FreeBSD/Linux Kernel Cross Reference
sys/kern/sysv_shm.c
1 /* $NetBSD: sysv_shm.c,v 1.23 1994/07/04 23:25:12 glass Exp $ */
2 /*-
3 * Copyright (c) 1994 Adam Glass and Charles Hannum. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
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
16 * Hannum.
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.
19 *
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.
30 */
31
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD: releng/5.4/sys/kern/sysv_shm.c 145335 2005-04-20 19:11:07Z cvs2svn $");
34
35 #include "opt_compat.h"
36 #include "opt_sysvipc.h"
37
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/kernel.h>
41 #include <sys/lock.h>
42 #include <sys/sysctl.h>
43 #include <sys/shm.h>
44 #include <sys/proc.h>
45 #include <sys/malloc.h>
46 #include <sys/mman.h>
47 #include <sys/module.h>
48 #include <sys/mutex.h>
49 #include <sys/resourcevar.h>
50 #include <sys/stat.h>
51 #include <sys/syscall.h>
52 #include <sys/syscallsubr.h>
53 #include <sys/sysent.h>
54 #include <sys/sysproto.h>
55 #include <sys/jail.h>
56
57 #include <vm/vm.h>
58 #include <vm/vm_param.h>
59 #include <vm/pmap.h>
60 #include <vm/vm_object.h>
61 #include <vm/vm_map.h>
62 #include <vm/vm_page.h>
63 #include <vm/vm_pager.h>
64
65 static MALLOC_DEFINE(M_SHM, "shm", "SVID compatible shared memory segments");
66
67 struct oshmctl_args;
68 static int oshmctl(struct thread *td, struct oshmctl_args *uap);
69
70 static int shmget_allocate_segment(struct thread *td,
71 struct shmget_args *uap, int mode);
72 static int shmget_existing(struct thread *td, struct shmget_args *uap,
73 int mode, int segnum);
74
75 /* XXX casting to (sy_call_t *) is bogus, as usual. */
76 static sy_call_t *shmcalls[] = {
77 (sy_call_t *)shmat, (sy_call_t *)oshmctl,
78 (sy_call_t *)shmdt, (sy_call_t *)shmget,
79 (sy_call_t *)shmctl
80 };
81
82 #define SHMSEG_FREE 0x0200
83 #define SHMSEG_REMOVED 0x0400
84 #define SHMSEG_ALLOCATED 0x0800
85 #define SHMSEG_WANTED 0x1000
86
87 static int shm_last_free, shm_nused, shm_committed, shmalloced;
88 static struct shmid_ds *shmsegs;
89
90 struct shmmap_state {
91 vm_offset_t va;
92 int shmid;
93 };
94
95 static void shm_deallocate_segment(struct shmid_ds *);
96 static int shm_find_segment_by_key(key_t);
97 static struct shmid_ds *shm_find_segment_by_shmid(int);
98 static struct shmid_ds *shm_find_segment_by_shmidx(int);
99 static int shm_delete_mapping(struct vmspace *vm, struct shmmap_state *);
100 static void shmrealloc(void);
101 static void shminit(void);
102 static int sysvshm_modload(struct module *, int, void *);
103 static int shmunload(void);
104 static void shmexit_myhook(struct vmspace *vm);
105 static void shmfork_myhook(struct proc *p1, struct proc *p2);
106 static int sysctl_shmsegs(SYSCTL_HANDLER_ARGS);
107
108 /*
109 * Tuneable values.
