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>
83 #include <sys/syscall.h>
84 #include <sys/syscallsubr.h>
85 #include <sys/sysent.h>
86 #include <sys/sysproto.h>
89 #include <security/mac/mac_framework.h>
92 #include <vm/vm_param.h>
94 #include <vm/vm_object.h>
95 #include <vm/vm_map.h>
96 #include <vm/vm_page.h>
97 #include <vm/vm_pager.h>
99 FEATURE(sysv_shm, "System V shared memory segments support");
101 static MALLOC_DEFINE(M_SHM, "shm", "SVID compatible shared memory segments");
103 static int shmget_allocate_segment(struct thread *td,
104 struct shmget_args *uap, int mode);
105 static int shmget_existing(struct thread *td, struct shmget_args *uap,
106 int mode, int segnum);
108 #define SHMSEG_FREE 0x0200
109 #define SHMSEG_REMOVED 0x0400
110 #define SHMSEG_ALLOCATED 0x0800
111 #define SHMSEG_WANTED 0x1000
113 static int shm_last_free, shm_nused, shmalloced;
114 vm_size_t shm_committed;
115 static struct shmid_kernel *shmsegs;
117 struct shmmap_state {
122 static void shm_deallocate_segment(struct shmid_kernel *);
123 static int shm_find_segment_by_key(key_t);
124 static struct shmid_kernel *shm_find_segment_by_shmid(int);
125 static struct shmid_kernel *shm_find_segment_by_shmidx(int);
126 static int shm_delete_mapping(struct vmspace *vm, struct shmmap_state *);
127 static void shmrealloc(void);
128 static int shminit(void);
129 static int sysvshm_modload(struct module *, int, void *);
130 static int shmunload(void);
131 static void shmexit_myhook(struct vmspace *vm);
132 static void shmfork_myhook(struct proc *p1, struct proc *p2);
133 static int sysctl_shmsegs(SYSCTL_HANDLER_ARGS);
139 #define SHMMAXPGS 131072 /* Note: sysv shared memory is swap backed. */
142 #define SHMMAX (SHMMAXPGS*PAGE_SIZE)
154 #define SHMALL (SHMMAXPGS)
157 struct shminfo shminfo = {
165 static int shm_use_phys;
166 static int shm_allow_removed;
168 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmax, CTLFLAG_RW, &shminfo.shmmax, 0,
169 "Maximum shared memory segment size");
170 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmin, CTLFLAG_RW, &shminfo.shmmin, 0,
171 "Minimum shared memory segment size");
172 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmni, CTLFLAG_RDTUN, &shminfo.shmmni, 0,
173 "Number of shared memory identifiers");
174 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmseg, CTLFLAG_RDTUN, &shminfo.shmseg, 0,
175 "Number of segments per process");
176 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmall, CTLFLAG_RW, &shminfo.shmall, 0,
177 "Maximum number of pages available for shared memory");
178 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_use_phys, CTLFLAG_RW,
179 &shm_use_phys, 0, "Enable/Disable locking of shared memory pages in core");
180 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_allow_removed, CTLFLAG_RW,
181 &shm_allow_removed, 0,
182 "Enable/Disable attachment to attached segments marked for removal");
183 SYSCTL_PROC(_kern_ipc, OID_AUTO, shmsegs, CTLTYPE_OPAQUE | CTLFLAG_RD,
184 NULL, 0, sysctl_shmsegs, "",
185 "Current number of shared memory segments allocated");
188 shm_find_segment_by_key(key)
193 for (i = 0; i < shmalloced; i++)
194 if ((shmsegs[i].u.shm_perm.mode & SHMSEG_ALLOCATED) &&
195 shmsegs[i].u.shm_perm.key == key)
200 static struct shmid_kernel *
201 shm_find_segment_by_shmid(int shmid)
204 struct shmid_kernel *shmseg;
206 segnum = IPCID_TO_IX(shmid);
207 if (segnum < 0 || segnum >= shmalloced)
209 shmseg = &shmsegs[segnum];
210 if ((shmseg->u.shm_perm.mode & SHMSEG_ALLOCATED) == 0 ||
211 (!shm_allow_removed &&
212 (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) != 0) ||
213 shmseg->u.shm_perm.seq != IPCID_TO_SEQ(shmid))
218 static struct shmid_kernel *
219 shm_find_segment_by_shmidx(int segnum)
221 struct shmid_kernel *shmseg;
223 if (segnum < 0 || segnum >= shmalloced)
225 shmseg = &shmsegs[segnum];
226 if ((shmseg->u.shm_perm.mode & SHMSEG_ALLOCATED) == 0 ||
227 (!shm_allow_removed &&
