goto done2;
326	for (i = 0; i < shminfo.shmseg; i++) {
327		if (shmmap_s->shmid == -1)
328			break;
329		shmmap_s++;
330	}
331	if (i >= shminfo.shmseg) {
332		error = EMFILE;
333		goto done2;
334	}
335	size = round_page(shmseg->shm_segsz);
336#ifdef VM_PROT_READ_IS_EXEC
337	prot = VM_PROT_READ | VM_PROT_EXECUTE;
338#else
339	prot = VM_PROT_READ;
340#endif
341	if ((uap->shmflg & SHM_RDONLY) == 0)
342		prot |= VM_PROT_WRITE;
343	flags = MAP_ANON | MAP_SHARED;
344	if (uap->shmaddr) {
345		flags |= MAP_FIXED;
346		if (uap->shmflg & SHM_RND) {
347			attach_va = (vm_offset_t)uap->shmaddr & ~(SHMLBA-1);
348		} else if (((vm_offset_t)uap->shmaddr & (SHMLBA-1)) == 0) {
349			attach_va = (vm_offset_t)uap->shmaddr;
350		} else {
351			error = EINVAL;
352			goto done2;
353		}
354	} else {
355		/*
356		 * This is just a hint to vm_map_find() about where to
357		 * put it.
358		 */
359		attach_va = round_page((vm_offset_t)p->p_vmspace->vm_taddr
360		    + MAXTSIZ + MAXDSIZ);
361	}
362
363	shm_handle = shmseg->shm_internal;
364	vm_object_reference(shm_handle->shm_object);
365	rv = vm_map_find(&p->p_vmspace->vm_map, shm_handle->shm_object,
366		0, &attach_va, size, (flags & MAP_FIXED)?0:1, prot, prot, 0);
367	if (rv != KERN_SUCCESS) {
368		error = ENOMEM;
369		goto done2;
370	}
371	vm_map_inherit(&p->p_vmspace->vm_map,
372		attach_va, attach_va + size, VM_INHERIT_SHARE);
373
374	shmmap_s->va = attach_va;
375	shmmap_s->shmid = uap->shmid;
376	shmseg->shm_lpid = p->p_pid;
377	shmseg->shm_atime = time_second;
378	shmseg->shm_nattch++;
379	p->p_retval[0] = attach_va;
380done2:
381	mtx_unlock(&Giant);
382	return (error);
383}
384
385struct oshmid_ds {
386	struct	ipc_perm shm_perm;	/* operation perms */
387	int	shm_segsz;		/* size of segment (bytes) */
388	ushort	shm_cpid;		/* pid, creator */
389	ushort	shm_lpid;		/* pid, last operation */
390	short	shm_nattch;		/* no. of current attaches */
391	time_t	shm_atime;		/* last attach time */
392	time_t	shm_dtime;		/* last detach time */
393	time_t	shm_ctime;		/* last change time */
394	void	*shm_handle;		/* internal handle for shm segment */
395};
396
397struct oshmctl_args {
398	int shmid;
399	int cmd;
400	struct oshmid_ds *ubuf;
401};
402
403/*
404 * MPSAFE
405 */
406static int
407oshmctl(p, uap)
408	struct proc *p;
409	struct oshmctl_args *uap;
410{
411#ifdef COMPAT_43
412	int error = 0;
413	struct shmid_ds *shmseg;
414	struct oshmid_ds outbuf;
415
416	mtx_lock(&Giant);
417
418	if (!jail_sysvipc_allowed && jailed(p->p_ucred)) {
419		error = ENOSYS;
420		goto done2;
421	}
422
423	shmseg = shm_find_segment_by_shmid(uap->shmid);
424	if (shmseg == NULL) {
425		error = EINVAL;
426		goto done2;
427	}
428	switch (uap->cmd) {
429	case IPC_STAT:
430		error = ipcperm(p, &shmseg->shm_perm, IPC_R);
431		if (error)
432			goto done2;
433		outbuf.shm_perm = shmseg->shm_perm;
434		outbuf.shm_segsz = shmseg->shm_segsz;
435		outbuf.shm_cpid = shmseg->shm_cpid;
436		outbuf.shm_lpid = shmseg->shm_lpid;
437		outbuf.shm_nattch = shmseg->shm_nattch;
438		outbuf.shm_atime = shmseg->shm_atime;
439		outbuf.shm_dtime = shmseg->shm_dtime;
440		outbuf.shm_ctime = shmseg->shm_ctime;
441		outbuf.shm_handle = shmseg->shm_internal;
442		error = copyout((caddr_t)&outbuf, uap->ubuf, sizeof(outbuf));
443		if (error)
444			goto done2;
445		break;
446	default:
447		/* XXX casting to (sy_call_t *) is bogus, as usual. */
448		error = ((sy_call_t *)shmctl)(p, uap);
449		break;
450	}
451done2:
452	mtx_unlock(&Giant);
453	return (error);
454#else
455	return EINVAL;
456#endif
457}
458
459#ifndef _SYS_SYSPROTO_H_
460struct shmctl_args {
461	int shmid;
462	int cmd;
463	struct shmid_ds *buf;
464};
465#endif
466
467/*
468 * MPSAFE
469 */
470int
471shmctl(p, uap)
472	struct proc *p;
