1 /* $NetBSD: sysv_shm.c,v 1.23 1994/07/04 23:25:12 glass Exp $ */
3 * Copyright (c) 1994 Adam Glass and Charles Hannum. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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
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.
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.
32 * Copyright (c) 2003-2005 McAfee, Inc.
33 * All rights reserved.
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
40 * Redistribution and use in source and binary forms, with or without
41 * modification, are permitted provided that the following conditions
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.
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
62 #include <sys/cdefs.h>
63 __FBSDID("$FreeBSD$");
65 #include "opt_compat.h"
66 #include "opt_sysvipc.h"
68 #include <sys/param.h>
69 #include <sys/systm.h>
70 #include <sys/kernel.h>
71 #include <sys/limits.h>
73 #include <sys/sysctl.h>
76 #include <sys/malloc.h>
78 #include <sys/module.h>
79 #include <sys/mutex.h>
80 #include <sys/resourcevar.h>
82 #include <sys/syscall.h>
83 #include <sys/syscallsubr.h>
84 #include <sys/sysent.h>
85 #include <sys/sysproto.h>
88 #include <security/mac/mac_framework.h>
91 #include <vm/vm_param.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>
98 static MALLOC_DEFINE(M_SHM, "shm", "SVID compatible shared memory segments");
100 static int shmget_allocate_segment(struct thread *td,
101 struct shmget_args *uap, int mode);
102 static int shmget_existing(struct thread *td, struct shmget_args *uap,
103 int mode, int segnum);
105 #define SHMSEG_FREE 0x0200
106 #define SHMSEG_REMOVED 0x0400
107 #define SHMSEG_ALLOCATED 0x0800
108 #define SHMSEG_WANTED 0x1000
110 static int shm_last_free, shm_nused, shmalloced;
111 vm_size_t shm_committed;
112 static struct shmid_kernel *shmsegs;
114 struct shmmap_state {
119 static void shm_deallocate_segment(struct shmid_kernel *);
120 static int shm_find_segment_by_key(key_t);
121 static struct shmid_kernel *shm_find_segment_by_shmid(int);
122 static struct shmid_kernel *shm_find_segment_by_shmidx(int);
123 static int shm_delete_mapping(struct vmspace *vm, struct shmmap_state *);
124 static void shmrealloc(void);
125 static int shminit(void);
126 static int sysvshm_modload(struct module *, int, void *);
127 static int shmunload(void);
128 static void shmexit_myhook(struct vmspace *vm);
129 static void shmfork_myhook(struct proc *p1, struct proc *p2);
130 static int sysctl_shmsegs(SYSCTL_HANDLER_ARGS);
136 #define SHMMAXPGS 131072 /* Note: sysv shared memory is swap backed. */
139 #define SHMMAX (SHMMAXPGS*PAGE_SIZE)
151 #define SHMALL (SHMMAXPGS)
154 struct shminfo shminfo = {
162 static int shm_use_phys;
163 static int shm_allow_removed;
165 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmax, CTLFLAG_RW, &shminfo.shmmax, 0,
166 "Maximum shared memory segment size");
167 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmin, CTLFLAG_RW, &shminfo.shmmin, 0,
168 "Minimum shared memory segment size");
169 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmni, CTLFLAG_RDTUN, &shminfo.shmmni, 0,
170 "Number of shared memory identifiers");
171 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmseg, CTLFLAG_RDTUN, &shminfo.shmseg, 0,
172 "Number of segments per process");
173 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmall, CTLFLAG_RW, &shminfo.shmall, 0,
174 "Maximum number of pages available for shared memory");
175 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_use_phys, CTLFLAG_RW,
176 &shm_use_phys, 0, "Enable/Disable locking of shared memory pages in core");