228 (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) != 0))
234 shm_deallocate_segment(shmseg)
235 struct shmid_kernel *shmseg;
241 vm_object_deallocate(shmseg->object);
242 shmseg->object = NULL;
243 size = round_page(shmseg->u.shm_segsz);
244 shm_committed -= btoc(size);
246 shmseg->u.shm_perm.mode = SHMSEG_FREE;
248 mac_sysvshm_cleanup(shmseg);
250 racct_sub_cred(shmseg->cred, RACCT_NSHM, 1);
251 racct_sub_cred(shmseg->cred, RACCT_SHMSIZE, size);
252 crfree(shmseg->cred);
257 shm_delete_mapping(struct vmspace *vm, struct shmmap_state *shmmap_s)
259 struct shmid_kernel *shmseg;
265 segnum = IPCID_TO_IX(shmmap_s->shmid);
266 shmseg = &shmsegs[segnum];
267 size = round_page(shmseg->u.shm_segsz);
268 result = vm_map_remove(&vm->vm_map, shmmap_s->va, shmmap_s->va + size);
269 if (result != KERN_SUCCESS)
271 shmmap_s->shmid = -1;
272 shmseg->u.shm_dtime = time_second;
273 if ((--shmseg->u.shm_nattch <= 0) &&
274 (shmseg->u.shm_perm.mode & SHMSEG_REMOVED)) {
275 shm_deallocate_segment(shmseg);
276 shm_last_free = segnum;
281 #ifndef _SYS_SYSPROTO_H_
289 struct shmdt_args *uap;
291 struct proc *p = td->td_proc;
292 struct shmmap_state *shmmap_s;
294 struct shmid_kernel *shmsegptr;
299 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
302 shmmap_s = p->p_vmspace->vm_shm;
303 if (shmmap_s == NULL) {
307 for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) {
308 if (shmmap_s->shmid != -1 &&
309 shmmap_s->va == (vm_offset_t)uap->shmaddr) {
313 if (i == shminfo.shmseg) {
318 shmsegptr = &shmsegs[IPCID_TO_IX(shmmap_s->shmid)];
319 error = mac_sysvshm_check_shmdt(td->td_ucred, shmsegptr);
323 error = shm_delete_mapping(p->p_vmspace, shmmap_s);
329 #ifndef _SYS_SYSPROTO_H_
337 kern_shmat(td, shmid, shmaddr, shmflg)
343 struct proc *p = td->td_proc;
345 struct shmid_kernel *shmseg;
346 struct shmmap_state *shmmap_s = NULL;
347 vm_offset_t attach_va;
353 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
356 shmmap_s = p->p_vmspace->vm_shm;
357 if (shmmap_s == NULL) {
358 shmmap_s = malloc(shminfo.shmseg * sizeof(struct shmmap_state),
360 for (i = 0; i < shminfo.shmseg; i++)
361 shmmap_s[i].shmid = -1;
362 p->p_vmspace->vm_shm = shmmap_s;
364 shmseg = shm_find_segment_by_shmid(shmid);
365 if (shmseg == NULL) {
369 error = ipcperm(td, &shmseg->u.shm_perm,
370 (shmflg & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W);
374 error = mac_sysvshm_check_shmat(td->td_ucred, shmseg, shmflg);
378 for (i = 0; i < shminfo.shmseg; i++) {
379 if (shmmap_s->shmid == -1)
383 if (i >= shminfo.shmseg) {
387 size = round_page(shmseg->u.shm_segsz);
389 if ((shmflg & SHM_RDONLY) == 0)
390 prot |= VM_PROT_WRITE;
391 flags = MAP_ANON | MAP_SHARED;
394 if (shmflg & SHM_RND) {
395 attach_va = (vm_offset_t)shmaddr & ~(SHMLBA-1);
396 } else if (((vm_offset_t)shmaddr & (SHMLBA-1)) == 0) {
397 attach_va = (vm_offset_t)shmaddr;
404 * This is just a hint to vm_map_find() about where to
408 attach_va = round_page((vm_offset_t)p->p_vmspace->vm_daddr +
409 lim_max(p, RLIMIT_DATA));
413 vm_object_reference(shmseg->object);
414 rv = vm_map_find(&p->p_vmspace->vm_map, shmseg->object,
415 0, &attach_va, size, (flags & MAP_FIXED) ? VMFS_NO_SPACE :
416 VMFS_OPTIMAL_SPACE, prot, prot, MAP_INHERIT_SHARE);
417 if (rv != KERN_SUCCESS) {
418 vm_object_deallocate(shmseg->object);
423 shmmap_s->va = attach_va;
424 shmmap_s->shmid = shmid;
425 shmseg->u.shm_lpid = p->p_pid;
426 shmseg->u.shm_atime = time_second;
427 shmseg->u.shm_nattch++;
428 td->td_retval[0] = attach_va;
437 struct shmat_args *uap;
439 return kern_shmat(td, uap->shmid, uap->shmaddr, uap->shmflg);
443 kern_shmctl(td, shmid, cmd, buf, bufsz)
451 struct shmid_kernel *shmseg;
453 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
459 * It is possible that kern_shmctl is being called from the Linux ABI
460 * layer, in which case, we will need to implement IPC_INFO. It should
461 * be noted that other shmctl calls will be funneled through here for
462 * Linix binaries as well.