473	struct shmctl_args *uap;
474{
475	int error = 0;
476	struct shmid_ds inbuf;
477	struct shmid_ds *shmseg;
478
479	mtx_lock(&Giant);
480
481	if (!jail_sysvipc_allowed && jailed(p->p_ucred)) {
482		error = ENOSYS;
483		goto done2;
484	}
485
486	shmseg = shm_find_segment_by_shmid(uap->shmid);
487	if (shmseg == NULL) {
488		error = EINVAL;
489		goto done2;
490	}
491	switch (uap->cmd) {
492	case IPC_STAT:
493		error = ipcperm(p, &shmseg->shm_perm, IPC_R);
494		if (error)
495			goto done2;
496		error = copyout((caddr_t)shmseg, uap->buf, sizeof(inbuf));
497		if (error)
498			goto done2;
499		break;
500	case IPC_SET:
501		error = ipcperm(p, &shmseg->shm_perm, IPC_M);
502		if (error)
503			goto done2;
504		error = copyin(uap->buf, (caddr_t)&inbuf, sizeof(inbuf));
505		if (error)
506			goto done2;
507		shmseg->shm_perm.uid = inbuf.shm_perm.uid;
508		shmseg->shm_perm.gid = inbuf.shm_perm.gid;
509		shmseg->shm_perm.mode =
510		    (shmseg->shm_perm.mode & ~ACCESSPERMS) |
511		    (inbuf.shm_perm.mode & ACCESSPERMS);
512		shmseg->shm_ctime = time_second;
513		break;
514	case IPC_RMID:
515		error = ipcperm(p, &shmseg->shm_perm, IPC_M);
516		if (error)
517			goto done2;
518		shmseg->shm_perm.key = IPC_PRIVATE;
519		shmseg->shm_perm.mode |= SHMSEG_REMOVED;
520		if (shmseg->shm_nattch <= 0) {
521			shm_deallocate_segment(shmseg);
522			shm_last_free = IPCID_TO_IX(uap->shmid);
523		}
524		break;
525#if 0
526	case SHM_LOCK:
527	case SHM_UNLOCK:
528#endif
529	default:
530		error = EINVAL;
531		break;
532	}
533done2:
534	mtx_unlock(&Giant);
535	return (error);
536}
537
538#ifndef _SYS_SYSPROTO_H_
539struct shmget_args {
540	key_t key;
541	size_t size;
542	int shmflg;
543};
544#endif
545
546static int
547shmget_existing(p, uap, mode, segnum)
548	struct proc *p;
549	struct shmget_args *uap;
550	int mode;
551	int segnum;
552{
553	struct shmid_ds *shmseg;
554	int error;
555
556	shmseg = &shmsegs[segnum];
557	if (shmseg->shm_perm.mode & SHMSEG_REMOVED) {
558		/*
559		 * This segment is in the process of being allocated.  Wait
560		 * until it's done, and look the key up again (in case the
561		 * allocation failed or it was freed).
562		 */
563		shmseg->shm_perm.mode |= SHMSEG_WANTED;
564		error = tsleep((caddr_t)shmseg, PLOCK | PCATCH, "shmget", 0);
565		if (error)
566			return error;
567		return EAGAIN;
568	}
569	if ((uap->shmflg & (IPC_CREAT | IPC_EXCL)) == (IPC_CREAT | IPC_EXCL))
570		return EEXIST;
571	error = ipcperm(p, &shmseg->shm_perm, mode);
572	if (error)
573		return error;
574	if (uap->size && uap->size > shmseg->shm_segsz)
575		return EINVAL;
576	p->p_retval[0] = IXSEQ_TO_IPCID(segnum, shmseg->shm_perm);
577	return 0;
578}
579
580static int
581shmget_allocate_segment(p, uap, mode)
582	struct proc *p;
583	struct shmget_args *uap;
584	int mode;
585{
586	int i, segnum, shmid, size;
587	struct ucred *cred = p->p_ucred;
588	struct shmid_ds *shmseg;
589	struct shm_handle *shm_handle;
590
591	GIANT_REQUIRED;
592
593	if (uap->size < shminfo.shmmin || uap->size > shminfo.shmmax)
594		return EINVAL;
595	if (shm_nused >= shminfo.shmmni) /* Any shmids left? */
596		return ENOSPC;
597	size = round_page(uap->size);
598	if (shm_committed + btoc(size) > shminfo.shmall)
599		return ENOMEM;
600	if (shm_last_free < 0) {
601		shmrealloc();	/* Maybe expand the shmsegs[] array. */
602		for (i = 0; i < shmalloced; i++)
603			if (shmsegs[i].shm_perm.mode & SHMSEG_FREE)
604				break;
605		if (i == shmalloced)
606			return ENOSPC;
607		segnum = i;
608	} else  {
609		segnum = shm_last_free;
610		shm_last_free = -1;
611	}
612	shmseg = &shmsegs[segnum];
613	/*
614	 * In case we sleep in malloc(), mark the segment present but deleted
615	 * so that noone else tries to create the same key.