177 SYSCTL_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");
180 SYSCTL_PROC(_kern_ipc, OID_AUTO, shmsegs, CTLFLAG_RD,
181 NULL, 0, sysctl_shmsegs, "",
182 "Current number of shared memory segments allocated");
185 shm_find_segment_by_key(key)
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)
197 static struct shmid_kernel *
198 shm_find_segment_by_shmid(int shmid)
201 struct shmid_kernel *shmseg;
203 segnum = IPCID_TO_IX(shmid);
204 if (segnum < 0 || segnum >= shmalloced)
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))
215 static struct shmid_kernel *
216 shm_find_segment_by_shmidx(int segnum)
218 struct shmid_kernel *shmseg;
220 if (segnum < 0 || segnum >= shmalloced)
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))
231 shm_deallocate_segment(shmseg)
232 struct shmid_kernel *shmseg;
238 vm_object_deallocate(shmseg->object);
239 shmseg->object = NULL;
240 size = round_page(shmseg->u.shm_segsz);
241 shm_committed -= btoc(size);
243 shmseg->u.shm_perm.mode = SHMSEG_FREE;
245 mac_sysvshm_cleanup(shmseg);
250 shm_delete_mapping(struct vmspace *vm, struct shmmap_state *shmmap_s)
252 struct shmid_kernel *shmseg;
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)
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;
274 #ifndef _SYS_SYSPROTO_H_
282 struct shmdt_args *uap;
284 struct proc *p = td->td_proc;
285 struct shmmap_state *shmmap_s;
287 struct shmid_kernel *shmsegptr;
292 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
295 shmmap_s = p->p_vmspace->vm_shm;
296 if (shmmap_s == NULL) {
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) {
306 if (i == shminfo.shmseg) {
311 shmsegptr = &shmsegs[IPCID_TO_IX(shmmap_s->shmid)];
312 error = mac_sysvshm_check_shmdt(td->td_ucred, shmsegptr);
316 error = shm_delete_mapping(p->p_vmspace, shmmap_s);
322 #ifndef _SYS_SYSPROTO_H_
330 kern_shmat(td, shmid, shmaddr, shmflg)
336 struct proc *p = td->td_proc;
338 struct shmid_kernel *shmseg;
339 struct shmmap_state *shmmap_s = NULL;
340 vm_offset_t attach_va;
346 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
349 shmmap_s = p->p_vmspace->vm_shm;
350 if (shmmap_s == NULL) {
351 shmmap_s = malloc(shminfo.shmseg * sizeof(struct shmmap_state),
353 for (i = 0; i < shminfo.shmseg; i++)
354 shmmap_s[i].shmid = -1;
355 p->p_vmspace->vm_shm = shmmap_s;
357 shmseg = shm_find_segment_by_shmid(shmid);
358 if (shmseg == NULL) {
362 error = ipcperm(td, &shmseg->u.shm_perm,
363 (shmflg & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W);
367 error = mac_sysvshm_check_shmat(td->td_ucred, shmseg, shmflg);
371 for (i = 0; i < shminfo.shmseg; i++) {
372 if (shmmap_s->shmid == -1)
376 if (i >= shminfo.shmseg) {
380 size = round_page(shmseg->u.shm_segsz);
382 if ((shmflg & SHM_RDONLY) == 0)
383 prot |= VM_PROT_WRITE;
384 flags = MAP_ANON | MAP_SHARED;
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;
397 * This is just a hint to vm_map_find() about where to
401 attach_va = round_page((vm_offset_t)p->p_vmspace->vm_daddr +
402 lim_max(p, RLIMIT_DATA));
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);
415 vm_map_inherit(&p->p_vmspace->vm_map,
416 attach_va, attach_va + size, VM_INHERIT_SHARE);
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;
432 struct shmat_args *uap;
434 return kern_shmat(td, uap->shmid, uap->shmaddr, uap->shmflg);
438 kern_shmctl(td, shmid, cmd, buf, bufsz)
446 struct shmid_kernel *shmseg;
448 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
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.
459 * NB: The Linux ABI layer will convert this data to structure(s) more
460 * consistent with the Linux ABI.
