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/racct.h>
81 #include <sys/resourcevar.h>
82 #include <sys/rwlock.h>
84 #include <sys/syscall.h>
85 #include <sys/syscallsubr.h>
86 #include <sys/sysent.h>
87 #include <sys/sysproto.h>
90 #include <security/mac/mac_framework.h>
93 #include <vm/vm_param.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>
100 FEATURE(sysv_shm, "System V shared memory segments support");
102 static MALLOC_DEFINE(M_SHM, "shm", "SVID compatible shared memory segments");
104 static int shmget_allocate_segment(struct thread *td,
105 struct shmget_args *uap, int mode);
106 static int shmget_existing(struct thread *td, struct shmget_args *uap,
107 int mode, int segnum);
109 #define SHMSEG_FREE 0x0200
110 #define SHMSEG_REMOVED 0x0400
111 #define SHMSEG_ALLOCATED 0x0800
112 #define SHMSEG_WANTED 0x1000
114 static int shm_last_free, shm_nused, shmalloced;
115 vm_size_t shm_committed;
116 static struct shmid_kernel *shmsegs;
118 struct shmmap_state {
123 static void shm_deallocate_segment(struct shmid_kernel *);
124 static int shm_find_segment_by_key(key_t);
125 static struct shmid_kernel *shm_find_segment_by_shmid(int);
126 static struct shmid_kernel *shm_find_segment_by_shmidx(int);
127 static int shm_delete_mapping(struct vmspace *vm, struct shmmap_state *);
128 static void shmrealloc(void);
129 static int shminit(void);
130 static int sysvshm_modload(struct module *, int, void *);
131 static int shmunload(void);
132 static void shmexit_myhook(struct vmspace *vm);
133 static void shmfork_myhook(struct proc *p1, struct proc *p2);
134 static int sysctl_shmsegs(SYSCTL_HANDLER_ARGS);
140 #define SHMMAXPGS 131072 /* Note: sysv shared memory is swap backed. */
143 #define SHMMAX (SHMMAXPGS*PAGE_SIZE)
155 #define SHMALL (SHMMAXPGS)
158 struct shminfo shminfo = {
166 static int shm_use_phys;
167 static int shm_allow_removed;
169 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmax, CTLFLAG_RW, &shminfo.shmmax, 0,
170 "Maximum shared memory segment size");
171 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmin, CTLFLAG_RW, &shminfo.shmmin, 0,
172 "Minimum shared memory segment size");
173 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmni, CTLFLAG_RDTUN, &shminfo.shmmni, 0,
174 "Number of shared memory identifiers");
175 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmseg, CTLFLAG_RDTUN, &shminfo.shmseg, 0,
176 "Number of segments per process");
177 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmall, CTLFLAG_RW, &shminfo.shmall, 0,
178 "Maximum number of pages available for shared memory");
179 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_use_phys, CTLFLAG_RW,
180 &shm_use_phys, 0, "Enable/Disable locking of shared memory pages in core");
181 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_allow_removed, CTLFLAG_RW,
182 &shm_allow_removed, 0,
183 "Enable/Disable attachment to attached segments marked for removal");
184 SYSCTL_PROC(_kern_ipc, OID_AUTO, shmsegs, CTLTYPE_OPAQUE | CTLFLAG_RD,
185 NULL, 0, sysctl_shmsegs, "",
186 "Current number of shared memory segments allocated");
189 shm_find_segment_by_key(key)
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)
201 static struct shmid_kernel *
202 shm_find_segment_by_shmid(int shmid)
205 struct shmid_kernel *shmseg;
207 segnum = IPCID_TO_IX(shmid);
208 if (segnum < 0 || segnum >= shmalloced)
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))
219 static struct shmid_kernel *
220 shm_find_segment_by_shmidx(int segnum)
222 struct shmid_kernel *shmseg;
224 if (segnum < 0 || segnum >= shmalloced)
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))