464 * NB: The Linux ABI layer will convert this data to structure(s) more
465 * consistent with the Linux ABI.
468 memcpy(buf, &shminfo, sizeof(shminfo));
470 *bufsz = sizeof(shminfo);
471 td->td_retval[0] = shmalloced;
474 struct shm_info shm_info;
475 shm_info.used_ids = shm_nused;
476 shm_info.shm_rss = 0; /*XXX where to get from ? */
477 shm_info.shm_tot = 0; /*XXX where to get from ? */
478 shm_info.shm_swp = 0; /*XXX where to get from ? */
479 shm_info.swap_attempts = 0; /*XXX where to get from ? */
480 shm_info.swap_successes = 0; /*XXX where to get from ? */
481 memcpy(buf, &shm_info, sizeof(shm_info));
483 *bufsz = sizeof(shm_info);
484 td->td_retval[0] = shmalloced;
489 shmseg = shm_find_segment_by_shmidx(shmid);
491 shmseg = shm_find_segment_by_shmid(shmid);
492 if (shmseg == NULL) {
497 error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, cmd);
504 error = ipcperm(td, &shmseg->u.shm_perm, IPC_R);
507 memcpy(buf, &shmseg->u, sizeof(struct shmid_ds));
509 *bufsz = sizeof(struct shmid_ds);
511 td->td_retval[0] = IXSEQ_TO_IPCID(shmid, shmseg->u.shm_perm);
514 struct shmid_ds *shmid;
516 shmid = (struct shmid_ds *)buf;
517 error = ipcperm(td, &shmseg->u.shm_perm, IPC_M);
520 shmseg->u.shm_perm.uid = shmid->shm_perm.uid;
521 shmseg->u.shm_perm.gid = shmid->shm_perm.gid;
522 shmseg->u.shm_perm.mode =
523 (shmseg->u.shm_perm.mode & ~ACCESSPERMS) |
524 (shmid->shm_perm.mode & ACCESSPERMS);
525 shmseg->u.shm_ctime = time_second;
529 error = ipcperm(td, &shmseg->u.shm_perm, IPC_M);
532 shmseg->u.shm_perm.key = IPC_PRIVATE;
533 shmseg->u.shm_perm.mode |= SHMSEG_REMOVED;
534 if (shmseg->u.shm_nattch <= 0) {
535 shm_deallocate_segment(shmseg);
536 shm_last_free = IPCID_TO_IX(shmid);
552 #ifndef _SYS_SYSPROTO_H_
556 struct shmid_ds *buf;
562 struct shmctl_args *uap;
569 * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support
570 * Linux binaries. If we see the call come through the FreeBSD ABI,
571 * return an error back to the user since we do not to support this.
573 if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO ||
574 uap->cmd == SHM_STAT)
577 /* IPC_SET needs to copyin the buffer before calling kern_shmctl */
578 if (uap->cmd == IPC_SET) {
579 if ((error = copyin(uap->buf, &buf, sizeof(struct shmid_ds))))
583 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz);
587 /* Cases in which we need to copyout */
590 error = copyout(&buf, uap->buf, bufsz);
596 /* Invalidate the return value */
597 td->td_retval[0] = -1;
604 shmget_existing(td, uap, mode, segnum)
606 struct shmget_args *uap;
610 struct shmid_kernel *shmseg;
613 shmseg = &shmsegs[segnum];
614 if (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) {
616 * This segment is in the process of being allocated. Wait
617 * until it's done, and look the key up again (in case the
618 * allocation failed or it was freed).
