1 /* $FreeBSD: src/sys/kern/sysv_shm.c,v 1.45.2.6 2002/10/22 20:45:03 fjoe Exp $ */
2 /* $DragonFly: src/sys/kern/sysv_shm.c,v 1.20 2006/12/23 23:47:54 swildner Exp $ */
3 /* $NetBSD: sysv_shm.c,v 1.23 1994/07/04 23:25:12 glass Exp $ */
6 * Copyright (c) 1994 Adam Glass and Charles Hannum. All rights reserved.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by Adam Glass and Charles
20 * 4. The names of the authors may not be used to endorse or promote products
21 * derived from this software without specific prior written permission.
23 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR
24 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
25 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
26 * IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT, INDIRECT,
27 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
28 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
32 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 #include "opt_compat.h"
36 #include "opt_sysvipc.h"
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/sysproto.h>
41 #include <sys/kernel.h>
42 #include <sys/sysctl.h>
45 #include <sys/malloc.h>
48 #include <sys/sysent.h>
52 #include <vm/vm_param.h>
55 #include <vm/vm_object.h>
56 #include <vm/vm_map.h>
57 #include <vm/vm_page.h>
58 #include <vm/vm_pager.h>
60 static MALLOC_DEFINE(M_SHM
, "shm", "SVID compatible shared memory segments");
63 static int sys_oshmctl (struct proc
*p
, struct oshmctl_args
*uap
);
65 static int shmget_allocate_segment (struct proc
*p
, struct shmget_args
*uap
, int mode
);
66 static int shmget_existing (struct proc
*p
, struct shmget_args
*uap
, int mode
, int segnum
);
68 /* XXX casting to (sy_call_t *) is bogus, as usual. */
69 static sy_call_t
*shmcalls
[] = {
70 (sy_call_t
*)sys_shmat
, (sy_call_t
*)sys_oshmctl
,
71 (sy_call_t
*)sys_shmdt
, (sy_call_t
*)sys_shmget
,
72 (sy_call_t
*)sys_shmctl
75 #define SHMSEG_FREE 0x0200
76 #define SHMSEG_REMOVED 0x0400
77 #define SHMSEG_ALLOCATED 0x0800
78 #define SHMSEG_WANTED 0x1000
80 static int shm_last_free
, shm_nused
, shm_committed
, shmalloced
;
81 static struct shmid_ds
*shmsegs
;
84 /* vm_offset_t kva; */
85 vm_object_t shm_object
;
93 static void shm_deallocate_segment (struct shmid_ds
*);
94 static int shm_find_segment_by_key (key_t
);
95 static struct shmid_ds
*shm_find_segment_by_shmid (int);
96 static int shm_delete_mapping (struct vmspace
*vm
, struct shmmap_state
*);
97 static void shmrealloc (void);
98 static void shminit (void *);
104 #define SHMMAXPGS 8192 /* note: sysv shared memory is swap backed */