235 shm_deallocate_segment(shmseg)
236 struct shmid_kernel *shmseg;
242 vm_object_deallocate(shmseg->object);
243 shmseg->object = NULL;
244 size = round_page(shmseg->u.shm_segsz);
245 shm_committed -= btoc(size);
247 shmseg->u.shm_perm.mode = SHMSEG_FREE;
249 mac_sysvshm_cleanup(shmseg);
251 racct_sub_cred(shmseg->cred, RACCT_NSHM, 1);
252 racct_sub_cred(shmseg->cred, RACCT_SHMSIZE, size);
253 crfree(shmseg->cred);
258 shm_delete_mapping(struct vmspace *vm, struct shmmap_state *shmmap_s)
260 struct shmid_kernel *shmseg;
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)
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;
282 #ifndef _SYS_SYSPROTO_H_
290 struct shmdt_args *uap;
292 struct proc *p = td->td_proc;
293 struct shmmap_state *shmmap_s;
295 struct shmid_kernel *shmsegptr;
300 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
303 shmmap_s = p->p_vmspace->vm_shm;
304 if (shmmap_s == NULL) {
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) {
314 if (i == shminfo.shmseg) {
319 shmsegptr = &shmsegs[IPCID_TO_IX(shmmap_s->shmid)];
320 error = mac_sysvshm_check_shmdt(td->td_ucred, shmsegptr);
324 error = shm_delete_mapping(p->p_vmspace, shmmap_s);
330 #ifndef _SYS_SYSPROTO_H_
338 kern_shmat(td, shmid, shmaddr, shmflg)
344 struct proc *p = td->td_proc;
346 struct shmid_kernel *shmseg;
347 struct shmmap_state *shmmap_s = NULL;
348 vm_offset_t attach_va;
354 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
357 shmmap_s = p->p_vmspace->vm_shm;
358 if (shmmap_s == NULL) {
359 shmmap_s = malloc(shminfo.shmseg * sizeof(struct shmmap_state),
361 for (i = 0; i < shminfo.shmseg; i++)
362 shmmap_s[i].shmid = -1;
363 p->p_vmspace->vm_shm = shmmap_s;
365 shmseg = shm_find_segment_by_shmid(shmid);
366 if (shmseg == NULL) {
370 error = ipcperm(td, &shmseg->u.shm_perm,
371 (shmflg & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W);
375 error = mac_sysvshm_check_shmat(td->td_ucred, shmseg, shmflg);
379 for (i = 0; i < shminfo.shmseg; i++) {
380 if (shmmap_s->shmid == -1)
384 if (i >= shminfo.shmseg) {
388 size = round_page(shmseg->u.shm_segsz);
390 if ((shmflg & SHM_RDONLY) == 0)
391 prot |= VM_PROT_WRITE;
392 flags = MAP_ANON | MAP_SHARED;
395 if (shmflg & SHM_RND) {
396 attach_va = (vm_offset_t)shmaddr & ~(SHMLBA-1);
397 } else if (((vm_offset_t)shmaddr & (SHMLBA-1)) == 0) {
398 attach_va = (vm_offset_t)shmaddr;
405 * This is just a hint to vm_map_find() about where to
409 attach_va = round_page((vm_offset_t)p->p_vmspace->vm_daddr +
410 lim_max(p, RLIMIT_DATA));
414 vm_object_reference(shmseg->object);
415 rv = vm_map_find(&p->p_vmspace->vm_map, shmseg->object,
416 0, &attach_va, size, 0, (flags & MAP_FIXED) ? VMFS_NO_SPACE :
417 VMFS_OPTIMAL_SPACE, prot, prot, MAP_INHERIT_SHARE);
418 if (rv != KERN_SUCCESS) {
419 vm_object_deallocate(shmseg->object);
424 shmmap_s->va = attach_va;
425 shmmap_s->shmid = shmid;
426 shmseg->u.shm_lpid = p->p_pid;
427 shmseg->u.shm_atime = time_second;
428 shmseg->u.shm_nattch++;
429 td->td_retval[0] = attach_va;
438 struct shmat_args *uap;
440 return kern_shmat(td, uap->shmid, uap->shmaddr, uap->shmflg);
444 kern_shmctl(td, shmid, cmd, buf, bufsz)
452 struct shmid_kernel *shmseg;
454 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
460 * It is possible that kern_shmctl is being called from the Linux ABI
461 * layer, in which case, we will need to implement IPC_INFO. It should
462 * be noted that other shmctl calls will be funneled through here for
463 * Linix binaries as well.
465 * NB: The Linux ABI layer will convert this data to structure(s) more
466 * consistent with the Linux ABI.