620 shmseg->u.shm_perm.mode |= SHMSEG_WANTED;
621 error = tsleep(shmseg, PLOCK | PCATCH, "shmget", 0);
626 if ((uap->shmflg & (IPC_CREAT | IPC_EXCL)) == (IPC_CREAT | IPC_EXCL))
629 error = mac_sysvshm_check_shmget(td->td_ucred, shmseg, uap->shmflg);
633 if (uap->size != 0 && uap->size > shmseg->u.shm_segsz)
635 td->td_retval[0] = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm);
640 shmget_allocate_segment(td, uap, mode)
642 struct shmget_args *uap;
645 int i, segnum, shmid;
647 struct ucred *cred = td->td_ucred;
648 struct shmid_kernel *shmseg;
649 vm_object_t shm_object;
653 if (uap->size < shminfo.shmmin || uap->size > shminfo.shmmax)
655 if (shm_nused >= shminfo.shmmni) /* Any shmids left? */
657 size = round_page(uap->size);
658 if (shm_committed + btoc(size) > shminfo.shmall)
660 if (shm_last_free < 0) {
661 shmrealloc(); /* Maybe expand the shmsegs[] array. */
662 for (i = 0; i < shmalloced; i++)
663 if (shmsegs[i].u.shm_perm.mode & SHMSEG_FREE)
669 segnum = shm_last_free;
672 shmseg = &shmsegs[segnum];
674 PROC_LOCK(td->td_proc);
675 if (racct_add(td->td_proc, RACCT_NSHM, 1)) {
676 PROC_UNLOCK(td->td_proc);
679 if (racct_add(td->td_proc, RACCT_SHMSIZE, size)) {
680 racct_sub(td->td_proc, RACCT_NSHM, 1);
681 PROC_UNLOCK(td->td_proc);
684 PROC_UNLOCK(td->td_proc);
687 * In case we sleep in malloc(), mark the segment present but deleted
688 * so that noone else tries to create the same key.
690 shmseg->u.shm_perm.mode = SHMSEG_ALLOCATED | SHMSEG_REMOVED;
691 shmseg->u.shm_perm.key = uap->key;
692 shmseg->u.shm_perm.seq = (shmseg->u.shm_perm.seq + 1) & 0x7fff;
693 shmid = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm);
696 * We make sure that we have allocated a pager before we need
699 shm_object = vm_pager_allocate(shm_use_phys ? OBJT_PHYS : OBJT_SWAP,
700 0, size, VM_PROT_DEFAULT, 0, cred);
701 if (shm_object == NULL) {
703 PROC_LOCK(td->td_proc);
704 racct_sub(td->td_proc, RACCT_NSHM, 1);
705 racct_sub(td->td_proc, RACCT_SHMSIZE, size);
706 PROC_UNLOCK(td->td_proc);
710 VM_OBJECT_LOCK(shm_object);
711 vm_object_clear_flag(shm_object, OBJ_ONEMAPPING);
712 vm_object_set_flag(shm_object, OBJ_NOSPLIT);
713 VM_OBJECT_UNLOCK(shm_object);
715 shmseg->object = shm_object;
716 shmseg->u.shm_perm.cuid = shmseg->u.shm_perm.uid = cred->cr_uid;
717 shmseg->u.shm_perm.cgid = shmseg->u.shm_perm.gid = cred->cr_gid;
718 shmseg->u.shm_perm.mode = (shmseg->u.shm_perm.mode & SHMSEG_WANTED) |
719 (mode & ACCESSPERMS) | SHMSEG_ALLOCATED;
720 shmseg->cred = crhold(cred);
721 shmseg->u.shm_segsz = uap->size;
722 shmseg->u.shm_cpid = td->td_proc->p_pid;
723 shmseg->u.shm_lpid = shmseg->u.shm_nattch = 0;
724 shmseg->u.shm_atime = shmseg->u.shm_dtime = 0;
726 mac_sysvshm_create(cred, shmseg);
728 shmseg->u.shm_ctime = time_second;
729 shm_committed += btoc(size);
731 if (shmseg->u.shm_perm.mode & SHMSEG_WANTED) {
733 * Somebody else wanted this key while we were asleep. Wake
736 shmseg->u.shm_perm.mode &= ~SHMSEG_WANTED;
739 td->td_retval[0] = shmid;
743 #ifndef _SYS_SYSPROTO_H_
753 struct shmget_args *uap;
758 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
761 mode = uap->shmflg & ACCESSPERMS;
762 if (uap->key != IPC_PRIVATE) {
764 segnum = shm_find_segment_by_key(uap->key);
766 error = shmget_existing(td, uap, mode, segnum);
771 if ((uap->shmflg & IPC_CREAT) == 0) {
776 error = shmget_allocate_segment(td, uap, mode);
783 shmfork_myhook(p1, p2)
784 struct proc *p1, *p2;
786 struct shmmap_state *shmmap_s;
791 size = shminfo.shmseg * sizeof(struct shmmap_state);
792 shmmap_s = malloc(size, M_SHM, M_WAITOK);