463 memcpy(buf, &shminfo, sizeof(shminfo));
465 *bufsz = sizeof(shminfo);
466 td->td_retval[0] = shmalloced;
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));
478 *bufsz = sizeof(shm_info);
479 td->td_retval[0] = shmalloced;
484 shmseg = shm_find_segment_by_shmidx(shmid);
486 shmseg = shm_find_segment_by_shmid(shmid);
487 if (shmseg == NULL) {
492 error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, cmd);
499 error = ipcperm(td, &shmseg->u.shm_perm, IPC_R);
502 memcpy(buf, &shmseg->u, sizeof(struct shmid_ds));
504 *bufsz = sizeof(struct shmid_ds);
506 td->td_retval[0] = IXSEQ_TO_IPCID(shmid, shmseg->u.shm_perm);
509 struct shmid_ds *shmid;
511 shmid = (struct shmid_ds *)buf;
512 error = ipcperm(td, &shmseg->u.shm_perm, IPC_M);
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;
524 error = ipcperm(td, &shmseg->u.shm_perm, IPC_M);
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);
547 #ifndef _SYS_SYSPROTO_H_
551 struct shmid_ds *buf;
557 struct shmctl_args *uap;
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.
568 if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO ||
569 uap->cmd == SHM_STAT)
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))))
578 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz);
582 /* Cases in which we need to copyout */
585 error = copyout(&buf, uap->buf, bufsz);
591 /* Invalidate the return value */
592 td->td_retval[0] = -1;
599 shmget_existing(td, uap, mode, segnum)
601 struct shmget_args *uap;
605 struct shmid_kernel *shmseg;
608 shmseg = &shmsegs[segnum];
609 if (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) {
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).
615 shmseg->u.shm_perm.mode |= SHMSEG_WANTED;
616 error = tsleep(shmseg, PLOCK | PCATCH, "shmget", 0);
621 if ((uap->shmflg & (IPC_CREAT | IPC_EXCL)) == (IPC_CREAT | IPC_EXCL))
624 error = mac_sysvshm_check_shmget(td->td_ucred, shmseg, uap->shmflg);
628 if (uap->size != 0 && uap->size > shmseg->u.shm_segsz)
630 td->td_retval[0] = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm);
635 shmget_allocate_segment(td, uap, mode)
637 struct shmget_args *uap;
640 int i, segnum, shmid;
642 struct ucred *cred = td->td_ucred;
643 struct shmid_kernel *shmseg;
644 vm_object_t shm_object;
648 if (uap->size < shminfo.shmmin || uap->size > shminfo.shmmax)
650 if (shm_nused >= shminfo.shmmni) /* Any shmids left? */
652 size = round_page(uap->size);
653 if (shm_committed + btoc(size) > shminfo.shmall)
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)
664 segnum = shm_last_free;
667 shmseg = &shmsegs[segnum];
669 * In case we sleep in malloc(), mark the segment present but deleted
670 * so that noone else tries to create the same key.
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);
678 * We make sure that we have allocated a pager before we need
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)
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);
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;
700 mac_sysvshm_create(cred, shmseg);
702 shmseg->u.shm_ctime = time_second;
703 shm_committed += btoc(size);