469 memcpy(buf, &shminfo, sizeof(shminfo));
471 *bufsz = sizeof(shminfo);
472 td->td_retval[0] = shmalloced;
475 struct shm_info shm_info;
476 shm_info.used_ids = shm_nused;
477 shm_info.shm_rss = 0; /*XXX where to get from ? */
478 shm_info.shm_tot = 0; /*XXX where to get from ? */
479 shm_info.shm_swp = 0; /*XXX where to get from ? */
480 shm_info.swap_attempts = 0; /*XXX where to get from ? */
481 shm_info.swap_successes = 0; /*XXX where to get from ? */
482 memcpy(buf, &shm_info, sizeof(shm_info));
484 *bufsz = sizeof(shm_info);
485 td->td_retval[0] = shmalloced;
490 shmseg = shm_find_segment_by_shmidx(shmid);
492 shmseg = shm_find_segment_by_shmid(shmid);
493 if (shmseg == NULL) {
498 error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, cmd);
505 error = ipcperm(td, &shmseg->u.shm_perm, IPC_R);
508 memcpy(buf, &shmseg->u, sizeof(struct shmid_ds));
510 *bufsz = sizeof(struct shmid_ds);
512 td->td_retval[0] = IXSEQ_TO_IPCID(shmid, shmseg->u.shm_perm);
515 struct shmid_ds *shmid;
517 shmid = (struct shmid_ds *)buf;
518 error = ipcperm(td, &shmseg->u.shm_perm, IPC_M);
521 shmseg->u.shm_perm.uid = shmid->shm_perm.uid;
522 shmseg->u.shm_perm.gid = shmid->shm_perm.gid;
523 shmseg->u.shm_perm.mode =
524 (shmseg->u.shm_perm.mode & ~ACCESSPERMS) |
525 (shmid->shm_perm.mode & ACCESSPERMS);
526 shmseg->u.shm_ctime = time_second;
530 error = ipcperm(td, &shmseg->u.shm_perm, IPC_M);
533 shmseg->u.shm_perm.key = IPC_PRIVATE;
534 shmseg->u.shm_perm.mode |= SHMSEG_REMOVED;
535 if (shmseg->u.shm_nattch <= 0) {
536 shm_deallocate_segment(shmseg);
537 shm_last_free = IPCID_TO_IX(shmid);
553 #ifndef _SYS_SYSPROTO_H_
557 struct shmid_ds *buf;
563 struct shmctl_args *uap;
570 * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support
571 * Linux binaries. If we see the call come through the FreeBSD ABI,
572 * return an error back to the user since we do not to support this.
574 if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO ||
575 uap->cmd == SHM_STAT)
578 /* IPC_SET needs to copyin the buffer before calling kern_shmctl */
579 if (uap->cmd == IPC_SET) {
580 if ((error = copyin(uap->buf, &buf, sizeof(struct shmid_ds))))
584 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz);
588 /* Cases in which we need to copyout */
591 error = copyout(&buf, uap->buf, bufsz);
597 /* Invalidate the return value */
598 td->td_retval[0] = -1;
605 shmget_existing(td, uap, mode, segnum)
607 struct shmget_args *uap;
611 struct shmid_kernel *shmseg;
614 shmseg = &shmsegs[segnum];
615 if (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) {
617 * This segment is in the process of being allocated. Wait
618 * until it's done, and look the key up again (in case the
619 * allocation failed or it was freed).
621 shmseg->u.shm_perm.mode |= SHMSEG_WANTED;
622 error = tsleep(shmseg, PLOCK | PCATCH, "shmget", 0);
627 if ((uap->shmflg & (IPC_CREAT | IPC_EXCL)) == (IPC_CREAT | IPC_EXCL))
630 error = mac_sysvshm_check_shmget(td->td_ucred, shmseg, uap->shmflg);
634 if (uap->size != 0 && uap->size > shmseg->u.shm_segsz)
636 td->td_retval[0] = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm);
641 shmget_allocate_segment(td, uap, mode)
643 struct shmget_args *uap;
646 int i, segnum, shmid;
648 struct ucred *cred = td->td_ucred;
649 struct shmid_kernel *shmseg;
650 vm_object_t shm_object;
654 if (uap->size < shminfo.shmmin || uap->size > shminfo.shmmax)
656 if (shm_nused >= shminfo.shmmni) /* Any shmids left? */
658 size = round_page(uap->size);
659 if (shm_committed + btoc(size) > shminfo.shmall)
661 if (shm_last_free < 0) {
662 shmrealloc(); /* Maybe expand the shmsegs[] array. */
663 for (i = 0; i < shmalloced; i++)
664 if (shmsegs[i].u.shm_perm.mode & SHMSEG_FREE)
670 segnum = shm_last_free;
673 shmseg = &shmsegs[segnum];
675 PROC_LOCK(td->td_proc);
676 if (racct_add(td->td_proc, RACCT_NSHM, 1)) {
677 PROC_UNLOCK(td->td_proc);
680 if (racct_add(td->td_proc, RACCT_SHMSIZE, size)) {
681 racct_sub(td->td_proc, RACCT_NSHM, 1);
682 PROC_UNLOCK(td->td_proc);
685 PROC_UNLOCK(td->td_proc);
688 * In case we sleep in malloc(), mark the segment present but deleted
689 * so that noone else tries to create the same key.