107 #define SHMMAX (SHMMAXPGS*PAGE_SIZE)
119 #define SHMALL (SHMMAXPGS)
122 struct shminfo shminfo
= {
130 static int shm_use_phys
;
132 TUNABLE_INT("kern.ipc.shmmin", &shminfo
.shmmin
);
133 TUNABLE_INT("kern.ipc.shmmni", &shminfo
.shmmni
);
134 TUNABLE_INT("kern.ipc.shmseg", &shminfo
.shmseg
);
135 TUNABLE_INT("kern.ipc.shmmaxpgs", &shminfo
.shmall
);
136 TUNABLE_INT("kern.ipc.shm_use_phys", &shm_use_phys
);
138 SYSCTL_INT(_kern_ipc
, OID_AUTO
, shmmax
, CTLFLAG_RW
, &shminfo
.shmmax
, 0, "");
139 SYSCTL_INT(_kern_ipc
, OID_AUTO
, shmmin
, CTLFLAG_RW
, &shminfo
.shmmin
, 0, "");
140 SYSCTL_INT(_kern_ipc
, OID_AUTO
, shmmni
, CTLFLAG_RD
, &shminfo
.shmmni
, 0, "");
141 SYSCTL_INT(_kern_ipc
, OID_AUTO
, shmseg
, CTLFLAG_RW
, &shminfo
.shmseg
, 0, "");
142 SYSCTL_INT(_kern_ipc
, OID_AUTO
, shmall
, CTLFLAG_RW
, &shminfo
.shmall
, 0, "");
143 SYSCTL_INT(_kern_ipc
, OID_AUTO
, shm_use_phys
, CTLFLAG_RW
, &shm_use_phys
, 0, "");
146 shm_find_segment_by_key(key_t key
)
150 for (i
= 0; i
< shmalloced
; i
++)
151 if ((shmsegs
[i
].shm_perm
.mode
& SHMSEG_ALLOCATED
) &&
152 shmsegs
[i
].shm_perm
.key
== key
)
157 static struct shmid_ds
*
158 shm_find_segment_by_shmid(int shmid
)
161 struct shmid_ds
*shmseg
;
163 segnum
= IPCID_TO_IX(shmid
);
164 if (segnum
< 0 || segnum
>= shmalloced
)
166 shmseg
= &shmsegs
[segnum
];
167 if ((shmseg
->shm_perm
.mode
& (SHMSEG_ALLOCATED
| SHMSEG_REMOVED
))
168 != SHMSEG_ALLOCATED
||
169 shmseg
->shm_perm
.seq
!= IPCID_TO_SEQ(shmid
))
175 shm_deallocate_segment(struct shmid_ds
*shmseg
)
177 struct shm_handle
*shm_handle
;
180 shm_handle
= shmseg
->shm_internal
;
181 vm_object_deallocate(shm_handle
->shm_object
);
182 kfree((caddr_t
)shm_handle
, M_SHM
);
183 shmseg
->shm_internal
= NULL
;
184 size
= round_page(shmseg
->shm_segsz
);
185 shm_committed
-= btoc(size
);
187 shmseg
->shm_perm
.mode
= SHMSEG_FREE
;
191 shm_delete_mapping(struct vmspace
*vm
, struct shmmap_state
*shmmap_s
)
193 struct shmid_ds
*shmseg
;
197 segnum
= IPCID_TO_IX(shmmap_s
->shmid
);
198 shmseg
= &shmsegs
[segnum
];
199 size
= round_page(shmseg
->shm_segsz
);
200 result
= vm_map_remove(&vm
->vm_map
, shmmap_s
->va
, shmmap_s
->va
+ size
);
201 if (result
!= KERN_SUCCESS
)
203 shmmap_s
->shmid
= -1;
204 shmseg
->shm_dtime
= time_second
;
205 if ((--shmseg
->shm_nattch
<= 0) &&
206 (shmseg
->shm_perm
.mode
& SHMSEG_REMOVED
)) {
207 shm_deallocate_segment(shmseg
);
208 shm_last_free
= segnum
;
214 sys_shmdt(struct shmdt_args
*uap
)
216 struct proc
*p
= curproc
;
217 struct shmmap_state
*shmmap_s
;
220 if (!jail_sysvipc_allowed
&& p
->p_ucred
->cr_prison
!= NULL