616	 */
617	shmseg->shm_perm.mode = SHMSEG_ALLOCATED | SHMSEG_REMOVED;
618	shmseg->shm_perm.key = uap->key;
619	shmseg->shm_perm.seq = (shmseg->shm_perm.seq + 1) & 0x7fff;
620	shm_handle = (struct shm_handle *)
621	    malloc(sizeof(struct shm_handle), M_SHM, M_WAITOK);
622	shmid = IXSEQ_TO_IPCID(segnum, shmseg->shm_perm);
623
624	/*
625	 * We make sure that we have allocated a pager before we need
626	 * to.
627	 */
628	if (shm_use_phys) {
629		shm_handle->shm_object =
630		    vm_pager_allocate(OBJT_PHYS, 0, size, VM_PROT_DEFAULT, 0);
631	} else {
632		shm_handle->shm_object =
633		    vm_pager_allocate(OBJT_SWAP, 0, size, VM_PROT_DEFAULT, 0);
634	}
635	vm_object_clear_flag(shm_handle->shm_object, OBJ_ONEMAPPING);
636	vm_object_set_flag(shm_handle->shm_object, OBJ_NOSPLIT);
637
638	shmseg->shm_internal = shm_handle;
639	shmseg->shm_perm.cuid = shmseg->shm_perm.uid = cred->cr_uid;
640	shmseg->shm_perm.cgid = shmseg->shm_perm.gid = cred->cr_gid;
641	shmseg->shm_perm.mode = (shmseg->shm_perm.mode & SHMSEG_WANTED) |
642	    (mode & ACCESSPERMS) | SHMSEG_ALLOCATED;
643	shmseg->shm_segsz = uap->size;
644	shmseg->shm_cpid = p->p_pid;
645	shmseg->shm_lpid = shmseg->shm_nattch = 0;
646	shmseg->shm_atime = shmseg->shm_dtime = 0;
647	shmseg->shm_ctime = time_second;
648	shm_committed += btoc(size);
649	shm_nused++;
650	if (shmseg->shm_perm.mode & SHMSEG_WANTED) {
651		/*
652		 * Somebody else wanted this key while we were asleep.  Wake
653		 * them up now.
654		 */
655		shmseg->shm_perm.mode &= ~SHMSEG_WANTED;
656		wakeup((caddr_t)shmseg);
657	}
658	p->p_retval[0] = shmid;
659	return 0;
660}
661
662/*
663 * MPSAFE
664 */
665int
666shmget(p, uap)
667	struct proc *p;
668	struct shmget_args *uap;
669{
670	int segnum, mode;
671	int error;
672
673	mtx_lock(&Giant);
674
675	if (!jail_sysvipc_allowed && jailed(p->p_ucred)) {
676		error = ENOSYS;
677		goto done2;
678	}
679
680	mode = uap->shmflg & ACCESSPERMS;
681	if (uap->key != IPC_PRIVATE) {
682	again:
683		segnum = shm_find_segment_by_key(uap->key);
684		if (segnum >= 0) {
685			error = shmget_existing(p, uap, mode, segnum);
686			if (error == EAGAIN)
687				goto again;
688			goto done2;
689		}
690		if ((uap->shmflg & IPC_CREAT) == 0) {
691			error = ENOENT;
692			goto done2;
693		}
694	}
695	error = shmget_allocate_segment(p, uap, mode);
696done2:
697	mtx_unlock(&Giant);
698	return (error);
699}
700
701/*
702 * MPSAFE
703 */
704int
705shmsys(p, uap)
706	struct proc *p;
707	/* XXX actually varargs. */
708	struct shmsys_args /* {
709		u_int	which;
710		int	a2;
711		int	a3;
712		int	a4;
713	} */ *uap;
714{
715	int error;
716
717	mtx_lock(&Giant);
718
719	if (!jail_sysvipc_allowed && jailed(p->p_ucred)) {
720		error = ENOSYS;
721		goto done2;
722	}
723
724	if (uap->which >= sizeof(shmcalls)/sizeof(shmcalls[0])) {
725		error = EINVAL;
726		goto done2;
727	}
728	error = (*shmcalls[uap->which])(p, &uap->a2);
729done2:
730	mtx_unlock(&Giant);
731	return (error);
732}
733
734static void
735shmfork_myhook(p1, p2)