691 shmseg->u.shm_perm.mode = SHMSEG_ALLOCATED | SHMSEG_REMOVED;
692 shmseg->u.shm_perm.key = uap->key;
693 shmseg->u.shm_perm.seq = (shmseg->u.shm_perm.seq + 1) & 0x7fff;
694 shmid = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm);
697 * We make sure that we have allocated a pager before we need
700 shm_object = vm_pager_allocate(shm_use_phys ? OBJT_PHYS : OBJT_SWAP,
701 0, size, VM_PROT_DEFAULT, 0, cred);
702 if (shm_object == NULL) {
704 PROC_LOCK(td->td_proc);
705 racct_sub(td->td_proc, RACCT_NSHM, 1);
706 racct_sub(td->td_proc, RACCT_SHMSIZE, size);
707 PROC_UNLOCK(td->td_proc);
711 VM_OBJECT_WLOCK(shm_object);
712 vm_object_clear_flag(shm_object, OBJ_ONEMAPPING);
713 vm_object_set_flag(shm_object, OBJ_NOSPLIT);
714 VM_OBJECT_WUNLOCK(shm_object);
716 shmseg->object = shm_object;
717 shmseg->u.shm_perm.cuid = shmseg->u.shm_perm.uid = cred->cr_uid;
718 shmseg->u.shm_perm.cgid = shmseg->u.shm_perm.gid = cred->cr_gid;
719 shmseg->u.shm_perm.mode = (shmseg->u.shm_perm.mode & SHMSEG_WANTED) |
720 (mode & ACCESSPERMS) | SHMSEG_ALLOCATED;
721 shmseg->cred = crhold(cred);
722 shmseg->u.shm_segsz = uap->size;
723 shmseg->u.shm_cpid = td->td_proc->p_pid;
724 shmseg->u.shm_lpid = shmseg->u.shm_nattch = 0;
725 shmseg->u.shm_atime = shmseg->u.shm_dtime = 0;
727 mac_sysvshm_create(cred, shmseg);
729 shmseg->u.shm_ctime = time_second;
730 shm_committed += btoc(size);
732 if (shmseg->u.shm_perm.mode & SHMSEG_WANTED) {
734 * Somebody else wanted this key while we were asleep. Wake
737 shmseg->u.shm_perm.mode &= ~SHMSEG_WANTED;
740 td->td_retval[0] = shmid;
744 #ifndef _SYS_SYSPROTO_H_
754 struct shmget_args *uap;
759 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
762 mode = uap->shmflg & ACCESSPERMS;
763 if (uap->key != IPC_PRIVATE) {
765 segnum = shm_find_segment_by_key(uap->key);
767 error = shmget_existing(td, uap, mode, segnum);
772 if ((uap->shmflg & IPC_CREAT) == 0) {
777 error = shmget_allocate_segment(td, uap, mode);
784 shmfork_myhook(p1, p2)
785 struct proc *p1, *p2;
787 struct shmmap_state *shmmap_s;
792 size = shminfo.shmseg * sizeof(struct shmmap_state);
793 shmmap_s = malloc(size, M_SHM, M_WAITOK);
794 bcopy(p1->p_vmspace->vm_shm, shmmap_s, size);
795 p2->p_vmspace->vm_shm = shmmap_s;
796 for (i = 0; i < shminfo.shmseg; i++, shmmap_s++)
797 if (shmmap_s->shmid != -1)
798 shmsegs[IPCID_TO_IX(shmmap_s->shmid)].u.shm_nattch++;
803 shmexit_myhook(struct vmspace *vm)
805 struct shmmap_state *base, *shm;
808 if ((base = vm->vm_shm) != NULL) {
811 for (i = 0, shm = base; i < shminfo.shmseg; i++, shm++) {
812 if (shm->shmid != -1)
813 shm_delete_mapping(vm, shm);
824 struct shmid_kernel *newsegs;
826 if (shmalloced >= shminfo.shmmni)
829 newsegs = malloc(shminfo.shmmni * sizeof(*newsegs), M_SHM, M_WAITOK);
832 for (i = 0; i < shmalloced; i++)
833 bcopy(&shmsegs[i], &newsegs[i], sizeof(newsegs[0]));
834 for (; i < shminfo.shmmni; i++) {