793 bcopy(p1->p_vmspace->vm_shm, shmmap_s, size);
794 p2->p_vmspace->vm_shm = shmmap_s;
795 for (i = 0; i < shminfo.shmseg; i++, shmmap_s++)
796 if (shmmap_s->shmid != -1)
797 shmsegs[IPCID_TO_IX(shmmap_s->shmid)].u.shm_nattch++;
802 shmexit_myhook(struct vmspace *vm)
804 struct shmmap_state *base, *shm;
807 if ((base = vm->vm_shm) != NULL) {
810 for (i = 0, shm = base; i < shminfo.shmseg; i++, shm++) {
811 if (shm->shmid != -1)
812 shm_delete_mapping(vm, shm);
823 struct shmid_kernel *newsegs;
825 if (shmalloced >= shminfo.shmmni)
828 newsegs = malloc(shminfo.shmmni * sizeof(*newsegs), M_SHM, M_WAITOK);
831 for (i = 0; i < shmalloced; i++)
832 bcopy(&shmsegs[i], &newsegs[i], sizeof(newsegs[0]));
833 for (; i < shminfo.shmmni; i++) {
834 shmsegs[i].u.shm_perm.mode = SHMSEG_FREE;
835 shmsegs[i].u.shm_perm.seq = 0;
837 mac_sysvshm_init(&shmsegs[i]);
840 free(shmsegs, M_SHM);
842 shmalloced = shminfo.shmmni;
845 static struct syscall_helper_data shm_syscalls[] = {
846 SYSCALL_INIT_HELPER(shmat),
847 SYSCALL_INIT_HELPER(shmctl),
848 SYSCALL_INIT_HELPER(shmdt),
849 SYSCALL_INIT_HELPER(shmget),
850 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
851 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
852 SYSCALL_INIT_HELPER_COMPAT(freebsd7_shmctl),
854 #if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43))
855 SYSCALL_INIT_HELPER(shmsys),
860 #ifdef COMPAT_FREEBSD32
861 #include <compat/freebsd32/freebsd32.h>
862 #include <compat/freebsd32/freebsd32_ipc.h>
863 #include <compat/freebsd32/freebsd32_proto.h>
864 #include <compat/freebsd32/freebsd32_signal.h>
865 #include <compat/freebsd32/freebsd32_syscall.h>
866 #include <compat/freebsd32/freebsd32_util.h>
868 static struct syscall_helper_data shm32_syscalls[] = {
869 SYSCALL32_INIT_HELPER_COMPAT(shmat),
870 SYSCALL32_INIT_HELPER_COMPAT(shmdt),
871 SYSCALL32_INIT_HELPER_COMPAT(shmget),
872 SYSCALL32_INIT_HELPER(freebsd32_shmsys),
873 SYSCALL32_INIT_HELPER(freebsd32_shmctl),
874 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
875 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
876 SYSCALL32_INIT_HELPER(freebsd7_freebsd32_shmctl),
888 if (TUNABLE_ULONG_FETCH("kern.ipc.shmmaxpgs", &shminfo.shmall) != 0)
889 printf("kern.ipc.shmmaxpgs is now called kern.ipc.shmall!\n");
891 TUNABLE_ULONG_FETCH("kern.ipc.shmall", &shminfo.shmall);
893 /* Initialize shmmax dealing with possible overflow. */
894 for (i = PAGE_SIZE; i > 0; i--) {
895 shminfo.shmmax = shminfo.shmall * i;
896 if (shminfo.shmmax >= shminfo.shmall)
900 TUNABLE_ULONG_FETCH("kern.ipc.shmmin", &shminfo.shmmin);
901 TUNABLE_ULONG_FETCH("kern.ipc.shmmni", &shminfo.shmmni);
902 TUNABLE_ULONG_FETCH("kern.ipc.shmseg", &shminfo.shmseg);
903 TUNABLE_INT_FETCH("kern.ipc.shm_use_phys", &shm_use_phys);
905 shmalloced = shminfo.shmmni;
906 shmsegs = malloc(shmalloced * sizeof(shmsegs[0]), M_SHM, M_WAITOK);
907 for (i = 0; i < shmalloced; i++) {
908 shmsegs[i].u.shm_perm.mode = SHMSEG_FREE;
909 shmsegs[i].u.shm_perm.seq = 0;
911 mac_sysvshm_init(&shmsegs[i]);
917 shmexit_hook = &shmexit_myhook;
918 shmfork_hook = &shmfork_myhook;
920 error = syscall_helper_register(shm_syscalls);
923 #ifdef COMPAT_FREEBSD32
924 error = syscall32_helper_register(shm32_syscalls);
939 #ifdef COMPAT_FREEBSD32
940 syscall32_helper_unregister(shm32_syscalls);
942 syscall_helper_unregister(shm_syscalls);
944 for (i = 0; i < shmalloced; i++) {
946 mac_sysvshm_destroy(&shmsegs[i]);
949 * Objects might be still mapped into the processes
950 * address spaces. Actual free would happen on the
951 * last mapping destruction.