)
223 shmmap_s
= (struct shmmap_state
*)p
->p_vmspace
->vm_shm
;
224 if (shmmap_s
== NULL
)
226 for (i
= 0; i
< shminfo
.shmseg
; i
++, shmmap_s
++)
227 if (shmmap_s
->shmid
!= -1 &&
228 shmmap_s
->va
== (vm_offset_t
)uap
->shmaddr
)
230 if (i
== shminfo
.shmseg
)
232 return shm_delete_mapping(p
->p_vmspace
, shmmap_s
);
236 sys_shmat(struct shmat_args
*uap
)
238 struct proc
*p
= curproc
;
240 struct shmid_ds
*shmseg
;
241 struct shmmap_state
*shmmap_s
= NULL
;
242 struct shm_handle
*shm_handle
;
243 vm_offset_t attach_va
;
248 if (!jail_sysvipc_allowed
&& p
->p_ucred
->cr_prison
!= NULL
)
251 shmmap_s
= (struct shmmap_state
*)p
->p_vmspace
->vm_shm
;
252 if (shmmap_s
== NULL
) {
253 size
= shminfo
.shmseg
* sizeof(struct shmmap_state
);
254 shmmap_s
= kmalloc(size
, M_SHM
, M_WAITOK
);
255 for (i
= 0; i
< shminfo
.shmseg
; i
++)
256 shmmap_s
[i
].shmid
= -1;
257 p
->p_vmspace
->vm_shm
= (caddr_t
)shmmap_s
;
259 shmseg
= shm_find_segment_by_shmid(uap
->shmid
);
262 error
= ipcperm(p
, &shmseg
->shm_perm
,
263 (uap
->shmflg
& SHM_RDONLY
) ? IPC_R
: IPC_R
|IPC_W
);
266 for (i
= 0; i
< shminfo
.shmseg
; i
++) {
267 if (shmmap_s
->shmid
== -1)
271 if (i
>= shminfo
.shmseg
)
273 size
= round_page(shmseg
->shm_segsz
);
274 #ifdef VM_PROT_READ_IS_EXEC
275 prot
= VM_PROT_READ
| VM_PROT_EXECUTE
;
279 if ((uap
->shmflg
& SHM_RDONLY
) == 0)
280 prot
|= VM_PROT_WRITE
;
281 flags
= MAP_ANON
| MAP_SHARED
;
284 if (uap
->shmflg
& SHM_RND
)
285 attach_va
= (vm_offset_t
)uap
->shmaddr
& ~(SHMLBA
-1);
286 else if (((vm_offset_t
)uap
->shmaddr
& (SHMLBA
-1)) == 0)
287 attach_va
= (vm_offset_t
)uap
->shmaddr
;
291 /* This is just a hint to vm_map_find() about where to put it. */
292 attach_va
= round_page((vm_offset_t
)p
->p_vmspace
->vm_taddr
+ maxtsiz
+ maxdsiz
);
295 shm_handle
= shmseg
->shm_internal
;
296 vm_object_reference(shm_handle
->shm_object
);
297 rv
= vm_map_find(&p
->p_vmspace
->vm_map
,
298 shm_handle
->shm_object
, 0,
300 ((flags
& MAP_FIXED
) ? 0 : 1),
304 if (rv
!= KERN_SUCCESS
) {
305 vm_object_deallocate(shm_handle
->shm_object
);
308 vm_map_inherit(&p
->p_vmspace
->vm_map
,
309 attach_va
, attach_va
+ size
, VM_INHERIT_SHARE
);
311 shmmap_s
->va
= attach_va
;
312 shmmap_s
->shmid
= uap
->shmid
;
313 shmseg
->shm_lpid
= p
->p_pid
;
314 shmseg
->shm_atime
= time_second
;
315 shmseg
->shm_nattch
++;
316 uap
->sysmsg_result
= attach_va
;
321 struct ipc_perm shm_perm
; /* operation perms */
322 int shm_segsz
; /* size of segment (bytes) */
323 ushort shm_cpid
; /* pid, creator */
324 ushort shm_lpid
; /* pid, last operation */
325 short shm_nattch