736	struct proc *p1, *p2;
737{
738	struct shmmap_state *shmmap_s;
739	size_t size;
740	int i;
741
742	size = shminfo.shmseg * sizeof(struct shmmap_state);
743	shmmap_s = malloc(size, M_SHM, M_WAITOK);
744	bcopy((caddr_t)p1->p_vmspace->vm_shm, (caddr_t)shmmap_s, size);
745	p2->p_vmspace->vm_shm = (caddr_t)shmmap_s;
746	for (i = 0; i < shminfo.shmseg; i++, shmmap_s++)
747		if (shmmap_s->shmid != -1)
748			shmsegs[IPCID_TO_IX(shmmap_s->shmid)].shm_nattch++;
749}
750
751static void
752shmexit_myhook(p)
753	struct proc *p;
754{
755	struct shmmap_state *shmmap_s;
756	int i;
757
758	GIANT_REQUIRED;
759
760	shmmap_s = (struct shmmap_state *)p->p_vmspace->vm_shm;
761	for (i = 0; i < shminfo.shmseg; i++, shmmap_s++)
762		if (shmmap_s->shmid != -1)
763			shm_delete_mapping(p, shmmap_s);
764	free((caddr_t)p->p_vmspace->vm_shm, M_SHM);
765	p->p_vmspace->vm_shm = NULL;
766}
767
768static void
769shmrealloc(void)
770{
771	int i;
772	struct shmid_ds *newsegs;
773
774	if (shmalloced >= shminfo.shmmni)
775		return;
776
777	newsegs = malloc(shminfo.shmmni * sizeof(*newsegs), M_SHM, M_WAITOK);
778	if (newsegs == NULL)
779		return;
780	for (i = 0; i < shmalloced; i++)
781		bcopy(&shmsegs[i], &newsegs[i], sizeof(newsegs[0]));
782	for (; i < shminfo.shmmni; i++) {
783		shmsegs[i].shm_perm.mode = SHMSEG_FREE;
784		shmsegs[i].shm_perm.seq = 0;
785	}
786	free(shmsegs, M_SHM);
787	shmsegs = newsegs;
788	shmalloced = shminfo.shmmni;
789}
790
791static void
792shminit()
793{
794	int i;
795
796	shmalloced = shminfo.shmmni;
797	shmsegs = malloc(shmalloced * sizeof(shmsegs[0]), M_SHM, M_WAITOK);
798	if (shmsegs == NULL)
799		panic("cannot allocate initial memory for sysvshm");
800	for (i = 0; i < shmalloced; i++) {
801		shmsegs[i].shm_perm.mode = SHMSEG_FREE;
802		shmsegs[i].shm_perm.seq = 0;
803	}
804	shm_last_free = 0;
805	shm_nused = 0;
806	shm_committed = 0;
807	shmexit_hook = &shmexit_myhook;
808	shmfork_hook = &shmfork_myhook;
809}
810
811static int
812shmunload()
813{
814
815	if (shm_nused > 0)
816		return (EBUSY);
817
818	free(shmsegs, M_SHM);
819	shmexit_hook = NULL;
820	shmfork_hook = NULL;
821	return (0);
822}
823
824static int
825sysctl_shmsegs(SYSCTL_HANDLER_ARGS)
826{
827
828	return (SYSCTL_OUT(req, shmsegs, shmalloced * sizeof(shmsegs[0])));
829}
830
831static int
832sysvshm_modload(struct module *module, int cmd, void *arg)
833{
834	int error = 0;
835
836	switch (cmd) {
837	case MOD_LOAD:
838		shminit();
839		break;
840	case MOD_UNLOAD:
841		error = shmunload();
842		break;
843	case MOD_SHUTDOWN:
844		break;
845	default:
846		error = EINVAL;
847		break;
848	}
849	return (error);
850}
851
852static moduledata_t sysvshm_mod = {
853	"sysvshm",
854	&sysvshm_modload,
855	NULL
856};
857
858SYSCALL_MODULE_HELPER(shmsys, 4);
859SYSCALL_MODULE_HELPER(shmat, 3);
860SYSCALL_MODULE_HELPER(shmctl, 3);
861SYSCALL_MODULE_HELPER(shmdt, 1);
862SYSCALL_MODULE_HELPER(shmget, 3);
863
864DECLARE_MODULE(sysvshm, sysvshm_mod,
865	SI_SUB_SYSV_SHM, SI_ORDER_FIRST);
866MODULE_VERSION(sysvshm, 1);
867