953 if (shmsegs[i].u.shm_perm.mode != SHMSEG_FREE)
954 vm_object_deallocate(shmsegs[i].object);
956 free(shmsegs, M_SHM);
963 sysctl_shmsegs(SYSCTL_HANDLER_ARGS)
966 return (SYSCTL_OUT(req, shmsegs, shmalloced * sizeof(shmsegs[0])));
969 #if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43))
971 struct ipc_perm_old shm_perm; /* operation perms */
972 int shm_segsz; /* size of segment (bytes) */
973 u_short shm_cpid; /* pid, creator */
974 u_short shm_lpid; /* pid, last operation */
975 short shm_nattch; /* no. of current attaches */
976 time_t shm_atime; /* last attach time */
977 time_t shm_dtime; /* last detach time */
978 time_t shm_ctime; /* last change time */
979 void *shm_handle; /* internal handle for shm segment */
982 struct oshmctl_args {
985 struct oshmid_ds *ubuf;
989 oshmctl(struct thread *td, struct oshmctl_args *uap)
993 struct shmid_kernel *shmseg;
994 struct oshmid_ds outbuf;
996 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
999 shmseg = shm_find_segment_by_shmid(uap->shmid);
1000 if (shmseg == NULL) {
1006 error = ipcperm(td, &shmseg->u.shm_perm, IPC_R);
1010 error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, uap->cmd);
1014 ipcperm_new2old(&shmseg->u.shm_perm, &outbuf.shm_perm);
1015 outbuf.shm_segsz = shmseg->u.shm_segsz;
1016 outbuf.shm_cpid = shmseg->u.shm_cpid;
1017 outbuf.shm_lpid = shmseg->u.shm_lpid;
1018 outbuf.shm_nattch = shmseg->u.shm_nattch;
1019 outbuf.shm_atime = shmseg->u.shm_atime;
1020 outbuf.shm_dtime = shmseg->u.shm_dtime;
1021 outbuf.shm_ctime = shmseg->u.shm_ctime;
1022 outbuf.shm_handle = shmseg->object;
1023 error = copyout(&outbuf, uap->ubuf, sizeof(outbuf));
1028 error = freebsd7_shmctl(td, (struct freebsd7_shmctl_args *)uap);
1039 /* XXX casting to (sy_call_t *) is bogus, as usual. */
1040 static sy_call_t *shmcalls[] = {
1041 (sy_call_t *)sys_shmat, (sy_call_t *)oshmctl,
1042 (sy_call_t *)sys_shmdt, (sy_call_t *)sys_shmget,
1043 (sy_call_t *)freebsd7_shmctl
1049 /* XXX actually varargs. */
1050 struct shmsys_args /* {
1059 if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
1061 if (uap->which < 0 ||
1062 uap->which >= sizeof(shmcalls)/sizeof(shmcalls[0]))
1065 error = (*shmcalls[uap->which])(td, &uap->a2);
1070 #endif /* i386 && (COMPAT_FREEBSD4 || COMPAT_43) */
1072 #ifdef COMPAT_FREEBSD32
1075 freebsd32_shmsys(struct thread *td, struct freebsd32_shmsys_args *uap)
1078 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1079 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1080 switch (uap->which) {
1081 case 0: { /* shmat */
1082 struct shmat_args ap;
1085 ap.shmaddr = PTRIN(uap->a3);
1086 ap.shmflg = uap->a4;
1087 return (sysent[SYS_shmat].sy_call(td, &ap));
1089 case 2: { /* shmdt */
1090 struct shmdt_args ap;
1092 ap.shmaddr = PTRIN(uap->a2);
1093 return (sysent[SYS_shmdt].sy_call(td, &ap));
1095 case 3: { /* shmget */
1096 struct shmget_args ap;
1100 ap.shmflg = uap->a4;
1101 return (sysent[SYS_shmget].sy_call(td, &ap));
1103 case 4: { /* shmctl */
1104 struct freebsd7_freebsd32_shmctl_args ap;
1108 ap.buf = PTRIN(uap->a4);
1109 return (freebsd7_freebsd32_shmctl(td, &ap));
1111 case 1: /* oshmctl */
1116 return (nosys(td, NULL));
1120 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1121 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1123 freebsd7_freebsd32_shmctl(struct thread *td,