; /* no. of current attaches */
326 time_t shm_atime
; /* last attach time */
327 time_t shm_dtime
; /* last detach time */
328 time_t shm_ctime
; /* last change time */
329 void *shm_handle
; /* internal handle for shm segment */
332 struct oshmctl_args
{
333 struct sysmsg sysmsg
;
336 struct oshmid_ds
*ubuf
;
340 sys_oshmctl(struct proc
*p
, struct oshmctl_args
*uap
)
344 struct shmid_ds
*shmseg
;
345 struct oshmid_ds outbuf
;
347 if (!jail_sysvipc_allowed
&& p
->p_ucred
->cr_prison
!= NULL
)
350 shmseg
= shm_find_segment_by_shmid(uap
->shmid
);
355 error
= ipcperm(p
, &shmseg
->shm_perm
, IPC_R
);
358 outbuf
.shm_perm
= shmseg
->shm_perm
;
359 outbuf
.shm_segsz
= shmseg
->shm_segsz
;
360 outbuf
.shm_cpid
= shmseg
->shm_cpid
;
361 outbuf
.shm_lpid
= shmseg
->shm_lpid
;
362 outbuf
.shm_nattch
= shmseg
->shm_nattch
;
363 outbuf
.shm_atime
= shmseg
->shm_atime
;
364 outbuf
.shm_dtime
= shmseg
->shm_dtime
;
365 outbuf
.shm_ctime
= shmseg
->shm_ctime
;
366 outbuf
.shm_handle
= shmseg
->shm_internal
;
367 error
= copyout((caddr_t
)&outbuf
, uap
->ubuf
, sizeof(outbuf
));
372 /* XXX casting to (sy_call_t *) is bogus, as usual. */
373 return (sys_shmctl((struct shmctl_args
*)uap
));
382 sys_shmctl(struct shmctl_args
*uap
)
384 struct proc
*p
= curproc
;
386 struct shmid_ds inbuf
;
387 struct shmid_ds
*shmseg
;
389 if (!jail_sysvipc_allowed
&& p
->p_ucred
->cr_prison
!= NULL
)
392 shmseg
= shm_find_segment_by_shmid(uap
->shmid
);
397 error
= ipcperm(p
, &shmseg
->shm_perm
, IPC_R
);
400 error
= copyout((caddr_t
)shmseg
, uap
->buf
, sizeof(inbuf
));
405 error
= ipcperm(p
, &shmseg
->shm_perm
, IPC_M
);
408 error
= copyin(uap
->buf
, (caddr_t
)&inbuf
, sizeof(inbuf
));
411 shmseg
->shm_perm
.uid
= inbuf
.shm_perm
.uid
;
412 shmseg
->shm_perm
.gid
= inbuf
.shm_perm
.gid
;
413 shmseg
->shm_perm
.mode
=
414 (shmseg
->shm_perm
.mode
& ~ACCESSPERMS
) |
415 (inbuf
.shm_perm
.mode
& ACCESSPERMS
);
416 shmseg
->shm_ctime
= time_second
;
419 error
= ipcperm(p
, &shmseg
->shm_perm
, IPC_M
);
422 shmseg
->shm_perm
.key
= IPC_PRIVATE
;
423 shmseg
->shm_perm
.mode
|= SHMSEG_REMOVED
;
424 if (shmseg
->shm_nattch
<= 0) {
425 shm_deallocate_segment(shmseg
);
426 shm_last_free
= IPCID_TO_IX(uap
->shmid
);
440 shmget_existing(struct proc
*p
, struct shmget_args
*uap
, int mode
, int segnum
)
442 struct shmid_ds
*shmseg
;
445 shmseg
= &shmsegs
[segnum
];
446 if (shmseg
->shm_perm
.mode
& SHMSEG_REMOVED
) {
448 * This segment is in the process of being allocated. Wait
449 * until it's done, and look the key up again (in case the
450 * allocation failed or it was freed).