835 shmsegs[i].u.shm_perm.mode = SHMSEG_FREE;
836 shmsegs[i].u.shm_perm.seq = 0;
838 mac_sysvshm_init(&shmsegs[i]);
841 free(shmsegs, M_SHM);
843 shmalloced = shminfo.shmmni;
846 static struct syscall_helper_data shm_syscalls[] = {
847 SYSCALL_INIT_HELPER(shmat),
848 SYSCALL_INIT_HELPER(shmctl),
849 SYSCALL_INIT_HELPER(shmdt),
850 SYSCALL_INIT_HELPER(shmget),
851 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
852 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
853 SYSCALL_INIT_HELPER_COMPAT(freebsd7_shmctl),
855 #if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43))
856 SYSCALL_INIT_HELPER(shmsys),
861 #ifdef COMPAT_FREEBSD32
862 #include <compat/freebsd32/freebsd32.h>
863 #include <compat/freebsd32/freebsd32_ipc.h>
864 #include <compat/freebsd32/freebsd32_proto.h>
865 #include <compat/freebsd32/freebsd32_signal.h>
866 #include <compat/freebsd32/freebsd32_syscall.h>
867 #include <compat/freebsd32/freebsd32_util.h>
869 static struct syscall_helper_data shm32_syscalls[] = {
870 SYSCALL32_INIT_HELPER_COMPAT(shmat),
871 SYSCALL32_INIT_HELPER_COMPAT(shmdt),
872 SYSCALL32_INIT_HELPER_COMPAT(shmget),
873 SYSCALL32_INIT_HELPER(freebsd32_shmsys),
874 SYSCALL32_INIT_HELPER(freebsd32_shmctl),
875 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
876 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
877 SYSCALL32_INIT_HELPER(freebsd7_freebsd32_shmctl),
889 if (TUNABLE_ULONG_FETCH("kern.ipc.shmmaxpgs", &shminfo.shmall) != 0)
890 printf("kern.ipc.shmmaxpgs is now called kern.ipc.shmall!\n");
892 TUNABLE_ULONG_FETCH("kern.ipc.shmall", &shminfo.shmall);
893 if (!TUNABLE_ULONG_FETCH("kern.ipc.shmmax", &shminfo.shmmax)) {
894 /* Initialize shmmax dealing with possible overflow. */
895 for (i = PAGE_SIZE; i > 0; i--) {
896 shminfo.shmmax = shminfo.shmall * i;
897 if (shminfo.shmmax >= shminfo.shmall)
901 TUNABLE_ULONG_FETCH("kern.ipc.shmmin", &shminfo.shmmin);
902 TUNABLE_ULONG_FETCH("kern.ipc.shmmni", &shminfo.shmmni);
903 TUNABLE_ULONG_FETCH("kern.ipc.shmseg", &shminfo.shmseg);
904 TUNABLE_INT_FETCH("kern.ipc.shm_use_phys", &shm_use_phys);
906 shmalloced = shminfo.shmmni;
907 shmsegs = malloc(shmalloced * sizeof(shmsegs[0]), M_SHM, M_WAITOK);
908 for (i = 0; i < shmalloced; i++) {
909 shmsegs[i].u.shm_perm.mode = SHMSEG_FREE;
910 shmsegs[i].u.shm_perm.seq = 0;
912 mac_sysvshm_init(&shmsegs[i]);
918 shmexit_hook = &shmexit_myhook;
919 shmfork_hook = &shmfork_myhook;
921 error = syscall_helper_register(shm_syscalls);
924 #ifdef COMPAT_FREEBSD32
925 error = syscall32_helper_register(shm32_syscalls);
940 #ifdef COMPAT_FREEBSD32
941 syscall32_helper_unregister(shm32_syscalls);
943 syscall_helper_unregister(shm_syscalls);
945 for (i = 0; i < shmalloced; i++) {
947 mac_sysvshm_destroy(&shmsegs[i]);
950 * Objects might be still mapped into the processes
951 * address spaces. Actual free would happen on the
952 * last mapping destruction.