705 if (shmseg->u.shm_perm.mode & SHMSEG_WANTED) {
707 * Somebody else wanted this key while we were asleep. Wake
710 shmseg->u.shm_perm.mode &= ~SHMSEG_WANTED;
713 td->td_retval[0] = shmid;
717 #ifndef _SYS_SYSPROTO_H_
727 struct shmget_args *uap;
732 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
735 mode = uap->shmflg & ACCESSPERMS;
736 if (uap->key != IPC_PRIVATE) {
738 segnum = shm_find_segment_by_key(uap->key);
740 error = shmget_existing(td, uap, mode, segnum);
745 if ((uap->shmflg & IPC_CREAT) == 0) {
750 error = shmget_allocate_segment(td, uap, mode);
757 shmfork_myhook(p1, p2)
758 struct proc *p1, *p2;
760 struct shmmap_state *shmmap_s;
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++;
776 shmexit_myhook(struct vmspace *vm)
778 struct shmmap_state *base, *shm;
781 if ((base = vm->vm_shm) != NULL) {
784 for (i = 0, shm = base; i < shminfo.shmseg; i++, shm++) {
785 if (shm->shmid != -1)
786 shm_delete_mapping(vm, shm);
797 struct shmid_kernel *newsegs;
799 if (shmalloced >= shminfo.shmmni)
802 newsegs = malloc(shminfo.shmmni * sizeof(*newsegs), M_SHM, M_WAITOK);
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;
811 mac_sysvshm_init(&shmsegs[i]);
814 free(shmsegs, M_SHM);
816 shmalloced = shminfo.shmmni;
819 static 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),
828 #if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43))
829 SYSCALL_INIT_HELPER(shmsys),
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>
842 static 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),
862 if (TUNABLE_ULONG_FETCH("kern.ipc.shmmaxpgs", &shminfo.shmall) != 0)
863 printf("kern.ipc.shmmaxpgs is now called kern.ipc.shmall!\n");
865 TUNABLE_ULONG_FETCH("kern.ipc.shmall", &shminfo.shmall);
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)
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);
879 shmalloced = shminfo.shmmni;
880 shmsegs = malloc(shmalloced * sizeof(shmsegs[0]), M_SHM, M_WAITOK);
882 panic("cannot allocate initial memory for sysvshm");
883 for (i = 0; i < shmalloced; i++) {
884 shmsegs[i].u.shm_perm.mode = SHMSEG_FREE;
885 shmsegs[i].u.shm_perm.seq = 0;
887 mac_sysvshm_init(&shmsegs[i]);
893 shmexit_hook = &shmexit_myhook;
894 shmfork_hook = &shmfork_myhook;
896 error = syscall_helper_register(shm_syscalls);
899 #ifdef COMPAT_FREEBSD32
900 error = syscall32_helper_register(shm32_syscalls);
915 #ifdef COMPAT_FREEBSD32
916 syscall32_helper_unregister(shm32_syscalls);
918 syscall_helper_unregister(shm_syscalls);
920 for (i = 0; i < shmalloced; i++) {
922 mac_sysvshm_destroy(&shmsegs[i]);
925 * Objects might be still mapped into the processes
926 * address spaces. Actual free would happen on the
927 * last mapping destruction.
929 if (shmsegs[i].u.shm_perm.mode != SHMSEG_FREE)
930 vm_object_deallocate(shmsegs[i].object);
932 free(shmsegs, M_SHM);
939 sysctl_shmsegs(SYSCTL_HANDLER_ARGS)
942 return (SYSCTL_OUT(req, shmsegs, shmalloced * sizeof(shmsegs[0])));
945 #if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43))
947 struct ipc_perm_old shm_perm; /* operation perms */
948 int shm_segsz; /* size of segment (bytes) */
949 u_short shm_cpid; /* pid, creator */