452 shmseg
->shm_perm
.mode
|= SHMSEG_WANTED
;
453 error
= tsleep((caddr_t
)shmseg
, PCATCH
, "shmget", 0);
458 if ((uap
->shmflg
& (IPC_CREAT
| IPC_EXCL
)) == (IPC_CREAT
| IPC_EXCL
))
460 error
= ipcperm(p
, &shmseg
->shm_perm
, mode
);
463 if (uap
->size
&& uap
->size
> shmseg
->shm_segsz
)
465 uap
->sysmsg_result
= IXSEQ_TO_IPCID(segnum
, shmseg
->shm_perm
);
470 shmget_allocate_segment(struct proc
*p
, struct shmget_args
*uap
, int mode
)
472 int i
, segnum
, shmid
, size
;
473 struct ucred
*cred
= p
->p_ucred
;
474 struct shmid_ds
*shmseg
;
475 struct shm_handle
*shm_handle
;
477 if (uap
->size
< shminfo
.shmmin
|| uap
->size
> shminfo
.shmmax
)
479 if (shm_nused
>= shminfo
.shmmni
) /* any shmids left? */
481 size
= round_page(uap
->size
);
482 if (shm_committed
+ btoc(size
) > shminfo
.shmall
)
484 if (shm_last_free
< 0) {
485 shmrealloc(); /* maybe expand the shmsegs[] array */
486 for (i
= 0; i
< shmalloced
; i
++)
487 if (shmsegs
[i
].shm_perm
.mode
& SHMSEG_FREE
)
493 segnum
= shm_last_free
;
496 shmseg
= &shmsegs
[segnum
];
498 * In case we sleep in malloc(), mark the segment present but deleted
499 * so that noone else tries to create the same key.
501 shmseg
->shm_perm
.mode
= SHMSEG_ALLOCATED
| SHMSEG_REMOVED
;
502 shmseg
->shm_perm
.key
= uap
->key
;
503 shmseg
->shm_perm
.seq
= (shmseg
->shm_perm
.seq
+ 1) & 0x7fff;
504 shm_handle
= kmalloc(sizeof(struct shm_handle
), M_SHM
, M_WAITOK
);
505 shmid
= IXSEQ_TO_IPCID(segnum
, shmseg
->shm_perm
);
508 * We make sure that we have allocated a pager before we need
512 shm_handle
->shm_object
=
513 vm_pager_allocate(OBJT_PHYS
, 0, size
, VM_PROT_DEFAULT
, 0);
515 shm_handle
->shm_object
=
516 vm_pager_allocate(OBJT_SWAP
, 0, size
, VM_PROT_DEFAULT
, 0);
518 vm_object_clear_flag(shm_handle
->shm_object
, OBJ_ONEMAPPING
);
519 vm_object_set_flag(shm_handle
->shm_object
, OBJ_NOSPLIT
);
521 shmseg
->shm_internal
= shm_handle
;
522 shmseg
->shm_perm
.cuid
= shmseg
->shm_perm
.uid
= cred
->cr_uid
;
523 shmseg
->shm_perm
.cgid
= shmseg
->shm_perm
.gid
= cred
->cr_gid
;
524 shmseg
->shm_perm
.mode
= (shmseg
->shm_perm
.mode
& SHMSEG_WANTED
) |
525 (mode
& ACCESSPERMS
) | SHMSEG_ALLOCATED
;
526 shmseg
->shm_segsz
= uap
->size
;
527 shmseg
->shm_cpid
= p
->p_pid
;
528 shmseg
->shm_lpid
= shmseg
->shm_nattch
= 0;
529 shmseg
->shm_atime
= shmseg
->shm_dtime
= 0;
530 shmseg
->shm_ctime
= time_second
;
531 shm_committed
+= btoc(size
);
533 if (shmseg
->shm_perm
.mode
& SHMSEG_WANTED
) {
535 * Somebody else wanted this key while we were asleep. Wake
538 shmseg
->shm_perm
.mode