954 if (shmsegs[i].u.shm_perm.mode != SHMSEG_FREE)
955 vm_object_deallocate(shmsegs[i].object);
957 free(shmsegs, M_SHM);
964 sysctl_shmsegs(SYSCTL_HANDLER_ARGS)
967 return (SYSCTL_OUT(req, shmsegs, shmalloced * sizeof(shmsegs[0])));
970 #if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43))
972 struct ipc_perm_old shm_perm; /* operation perms */
973 int shm_segsz; /* size of segment (bytes) */
974 u_short shm_cpid; /* pid, creator */
975 u_short shm_lpid; /* pid, last operation */
976 short shm_nattch; /* no. of current attaches */
977 time_t shm_atime; /* last attach time */
978 time_t shm_dtime; /* last detach time */
979 time_t shm_ctime; /* last change time */
980 void *shm_handle; /* internal handle for shm segment */
983 struct oshmctl_args {
986 struct oshmid_ds *ubuf;
990 oshmctl(struct thread *td, struct oshmctl_args *uap)
994 struct shmid_kernel *shmseg;
995 struct oshmid_ds outbuf;
997 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
1000 shmseg = shm_find_segment_by_shmid(uap->shmid);
1001 if (shmseg == NULL) {
1007 error = ipcperm(td, &shmseg->u.shm_perm, IPC_R);
1011 error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, uap->cmd);
1015 ipcperm_new2old(&shmseg->u.shm_perm, &outbuf.shm_perm);
1016 outbuf.shm_segsz = shmseg->u.shm_segsz;
1017 outbuf.shm_cpid = shmseg->u.shm_cpid;
1018 outbuf.shm_lpid = shmseg->u.shm_lpid;
1019 outbuf.shm_nattch = shmseg->u.shm_nattch;
1020 outbuf.shm_atime = shmseg->u.shm_atime;
1021 outbuf.shm_dtime = shmseg->u.shm_dtime;
1022 outbuf.shm_ctime = shmseg->u.shm_ctime;
1023 outbuf.shm_handle = shmseg->object;
1024 error = copyout(&outbuf, uap->ubuf, sizeof(outbuf));
1029 error = freebsd7_shmctl(td, (struct freebsd7_shmctl_args *)uap);
1040 /* XXX casting to (sy_call_t *) is bogus, as usual. */
1041 static sy_call_t *shmcalls[] = {
1042 (sy_call_t *)sys_shmat, (sy_call_t *)oshmctl,
1043 (sy_call_t *)sys_shmdt, (sy_call_t *)sys_shmget,
1044 (sy_call_t *)freebsd7_shmctl
1050 /* XXX actually varargs. */
1051 struct shmsys_args /* {
1060 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
1062 if (uap->which < 0 ||
1063 uap->which >= sizeof(shmcalls)/sizeof(shmcalls[0]))
1066 error = (*shmcalls[uap->which])(td, &uap->a2);
1071 #endif /* i386 && (COMPAT_FREEBSD4 || COMPAT_43) */
1073 #ifdef COMPAT_FREEBSD32
1076 freebsd32_shmsys(struct thread *td, struct freebsd32_shmsys_args *uap)
1079 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1080 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1081 switch (uap->which) {
1082 case 0: { /* shmat */
1083 struct shmat_args ap;
1086 ap.shmaddr = PTRIN(uap->a3);
1087 ap.shmflg = uap->a4;
1088 return (sysent[SYS_shmat].sy_call(td, &ap));
1090 case 2: { /* shmdt */
1091 struct shmdt_args ap;
1093 ap.shmaddr = PTRIN(uap->a2);
1094 return (sysent[SYS_shmdt].sy_call(td, &ap));
1096 case 3: { /* shmget */
1097 struct shmget_args ap;
1101 ap.shmflg = uap->a4;
1102 return (sysent[SYS_shmget].sy_call(td, &ap));
1104 case 4: { /* shmctl */
1105 struct freebsd7_freebsd32_shmctl_args ap;
1109 ap.buf = PTRIN(uap->a4);
1110 return (freebsd7_freebsd32_shmctl(td, &ap));
1112 case 1: /* oshmctl */
1117 return (nosys(td, NULL));
1121 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1122 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1124 freebsd7_freebsd32_shmctl(struct thread *td,