950 u_short shm_lpid; /* pid, last operation */
951 short shm_nattch; /* no. of current attaches */
952 time_t shm_atime; /* last attach time */
953 time_t shm_dtime; /* last detach time */
954 time_t shm_ctime; /* last change time */
955 void *shm_handle; /* internal handle for shm segment */
958 struct oshmctl_args {
961 struct oshmid_ds *ubuf;
965 oshmctl(struct thread *td, struct oshmctl_args *uap)
969 struct shmid_kernel *shmseg;
970 struct oshmid_ds outbuf;
972 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
975 shmseg = shm_find_segment_by_shmid(uap->shmid);
976 if (shmseg == NULL) {
982 error = ipcperm(td, &shmseg->u.shm_perm, IPC_R);
986 error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, uap->cmd);
990 ipcperm_new2old(&shmseg->u.shm_perm, &outbuf.shm_perm);
991 outbuf.shm_segsz = shmseg->u.shm_segsz;
992 outbuf.shm_cpid = shmseg->u.shm_cpid;
993 outbuf.shm_lpid = shmseg->u.shm_lpid;
994 outbuf.shm_nattch = shmseg->u.shm_nattch;
995 outbuf.shm_atime = shmseg->u.shm_atime;
996 outbuf.shm_dtime = shmseg->u.shm_dtime;
997 outbuf.shm_ctime = shmseg->u.shm_ctime;
998 outbuf.shm_handle = shmseg->object;
999 error = copyout(&outbuf, uap->ubuf, sizeof(outbuf));
1004 error = freebsd7_shmctl(td, (struct freebsd7_shmctl_args *)uap);
1015 /* XXX casting to (sy_call_t *) is bogus, as usual. */
1016 static sy_call_t *shmcalls[] = {
1017 (sy_call_t *)shmat, (sy_call_t *)oshmctl,
1018 (sy_call_t *)shmdt, (sy_call_t *)shmget,
1019 (sy_call_t *)freebsd7_shmctl
1025 /* XXX actually varargs. */
1026 struct shmsys_args /* {
1035 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
1037 if (uap->which < 0 ||
1038 uap->which >= sizeof(shmcalls)/sizeof(shmcalls[0]))
1041 error = (*shmcalls[uap->which])(td, &uap->a2);
1046 #endif /* i386 && (COMPAT_FREEBSD4 || COMPAT_43) */
1048 #ifdef COMPAT_FREEBSD32
1051 freebsd32_shmsys(struct thread *td, struct freebsd32_shmsys_args *uap)
1054 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1055 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1056 switch (uap->which) {
1057 case 0: { /* shmat */
1058 struct shmat_args ap;
1061 ap.shmaddr = PTRIN(uap->a3);
1062 ap.shmflg = uap->a4;
1063 return (sysent[SYS_shmat].sy_call(td, &ap));
1065 case 2: { /* shmdt */
1066 struct shmdt_args ap;
1068 ap.shmaddr = PTRIN(uap->a2);
1069 return (sysent[SYS_shmdt].sy_call(td, &ap));
1071 case 3: { /* shmget */
1072 struct shmget_args ap;
1076 ap.shmflg = uap->a4;
1077 return (sysent[SYS_shmget].sy_call(td, &ap));
1079 case 4: { /* shmctl */
1080 struct freebsd7_freebsd32_shmctl_args ap;
1084 ap.buf = PTRIN(uap->a4);
1085 return (freebsd7_freebsd32_shmctl(td, &ap));
1087 case 1: /* oshmctl */
1092 return (nosys(td, NULL));
1096 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1097 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1099 freebsd7_freebsd32_shmctl(struct thread *td,
1100 struct freebsd7_freebsd32_shmctl_args *uap)
1104 struct shmid_ds shmid_ds;
1105 struct shm_info shm_info;
1106 struct shminfo shminfo;
1109 struct shmid_ds32_old shmid_ds32;
1110 struct shm_info32 shm_info32;
1111 struct shminfo32 shminfo32;
1115 if (uap->cmd == IPC_SET) {