110 */
111 #ifndef SHMMAXPGS
112 #define SHMMAXPGS 8192 /* Note: sysv shared memory is swap backed. */
113 #endif
114 #ifndef SHMMAX
115 #define SHMMAX (SHMMAXPGS*PAGE_SIZE)
116 #endif
117 #ifndef SHMMIN
118 #define SHMMIN 1
119 #endif
120 #ifndef SHMMNI
121 #define SHMMNI 192
122 #endif
123 #ifndef SHMSEG
124 #define SHMSEG 128
125 #endif
126 #ifndef SHMALL
127 #define SHMALL (SHMMAXPGS)
128 #endif
129
130 struct shminfo shminfo = {
131 SHMMAX,
132 SHMMIN,
133 SHMMNI,
134 SHMSEG,
135 SHMALL
136 };
137
138 static int shm_use_phys;
139 static int shm_allow_removed;
140
141 SYSCTL_DECL(_kern_ipc);
142 SYSCTL_INT(_kern_ipc, OID_AUTO, shmmax, CTLFLAG_RW, &shminfo.shmmax, 0,
143 "Maximum shared memory segment size");
144 SYSCTL_INT(_kern_ipc, OID_AUTO, shmmin, CTLFLAG_RW, &shminfo.shmmin, 0,
145 "Minimum shared memory segment size");
146 SYSCTL_INT(_kern_ipc, OID_AUTO, shmmni, CTLFLAG_RDTUN, &shminfo.shmmni, 0,
147 "Number of shared memory identifiers");
148 SYSCTL_INT(_kern_ipc, OID_AUTO, shmseg, CTLFLAG_RDTUN, &shminfo.shmseg, 0,
149 "Number of segments per process");
150 SYSCTL_INT(_kern_ipc, OID_AUTO, shmall, CTLFLAG_RW, &shminfo.shmall, 0,
151 "Maximum number of pages available for shared memory");
152 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_use_phys, CTLFLAG_RW,
153 &shm_use_phys, 0, "Enable/Disable locking of shared memory pages in core");
154 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_allow_removed, CTLFLAG_RW,
155 &shm_allow_removed, 0,
156 "Enable/Disable attachment to attached segments marked for removal");
157 SYSCTL_PROC(_kern_ipc, OID_AUTO, shmsegs, CTLFLAG_RD,
158 NULL, 0, sysctl_shmsegs, "",
159 "Current number of shared memory segments allocated");
160
161 static int
162 shm_find_segment_by_key(key)
163 key_t key;
164 {
165 int i;
166
167 for (i = 0; i < shmalloced; i++)
168 if ((shmsegs[i].shm_perm.mode & SHMSEG_ALLOCATED) &&
169 shmsegs[i].shm_perm.key == key)
170 return (i);
171 return (-1);
172 }
173
174 static struct shmid_ds *
175 shm_find_segment_by_shmid(int shmid)
176 {
177 int segnum;
178 struct shmid_ds *shmseg;
179
180 segnum = IPCID_TO_IX(shmid);
181 if (segnum < 0 || segnum >= shmalloced)
182 return (NULL);
183 shmseg = &shmsegs[segnum];
184 if ((shmseg->shm_perm.mode & SHMSEG_ALLOCATED) == 0 ||
185 (!shm_allow_removed &&
186 (shmseg->shm_perm.mode & SHMSEG_REMOVED) != 0) ||
187 shmseg->shm_perm.seq != IPCID_TO_SEQ(shmid))
188 return (NULL);
189 return (shmseg);
190 }
191
192 static struct shmid_ds *
193 shm_find_segment_by_shmidx(int segnum)
194 {
195 struct shmid_ds *shmseg;
196
197 if (segnum < 0 || segnum >= shmalloced)
198 return (NULL);
199 shmseg = &shmsegs[segnum];
200 if ((shmseg->shm_perm.mode & SHMSEG_ALLOCATED) == 0 ||
201 (!shm_allow_removed &&
202 (shmseg->shm_perm.mode & SHMSEG_REMOVED) != 0))
203 return (NULL);
204 return (shmseg);
205 }
206
207 static void
208 shm_deallocate_segment(shmseg)
209 struct shmid_ds *shmseg;
210 {
211 size_t size;
212
213 GIANT_REQUIRED;
214
215 vm_object_deallocate(shmseg->shm_internal);
216 shmseg->shm_internal = NULL;
217 size = round_page(shmseg->shm_segsz);
218 shm_committed -= btoc(size);
219 shm_nused--;
220 shmseg->shm_perm.mode = SHMSEG_FREE;
221 }
222
223 static int
224 shm_delete_mapping(struct vmspace *vm, struct shmmap_state *shmmap_s)
225 {
226 struct shmid_ds *shmseg;
227 int segnum, result;
228 size_t size;