1125 struct freebsd7_freebsd32_shmctl_args *uap)
1129 struct shmid_ds shmid_ds;
1130 struct shm_info shm_info;
1131 struct shminfo shminfo;
1134 struct shmid_ds32_old shmid_ds32;
1135 struct shm_info32 shm_info32;
1136 struct shminfo32 shminfo32;
1140 if (uap->cmd == IPC_SET) {
1141 if ((error = copyin(uap->buf, &u32.shmid_ds32,
1142 sizeof(u32.shmid_ds32))))
1144 freebsd32_ipcperm_old_in(&u32.shmid_ds32.shm_perm,
1145 &u.shmid_ds.shm_perm);
1146 CP(u32.shmid_ds32, u.shmid_ds, shm_segsz);
1147 CP(u32.shmid_ds32, u.shmid_ds, shm_lpid);
1148 CP(u32.shmid_ds32, u.shmid_ds, shm_cpid);
1149 CP(u32.shmid_ds32, u.shmid_ds, shm_nattch);
1150 CP(u32.shmid_ds32, u.shmid_ds, shm_atime);
1151 CP(u32.shmid_ds32, u.shmid_ds, shm_dtime);
1152 CP(u32.shmid_ds32, u.shmid_ds, shm_ctime);
1155 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz);
1159 /* Cases in which we need to copyout */
1162 CP(u.shminfo, u32.shminfo32, shmmax);
1163 CP(u.shminfo, u32.shminfo32, shmmin);
1164 CP(u.shminfo, u32.shminfo32, shmmni);
1165 CP(u.shminfo, u32.shminfo32, shmseg);
1166 CP(u.shminfo, u32.shminfo32, shmall);
1167 error = copyout(&u32.shminfo32, uap->buf,
1168 sizeof(u32.shminfo32));
1171 CP(u.shm_info, u32.shm_info32, used_ids);
1172 CP(u.shm_info, u32.shm_info32, shm_rss);
1173 CP(u.shm_info, u32.shm_info32, shm_tot);
1174 CP(u.shm_info, u32.shm_info32, shm_swp);
1175 CP(u.shm_info, u32.shm_info32, swap_attempts);
1176 CP(u.shm_info, u32.shm_info32, swap_successes);
1177 error = copyout(&u32.shm_info32, uap->buf,
1178 sizeof(u32.shm_info32));
1182 freebsd32_ipcperm_old_out(&u.shmid_ds.shm_perm,
1183 &u32.shmid_ds32.shm_perm);
1184 if (u.shmid_ds.shm_segsz > INT32_MAX)
1185 u32.shmid_ds32.shm_segsz = INT32_MAX;
1187 CP(u.shmid_ds, u32.shmid_ds32, shm_segsz);
1188 CP(u.shmid_ds, u32.shmid_ds32, shm_lpid);
1189 CP(u.shmid_ds, u32.shmid_ds32, shm_cpid);
1190 CP(u.shmid_ds, u32.shmid_ds32, shm_nattch);
1191 CP(u.shmid_ds, u32.shmid_ds32, shm_atime);
1192 CP(u.shmid_ds, u32.shmid_ds32, shm_dtime);
1193 CP(u.shmid_ds, u32.shmid_ds32, shm_ctime);
1194 u32.shmid_ds32.shm_internal = 0;
1195 error = copyout(&u32.shmid_ds32, uap->buf,
1196 sizeof(u32.shmid_ds32));
1202 /* Invalidate the return value */
1203 td->td_retval[0] = -1;
1210 freebsd32_shmctl(struct thread *td, struct freebsd32_shmctl_args *uap)
1214 struct shmid_ds shmid_ds;
1215 struct shm_info shm_info;
1216 struct shminfo shminfo;
1219 struct shmid_ds32 shmid_ds32;
1220 struct shm_info32 shm_info32;
1221 struct shminfo32 shminfo32;
1225 if (uap->cmd == IPC_SET) {
1226 if ((error = copyin(uap->buf, &u32.shmid_ds32,
1227 sizeof(u32.shmid_ds32))))
1229 freebsd32_ipcperm_in(&u32.shmid_ds32.shm_perm,
1230 &u.shmid_ds.shm_perm);
1231 CP(u32.shmid_ds32, u.shmid_ds, shm_segsz);
1232 CP(u32.shmid_ds32, u.shmid_ds, shm_lpid);
1233 CP(u32.shmid_ds32, u.shmid_ds, shm_cpid);
1234 CP(u32.shmid_ds32, u.shmid_ds, shm_nattch);
1235 CP(u32.shmid_ds32, u.shmid_ds, shm_atime);
1236 CP(u32.shmid_ds32, u.shmid_ds, shm_dtime);
1237 CP(u32.shmid_ds32, u.shmid_ds, shm_ctime);
1240 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz);
1244 /* Cases in which we need to copyout */
1247 CP(u.shminfo, u32.shminfo32, shmmax);
1248 CP(u.shminfo, u32.shminfo32, shmmin);