1116 if ((error = copyin(uap->buf, &u32.shmid_ds32,
1117 sizeof(u32.shmid_ds32))))
1119 freebsd32_ipcperm_old_in(&u32.shmid_ds32.shm_perm,
1120 &u.shmid_ds.shm_perm);
1121 CP(u32.shmid_ds32, u.shmid_ds, shm_segsz);
1122 CP(u32.shmid_ds32, u.shmid_ds, shm_lpid);
1123 CP(u32.shmid_ds32, u.shmid_ds, shm_cpid);
1124 CP(u32.shmid_ds32, u.shmid_ds, shm_nattch);
1125 CP(u32.shmid_ds32, u.shmid_ds, shm_atime);
1126 CP(u32.shmid_ds32, u.shmid_ds, shm_dtime);
1127 CP(u32.shmid_ds32, u.shmid_ds, shm_ctime);
1130 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz);
1134 /* Cases in which we need to copyout */
1137 CP(u.shminfo, u32.shminfo32, shmmax);
1138 CP(u.shminfo, u32.shminfo32, shmmin);
1139 CP(u.shminfo, u32.shminfo32, shmmni);
1140 CP(u.shminfo, u32.shminfo32, shmseg);
1141 CP(u.shminfo, u32.shminfo32, shmall);
1142 error = copyout(&u32.shminfo32, uap->buf,
1143 sizeof(u32.shminfo32));
1146 CP(u.shm_info, u32.shm_info32, used_ids);
1147 CP(u.shm_info, u32.shm_info32, shm_rss);
1148 CP(u.shm_info, u32.shm_info32, shm_tot);
1149 CP(u.shm_info, u32.shm_info32, shm_swp);
1150 CP(u.shm_info, u32.shm_info32, swap_attempts);
1151 CP(u.shm_info, u32.shm_info32, swap_successes);
1152 error = copyout(&u32.shm_info32, uap->buf,
1153 sizeof(u32.shm_info32));
1157 freebsd32_ipcperm_old_out(&u.shmid_ds.shm_perm,
1158 &u32.shmid_ds32.shm_perm);
1159 if (u.shmid_ds.shm_segsz > INT32_MAX)
1160 u32.shmid_ds32.shm_segsz = INT32_MAX;
1162 CP(u.shmid_ds, u32.shmid_ds32, shm_segsz);
1163 CP(u.shmid_ds, u32.shmid_ds32, shm_lpid);
1164 CP(u.shmid_ds, u32.shmid_ds32, shm_cpid);
1165 CP(u.shmid_ds, u32.shmid_ds32, shm_nattch);
1166 CP(u.shmid_ds, u32.shmid_ds32, shm_atime);
1167 CP(u.shmid_ds, u32.shmid_ds32, shm_dtime);
1168 CP(u.shmid_ds, u32.shmid_ds32, shm_ctime);
1169 u32.shmid_ds32.shm_internal = 0;
1170 error = copyout(&u32.shmid_ds32, uap->buf,
1171 sizeof(u32.shmid_ds32));
1177 /* Invalidate the return value */
1178 td->td_retval[0] = -1;
1185 freebsd32_shmctl(struct thread *td, struct freebsd32_shmctl_args *uap)
1189 struct shmid_ds shmid_ds;
1190 struct shm_info shm_info;
1191 struct shminfo shminfo;
1194 struct shmid_ds32 shmid_ds32;
1195 struct shm_info32 shm_info32;
1196 struct shminfo32 shminfo32;
1200 if (uap->cmd == IPC_SET) {
1201 if ((error = copyin(uap->buf, &u32.shmid_ds32,
1202 sizeof(u32.shmid_ds32))))
1204 freebsd32_ipcperm_in(&u32.shmid_ds32.shm_perm,
1205 &u.shmid_ds.shm_perm);
1206 CP(u32.shmid_ds32, u.shmid_ds, shm_segsz);
1207 CP(u32.shmid_ds32, u.shmid_ds, shm_lpid);
1208 CP(u32.shmid_ds32, u.shmid_ds, shm_cpid);
1209 CP(u32.shmid_ds32, u.shmid_ds, shm_nattch);
1210 CP(u32.shmid_ds32, u.shmid_ds, shm_atime);
1211 CP(u32.shmid_ds32, u.shmid_ds, shm_dtime);
1212 CP(u32.shmid_ds32, u.shmid_ds, shm_ctime);
1215 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz);
1219 /* Cases in which we need to copyout */
1222 CP(u.shminfo, u32.shminfo32, shmmax);
1223 CP(u.shminfo, u32.shminfo32, shmmin);
1224 CP(u.shminfo, u32.shminfo32, shmmni);
1225 CP(u.shminfo, u32.shminfo32, shmseg);
1226 CP(u.shminfo, u32.shminfo32, shmall);
1227 error = copyout(&u32.shminfo32, uap->buf,