229
230 GIANT_REQUIRED;
231
232 segnum = IPCID_TO_IX(shmmap_s->shmid);
233 shmseg = &shmsegs[segnum];
234 size = round_page(shmseg->shm_segsz);
235 result = vm_map_remove(&vm->vm_map, shmmap_s->va, shmmap_s->va + size);
236 if (result != KERN_SUCCESS)
237 return (EINVAL);
238 shmmap_s->shmid = -1;
239 shmseg->shm_dtime = time_second;
240 if ((--shmseg->shm_nattch <= 0) &&
241 (shmseg->shm_perm.mode & SHMSEG_REMOVED)) {
242 shm_deallocate_segment(shmseg);
243 shm_last_free = segnum;
244 }
245 return (0);
246 }
247
248 #ifndef _SYS_SYSPROTO_H_
249 struct shmdt_args {
250 const void *shmaddr;
251 };
252 #endif
253
254 /*
255 * MPSAFE
256 */
257 int
258 shmdt(td, uap)
259 struct thread *td;
260 struct shmdt_args *uap;
261 {
262 struct proc *p = td->td_proc;
263 struct shmmap_state *shmmap_s;
264 int i;
265 int error = 0;
266
267 if (!jail_sysvipc_allowed && jailed(td->td_ucred))
268 return (ENOSYS);
269 mtx_lock(&Giant);
270 shmmap_s = p->p_vmspace->vm_shm;
271 if (shmmap_s == NULL) {
272 error = EINVAL;
273 goto done2;
274 }
275 for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) {
276 if (shmmap_s->shmid != -1 &&
277 shmmap_s->va == (vm_offset_t)uap->shmaddr) {
278 break;
279 }
280 }
281 if (i == shminfo.shmseg) {
282 error = EINVAL;
283 goto done2;
284 }
285 error = shm_delete_mapping(p->p_vmspace, shmmap_s);
286 done2:
287 mtx_unlock(&Giant);
288 return (error);
289 }
290
291 #ifndef _SYS_SYSPROTO_H_
292 struct shmat_args {
293 int shmid;
294 const void *shmaddr;
295 int shmflg;
296 };
297 #endif
298
299 /*
300 * MPSAFE
301 */
302 int
303 kern_shmat(td, shmid, shmaddr, shmflg)
304 struct thread *td;
305 int shmid;
306 const void *shmaddr;
307 int shmflg;
308 {
309 struct proc *p = td->td_proc;
310 int i, flags;
311 struct shmid_ds *shmseg;
312 struct shmmap_state *shmmap_s = NULL;
313 vm_offset_t attach_va;
314 vm_prot_t prot;
315 vm_size_t size;
316 int rv;
317 int error = 0;
318
319 if (!jail_sysvipc_allowed && jailed(td->td_ucred))
320 return (ENOSYS);
321 mtx_lock(&Giant);
322 shmmap_s = p->p_vmspace->vm_shm;
323 if (shmmap_s == NULL) {
324 size = shminfo.shmseg * sizeof(struct shmmap_state);
325 shmmap_s = malloc(size, M_SHM, M_WAITOK);
326 for (i = 0; i < shminfo.shmseg; i++)
327 shmmap_s[i].shmid = -1;
328 p->p_vmspace->vm_shm = shmmap_s;
329 }
330 shmseg = shm_find_segment_by_shmid(shmid);
331 if (shmseg == NULL) {
332 error = EINVAL;
333 goto done2;
334 }
335 error = ipcperm(td, &shmseg->shm_perm,
336 (shmflg & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W);
337 if (error)
338 goto done2;
339 for (i = 0; i < shminfo.shmseg; i++) {
340 if (shmmap_s->shmid == -1)
341 break;
342 shmmap_s++;
343 }
344 if (i >= shminfo.shmseg) {
345 error = EMFILE;
346 goto done2;
347 }
348 size = round_page(shmseg->shm_segsz);
349 #ifdef VM_PROT_READ_IS_EXEC
350 prot = VM_PROT_READ | VM_PROT_EXECUTE;
351 #else
352 prot = VM_PROT_READ;
353 #endif
354 if ((shmflg & SHM_RDONLY) == 0)
355 prot |= VM_PROT_WRITE;
356 flags = MAP_ANON | MAP_SHARED;
357 if (shmaddr) {
358 flags |= MAP_FIXED;
359 if (shmflg & SHM_RND) {
360 attach_va = (vm_offset_t)shmaddr & ~(SHMLBA-1);
361 } else if (((vm_offset_t)shmaddr & (SHMLBA-1)) == 0) {
362 attach_va = (vm_offset_t)shmaddr;
363 } else {
364 error = EINVAL;
365 goto done2;
366 }
367 } else {
368 /*
369 * This is just a hint to vm_map_find() about where to
370 * put it.