1249 CP(u.shminfo, u32.shminfo32, shmmni);
1250 CP(u.shminfo, u32.shminfo32, shmseg);
1251 CP(u.shminfo, u32.shminfo32, shmall);
1252 error = copyout(&u32.shminfo32, uap->buf,
1253 sizeof(u32.shminfo32));
1256 CP(u.shm_info, u32.shm_info32, used_ids);
1257 CP(u.shm_info, u32.shm_info32, shm_rss);
1258 CP(u.shm_info, u32.shm_info32, shm_tot);
1259 CP(u.shm_info, u32.shm_info32, shm_swp);
1260 CP(u.shm_info, u32.shm_info32, swap_attempts);
1261 CP(u.shm_info, u32.shm_info32, swap_successes);
1262 error = copyout(&u32.shm_info32, uap->buf,
1263 sizeof(u32.shm_info32));
1267 freebsd32_ipcperm_out(&u.shmid_ds.shm_perm,
1268 &u32.shmid_ds32.shm_perm);
1269 if (u.shmid_ds.shm_segsz > INT32_MAX)
1270 u32.shmid_ds32.shm_segsz = INT32_MAX;
1272 CP(u.shmid_ds, u32.shmid_ds32, shm_segsz);
1273 CP(u.shmid_ds, u32.shmid_ds32, shm_lpid);
1274 CP(u.shmid_ds, u32.shmid_ds32, shm_cpid);
1275 CP(u.shmid_ds, u32.shmid_ds32, shm_nattch);
1276 CP(u.shmid_ds, u32.shmid_ds32, shm_atime);
1277 CP(u.shmid_ds, u32.shmid_ds32, shm_dtime);
1278 CP(u.shmid_ds, u32.shmid_ds32, shm_ctime);
1279 error = copyout(&u32.shmid_ds32, uap->buf,
1280 sizeof(u32.shmid_ds32));
1286 /* Invalidate the return value */
1287 td->td_retval[0] = -1;
1293 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1294 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1297 #define CP(src, dst, fld) do { (dst).fld = (src).fld; } while (0)
1300 #ifndef _SYS_SYSPROTO_H_
1301 struct freebsd7_shmctl_args {
1304 struct shmid_ds_old *buf;
1308 freebsd7_shmctl(td, uap)
1310 struct freebsd7_shmctl_args *uap;
1313 struct shmid_ds_old old;
1314 struct shmid_ds buf;
1318 * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support
1319 * Linux binaries. If we see the call come through the FreeBSD ABI,
1320 * return an error back to the user since we do not to support this.
1322 if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO ||
1323 uap->cmd == SHM_STAT)
1326 /* IPC_SET needs to copyin the buffer before calling kern_shmctl */
1327 if (uap->cmd == IPC_SET) {
1328 if ((error = copyin(uap->buf, &old, sizeof(old))))
1330 ipcperm_old2new(&old.shm_perm, &buf.shm_perm);
1331 CP(old, buf, shm_segsz);
1332 CP(old, buf, shm_lpid);
1333 CP(old, buf, shm_cpid);
1334 CP(old, buf, shm_nattch);
1335 CP(old, buf, shm_atime);
1336 CP(old, buf, shm_dtime);
1337 CP(old, buf, shm_ctime);
1340 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz);
1344 /* Cases in which we need to copyout */
1347 ipcperm_new2old(&buf.shm_perm, &old.shm_perm);
1348 if (buf.shm_segsz > INT_MAX)
1349 old.shm_segsz = INT_MAX;
1351 CP(buf, old, shm_segsz);
1352 CP(buf, old, shm_lpid);
1353 CP(buf, old, shm_cpid);
1354 if (buf.shm_nattch > SHRT_MAX)
1355 old.shm_nattch = SHRT_MAX;
1357 CP(buf, old, shm_nattch);
1358 CP(buf, old, shm_atime);
1359 CP(buf, old, shm_dtime);
1360 CP(buf, old, shm_ctime);
1361 old.shm_internal = NULL;
1362 error = copyout(&old, uap->buf, sizeof(old));
1368 /* Invalidate the return value */
1369 td->td_retval[0] = -1;
1374 #endif /* COMPAT_FREEBSD4 || COMPAT_FREEBSD5 || COMPAT_FREEBSD6 ||
1378 sysvshm_modload(struct module *module, int cmd, void *arg)
1389 error = shmunload();
1400 static moduledata_t sysvshm_mod = {
1406 DECLARE_MODULE(sysvshm, sysvshm_mod, SI_SUB_SYSV_SHM, SI_ORDER_FIRST);
1407 MODULE_VERSION(sysvshm, 1);