1228 sizeof(u32.shminfo32));
1231 CP(u.shm_info, u32.shm_info32, used_ids);
1232 CP(u.shm_info, u32.shm_info32, shm_rss);
1233 CP(u.shm_info, u32.shm_info32, shm_tot);
1234 CP(u.shm_info, u32.shm_info32, shm_swp);
1235 CP(u.shm_info, u32.shm_info32, swap_attempts);
1236 CP(u.shm_info, u32.shm_info32, swap_successes);
1237 error = copyout(&u32.shm_info32, uap->buf,
1238 sizeof(u32.shm_info32));
1242 freebsd32_ipcperm_out(&u.shmid_ds.shm_perm,
1243 &u32.shmid_ds32.shm_perm);
1244 if (u.shmid_ds.shm_segsz > INT32_MAX)
1245 u32.shmid_ds32.shm_segsz = INT32_MAX;
1247 CP(u.shmid_ds, u32.shmid_ds32, shm_segsz);
1248 CP(u.shmid_ds, u32.shmid_ds32, shm_lpid);
1249 CP(u.shmid_ds, u32.shmid_ds32, shm_cpid);
1250 CP(u.shmid_ds, u32.shmid_ds32, shm_nattch);
1251 CP(u.shmid_ds, u32.shmid_ds32, shm_atime);
1252 CP(u.shmid_ds, u32.shmid_ds32, shm_dtime);
1253 CP(u.shmid_ds, u32.shmid_ds32, shm_ctime);
1254 error = copyout(&u32.shmid_ds32, uap->buf,
1255 sizeof(u32.shmid_ds32));
1261 /* Invalidate the return value */
1262 td->td_retval[0] = -1;
1268 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1269 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1272 #define CP(src, dst, fld) do { (dst).fld = (src).fld; } while (0)
1275 #ifndef _SYS_SYSPROTO_H_
1276 struct freebsd7_shmctl_args {
1279 struct shmid_ds_old *buf;
1283 freebsd7_shmctl(td, uap)
1285 struct freebsd7_shmctl_args *uap;
1288 struct shmid_ds_old old;
1289 struct shmid_ds buf;
1293 * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support
1294 * Linux binaries. If we see the call come through the FreeBSD ABI,
1295 * return an error back to the user since we do not to support this.
1297 if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO ||
1298 uap->cmd == SHM_STAT)
1301 /* IPC_SET needs to copyin the buffer before calling kern_shmctl */
1302 if (uap->cmd == IPC_SET) {
1303 if ((error = copyin(uap->buf, &old, sizeof(old))))
1305 ipcperm_old2new(&old.shm_perm, &buf.shm_perm);
1306 CP(old, buf, shm_segsz);
1307 CP(old, buf, shm_lpid);
1308 CP(old, buf, shm_cpid);
1309 CP(old, buf, shm_nattch);
1310 CP(old, buf, shm_atime);
1311 CP(old, buf, shm_dtime);
1312 CP(old, buf, shm_ctime);
1315 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz);
1319 /* Cases in which we need to copyout */
1322 ipcperm_new2old(&buf.shm_perm, &old.shm_perm);
1323 if (buf.shm_segsz > INT_MAX)
1324 old.shm_segsz = INT_MAX;
1326 CP(buf, old, shm_segsz);
1327 CP(buf, old, shm_lpid);
1328 CP(buf, old, shm_cpid);
1329 if (buf.shm_nattch > SHRT_MAX)
1330 old.shm_nattch = SHRT_MAX;
1332 CP(buf, old, shm_nattch);
1333 CP(buf, old, shm_atime);
1334 CP(buf, old, shm_dtime);
1335 CP(buf, old, shm_ctime);
1336 old.shm_internal = NULL;
1337 error = copyout(&old, uap->buf, sizeof(old));
1343 /* Invalidate the return value */
1344 td->td_retval[0] = -1;
1349 #endif /* COMPAT_FREEBSD4 || COMPAT_FREEBSD5 || COMPAT_FREEBSD6 ||
1353 sysvshm_modload(struct module *module, int cmd, void *arg)
1364 error = shmunload();
1375 static moduledata_t sysvshm_mod = {
1381 DECLARE_MODULE(sysvshm, sysvshm_mod, SI_SUB_SYSV_SHM, SI_ORDER_FIRST);
1382 MODULE_VERSION(sysvshm, 1);