&= ~SHMSEG_WANTED
;
539 wakeup((caddr_t
)shmseg
);
541 uap
->sysmsg_result
= shmid
;
546 sys_shmget(struct shmget_args
*uap
)
548 struct proc
*p
= curproc
;
549 int segnum
, mode
, error
;
551 if (!jail_sysvipc_allowed
&& p
->p_ucred
->cr_prison
!= NULL
)
554 mode
= uap
->shmflg
& ACCESSPERMS
;
555 if (uap
->key
!= IPC_PRIVATE
) {
557 segnum
= shm_find_segment_by_key(uap
->key
);
559 error
= shmget_existing(p
, uap
, mode
, segnum
);
564 if ((uap
->shmflg
& IPC_CREAT
) == 0)
567 return shmget_allocate_segment(p
, uap
, mode
);
571 * shmsys_args(int which, int a2, ...) (VARARGS)
574 sys_shmsys(struct shmsys_args
*uap
)
576 struct proc
*p
= curproc
;
577 unsigned int which
= (unsigned int)uap
->which
;
580 if (!jail_sysvipc_allowed
&& p
->p_ucred
->cr_prison
!= NULL
)
583 if (which
>= sizeof(shmcalls
)/sizeof(shmcalls
[0]))
585 bcopy(&uap
->a2
, &uap
->which
,
586 sizeof(struct shmsys_args
) - offsetof(struct shmsys_args
, a2
));
587 error
= ((*shmcalls
[which
])(uap
));
592 shmfork(struct proc
*p1
, struct proc
*p2
)
594 struct shmmap_state
*shmmap_s
;
598 size
= shminfo
.shmseg
* sizeof(struct shmmap_state
);
599 shmmap_s
= kmalloc(size
, M_SHM
, M_WAITOK
);
600 bcopy((caddr_t
)p1
->p_vmspace
->vm_shm
, (caddr_t
)shmmap_s
, size
);
601 p2
->p_vmspace
->vm_shm
= (caddr_t
)shmmap_s
;
602 for (i
= 0; i
< shminfo
.shmseg
; i
++, shmmap_s
++)
603 if (shmmap_s
->shmid
!= -1)
604 shmsegs
[IPCID_TO_IX(shmmap_s
->shmid
)].shm_nattch
++;
608 shmexit(struct vmspace
*vm
)
610 struct shmmap_state
*base
, *shm
;
613 if ((base
= (struct shmmap_state
*)vm
->vm_shm
) != NULL
) {
615 for (i
= 0, shm
= base
; i
< shminfo
.shmseg
; i
++, shm
++) {
616 if (shm
->shmid
!= -1)
617 shm_delete_mapping(vm
, shm
);
627 struct shmid_ds
*newsegs
;
629 if (shmalloced
>= shminfo
.shmmni
)
632 newsegs
= kmalloc(shminfo
.shmmni
* sizeof(*newsegs
), M_SHM
, M_WAITOK
);
635 for (i
= 0; i
< shmalloced
; i
++)
636 bcopy(&shmsegs
[i
], &newsegs
[i
], sizeof(newsegs
[0]));
637 for (; i
< shminfo
.shmmni
; i
++) {
638 shmsegs
[i
].shm_perm
.mode
= SHMSEG_FREE
;
639 shmsegs
[i
].shm_perm
.seq
= 0;
641 kfree(shmsegs
, M_SHM
);
643 shmalloced
= shminfo
.shmmni
;
651 shminfo
.shmmax
= shminfo
.shmall
* PAGE_SIZE
;
652 shmalloced
= shminfo
.shmmni
;
653 shmsegs
= kmalloc(shmalloced
* sizeof(shmsegs
[0]), M_SHM
, M_WAITOK
);
655 panic("cannot allocate initial memory for sysvshm");
656 for (i
= 0; i
< shmalloced
; i
++) {
657 shmsegs
[i
].shm_perm
.mode
= SHMSEG_FREE
;
658 shmsegs
[i
].shm_perm
.seq
= 0;
664 SYSINIT(sysv_shm
, SI_SUB_SYSV_SHM
, SI_ORDER_FIRST
, shminit
, NULL
);