1124 struct freebsd7_freebsd32_shmctl_args *uap)
1128 struct shmid_ds shmid_ds;
1129 struct shm_info shm_info;
1130 struct shminfo shminfo;
1133 struct shmid_ds32_old shmid_ds32;
1134 struct shm_info32 shm_info32;
1135 struct shminfo32 shminfo32;
1139 if (uap->cmd == IPC_SET) {
1140 if ((error = copyin(uap->buf, &u32.shmid_ds32,
1141 sizeof(u32.shmid_ds32))))
1143 freebsd32_ipcperm_old_in(&u32.shmid_ds32.shm_perm,
1144 &u.shmid_ds.shm_perm);
1145 CP(u32.shmid_ds32, u.shmid_ds, shm_segsz);
1146 CP(u32.shmid_ds32, u.shmid_ds, shm_lpid);
1147 CP(u32.shmid_ds32, u.shmid_ds, shm_cpid);
1148 CP(u32.shmid_ds32, u.shmid_ds, shm_nattch);
1149 CP(u32.shmid_ds32, u.shmid_ds, shm_atime);
1150 CP(u32.shmid_ds32, u.shmid_ds, shm_dtime);
1151 CP(u32.shmid_ds32, u.shmid_ds, shm_ctime);
1154 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz);
1158 /* Cases in which we need to copyout */
1161 CP(u.shminfo, u32.shminfo32, shmmax);
1162 CP(u.shminfo, u32.shminfo32, shmmin);
1163 CP(u.shminfo, u32.shminfo32, shmmni);
1164 CP(u.shminfo, u32.shminfo32, shmseg);
1165 CP(u.shminfo, u32.shminfo32, shmall);
1166 error = copyout(&u32.shminfo32, uap->buf,
1167 sizeof(u32.shminfo32));
1170 CP(u.shm_info, u32.shm_info32, used_ids);
1171 CP(u.shm_info, u32.shm_info32, shm_rss);
1172 CP(u.shm_info, u32.shm_info32, shm_tot);
1173 CP(u.shm_info, u32.shm_info32, shm_swp);
1174 CP(u.shm_info, u32.shm_info32, swap_attempts);
1175 CP(u.shm_info, u32.shm_info32, swap_successes);
1176 error = copyout(&u32.shm_info32, uap->buf,
1177 sizeof(u32.shm_info32));
1181 freebsd32_ipcperm_old_out(&u.shmid_ds.shm_perm,
1182 &u32.shmid_ds32.shm_perm);
1183 if (u.shmid_ds.shm_segsz > INT32_MAX)
1184 u32.shmid_ds32.shm_segsz = INT32_MAX;
1186 CP(u.shmid_ds, u32.shmid_ds32, shm_segsz);
1187 CP(u.shmid_ds, u32.shmid_ds32, shm_lpid);
1188 CP(u.shmid_ds, u32.shmid_ds32, shm_cpid);
1189 CP(u.shmid_ds, u32.shmid_ds32, shm_nattch);
1190 CP(u.shmid_ds, u32.shmid_ds32, shm_atime);
1191 CP(u.shmid_ds, u32.shmid_ds32, shm_dtime);
1192 CP(u.shmid_ds, u32.shmid_ds32, shm_ctime);
1193 u32.shmid_ds32.shm_internal = 0;
1194 error = copyout(&u32.shmid_ds32, uap->buf,
1195 sizeof(u32.shmid_ds32));
1201 /* Invalidate the return value */
1202 td->td_retval[0] = -1;
1209 freebsd32_shmctl(struct thread *td, struct freebsd32_shmctl_args *uap)
1213 struct shmid_ds shmid_ds;
1214 struct shm_info shm_info;
1215 struct shminfo shminfo;
1218 struct shmid_ds32 shmid_ds32;
1219 struct shm_info32 shm_info32;
1220 struct shminfo32 shminfo32;
1224 if (uap->cmd == IPC_SET) {
1225 if ((error = copyin(uap->buf, &u32.shmid_ds32,
1226 sizeof(u32.shmid_ds32))))
1228 freebsd32_ipcperm_in(&u32.shmid_ds32.shm_perm,
1229 &u.shmid_ds.shm_perm);
1230 CP(u32.shmid_ds32, u.shmid_ds, shm_segsz);
1231 CP(u32.shmid_ds32, u.shmid_ds, shm_lpid);
1232 CP(u32.shmid_ds32, u.shmid_ds, shm_cpid);
1233 CP(u32.shmid_ds32, u.shmid_ds, shm_nattch);
1234 CP(u32.shmid_ds32, u.shmid_ds, shm_atime);
1235 CP(u32.shmid_ds32, u.shmid_ds, shm_dtime);
1236 CP(u32.shmid_ds32, u.shmid_ds, shm_ctime);
1239 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz);
1243 /* Cases in which we need to copyout */
1246 CP(u.shminfo, u32.shminfo32, shmmax);
1247 CP(u.shminfo, u32.shminfo32, shmmin);
1248 CP(u.shminfo, u32.shminfo32, shmmni);
1249 CP(u.shminfo, u32.shminfo32, shmseg);
1250 CP(u.shminfo, u32.shminfo32, shmall);
1251 error = copyout(&u32.shminfo32, uap->buf,
1252 sizeof(u32.shminfo32));
1255 CP(u.shm_info, u32.shm_info32, used_ids);
1256 CP(u.shm_info, u32.shm_info32, shm_rss);
1257 CP(u.shm_info, u32.shm_info32, shm_tot);
1258 CP(u.shm_info, u32.shm_info32, shm_swp);
1259 CP(u.shm_info, u32.shm_info32, swap_attempts);
1260 CP(u.shm_info, u32.shm_info32, swap_successes);
1261 error = copyout(&u32.shm_info32, uap->buf,
1262 sizeof(u32.shm_info32));
1266 freebsd32_ipcperm_out(&u.shmid_ds.shm_perm,
1267 &u32.shmid_ds32.shm_perm);
1268 if (u.shmid_ds.shm_segsz > INT32_MAX)
1269 u32.shmid_ds32.shm_segsz = INT32_MAX;
1271 CP(u.shmid_ds, u32.shmid_ds32, shm_segsz);
1272 CP(u.shmid_ds, u32.shmid_ds32, shm_lpid);
1273 CP(u.shmid_ds, u32.shmid_ds32, shm_cpid);
1274 CP(u.shmid_ds, u32.shmid_ds32, shm_nattch);
1275 CP(u.shmid_ds, u32.shmid_ds32, shm_atime);
1276 CP(u.shmid_ds, u32.shmid_ds32, shm_dtime);
1277 CP(u.shmid_ds, u32.shmid_ds32, shm_ctime);
1278 error = copyout(&u32.shmid_ds32, uap->buf,
1279 sizeof(u32.shmid_ds32));
1285 /* Invalidate the return value */
1286 td->td_retval[0] = -1;
1292 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1293 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1296 #define CP(src, dst, fld) do { (dst).fld = (src).fld; } while (0)
1299 #ifndef _SYS_SYSPROTO_H_
1300 struct freebsd7_shmctl_args {
1303 struct shmid_ds_old *buf;
1307 freebsd7_shmctl(td, uap)
1309 struct freebsd7_shmctl_args *uap;
1312 struct shmid_ds_old old;
1313 struct shmid_ds buf;
1317 * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support
1318 * Linux binaries. If we see the call come through the FreeBSD ABI,
1319 * return an error back to the user since we do not to support this.
1321 if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO ||
1322 uap->cmd == SHM_STAT)
1325 /* IPC_SET needs to copyin the buffer before calling kern_shmctl */
1326 if (uap->cmd == IPC_SET) {
1327 if ((error = copyin(uap->buf, &old, sizeof(old))))
1329 ipcperm_old2new(&old.shm_perm, &buf.shm_perm);
1330 CP(old, buf, shm_segsz);
1331 CP(old, buf, shm_lpid);
1332 CP(old, buf, shm_cpid);
1333 CP(old, buf, shm_nattch);
1334 CP(old, buf, shm_atime);
1335 CP(old, buf, shm_dtime);
1336 CP(old, buf, shm_ctime);
1339 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz);
1343 /* Cases in which we need to copyout */
1346 ipcperm_new2old(&buf.shm_perm, &old.shm_perm);
1347 if (buf.shm_segsz > INT_MAX)
1348 old.shm_segsz = INT_MAX;
1350 CP(buf, old, shm_segsz);
1351 CP(buf, old, shm_lpid);
1352 CP(buf, old, shm_cpid);
1353 if (buf.shm_nattch > SHRT_MAX)
1354 old.shm_nattch = SHRT_MAX;
1356 CP(buf, old, shm_nattch);
1357 CP(buf, old, shm_atime);
1358 CP(buf, old, shm_dtime);
1359 CP(buf, old, shm_ctime);
1360 old.shm_internal = NULL;
1361 error = copyout(&old, uap->buf, sizeof(old));
1367 /* Invalidate the return value */
1368 td->td_retval[0] = -1;
1373 #endif /* COMPAT_FREEBSD4 || COMPAT_FREEBSD5 || COMPAT_FREEBSD6 ||
1377 sysvshm_modload(struct module *module, int cmd, void *arg)
1388 error = shmunload();
1399 static moduledata_t sysvshm_mod = {
1405 DECLARE_MODULE(sysvshm, sysvshm_mod, SI_SUB_SYSV_SHM, SI_ORDER_FIRST);
1406 MODULE_VERSION(sysvshm, 1);