371 */
372 PROC_LOCK(p);
373 attach_va = round_page((vm_offset_t)p->p_vmspace->vm_daddr +
374 lim_max(p, RLIMIT_DATA));
375 PROC_UNLOCK(p);
376 }
377
378 vm_object_reference(shmseg->shm_internal);
379 rv = vm_map_find(&p->p_vmspace->vm_map, shmseg->shm_internal,
380 0, &attach_va, size, (flags & MAP_FIXED)?0:1, prot, prot, 0);
381 if (rv != KERN_SUCCESS) {
382 vm_object_deallocate(shmseg->shm_internal);
383 error = ENOMEM;
384 goto done2;
385 }
386 vm_map_inherit(&p->p_vmspace->vm_map,
387 attach_va, attach_va + size, VM_INHERIT_SHARE);
388
389 shmmap_s->va = attach_va;
390 shmmap_s->shmid = shmid;
391 shmseg->shm_lpid = p->p_pid;
392 shmseg->shm_atime = time_second;
393 shmseg->shm_nattch++;
394 td->td_retval[0] = attach_va;
395 done2:
396 mtx_unlock(&Giant);
397 return (error);
398 }
399
400 int
401 shmat(td, uap)
402 struct thread *td;
403 struct shmat_args *uap;
404 {
405 return kern_shmat(td, uap->shmid, uap->shmaddr, uap->shmflg);
406 }
407
408 struct oshmid_ds {
409 struct ipc_perm shm_perm; /* operation perms */
410 int shm_segsz; /* size of segment (bytes) */
411 u_short shm_cpid; /* pid, creator */
412 u_short shm_lpid; /* pid, last operation */
413 short shm_nattch; /* no. of current attaches */
414 time_t shm_atime; /* last attach time */
415 time_t shm_dtime; /* last detach time */
416 time_t shm_ctime; /* last change time */
417 void *shm_handle; /* internal handle for shm segment */
418 };
419
420 struct oshmctl_args {
421 int shmid;
422 int cmd;
423 struct oshmid_ds *ubuf;
424 };
425
426 /*
427 * MPSAFE
428 */
429 static int
430 oshmctl(td, uap)
431 struct thread *td;
432 struct oshmctl_args *uap;
433 {
434 #ifdef COMPAT_43
435 int error = 0;
436 struct shmid_ds *shmseg;
437 struct oshmid_ds outbuf;
438
439 if (!jail_sysvipc_allowed && jailed(td->td_ucred))
440 return (ENOSYS);
441 mtx_lock(&Giant);
442 shmseg = shm_find_segment_by_shmid(uap->shmid);
443 if (shmseg == NULL) {
444 error = EINVAL;
445 goto done2;
446 }
447 switch (uap->cmd) {
448 case IPC_STAT:
449 error = ipcperm(td, &shmseg->shm_perm, IPC_R);
450 if (error)
451 goto done2;
452 outbuf.shm_perm = shmseg->shm_perm;
453 outbuf.shm_segsz = shmseg->shm_segsz;
454 outbuf.shm_cpid = shmseg->shm_cpid;
455 outbuf.shm_lpid = shmseg->shm_lpid;
456 outbuf.shm_nattch = shmseg->shm_nattch;
457 outbuf.shm_atime = shmseg->shm_atime;
458 outbuf.shm_dtime = shmseg->shm_dtime;
459 outbuf.shm_ctime = shmseg->shm_ctime;
460 outbuf.shm_handle = shmseg->shm_internal;
461 error = copyout(&outbuf, uap->ubuf, sizeof(outbuf));
462 if (error)
463 goto done2;
464 break;
465 default:
466 error = shmctl(td, (struct shmctl_args *)uap);
467 break;
468 }
469 done2:
470 mtx_unlock(&Giant);
471 return (error);
472 #else
473 return (EINVAL);
474 #endif
475 }
476
477 #ifndef _SYS_SYSPROTO_H_
478 struct shmctl_args {
479 int shmid;
480 int cmd;
481 struct shmid_ds *buf;
482 };
483 #endif
484
485 /*
486 * MPSAFE
487 */
488 int
489 kern_shmctl(td, shmid, cmd, buf, bufsz)
490 struct thread *td;
491 int shmid;
492 int cmd;
493 void *buf;
494 size_t *bufsz;
495 {
496 int error = 0;
497 struct shmid_ds *shmseg;
498
499 if (!jail_sysvipc_allowed && jailed(td->td_ucred))
500 return (ENOSYS);
501
502 mtx_lock(&Giant);
503 switch (cmd) {
504 case IPC_INFO:
505 memcpy(buf, &shminfo, sizeof(shminfo));
506 if (bufsz)
507 *bufsz = sizeof(shminfo);
508 td->td_retval[0] = shmalloced;
509 goto done2;
510 case SHM_INFO: {
511 struct shm_info shm_info;
512 shm_info.used_ids = shm_nused;
513 shm_info.shm_rss = 0; /*XXX where to get from ? */
514 shm_info.shm_tot = 0; /*XXX where to get from ? */
515 shm_info.shm_swp = 0; /*XXX where to get from ? */
516 shm_info.swap_attempts = 0; /*XXX where to get from ? */
517 shm_info.swap_successes = 0; /*XXX where to get from ? */
518 memcpy(buf, &shm_info, sizeof(shm_info));
519 if (bufsz)
520 *bufsz = sizeof(shm_info);
521 td->td_retval[0] = shmalloced;
522 goto done2;
523 }
524 }
525 if (cmd == SHM_STAT)
526 shmseg = shm_find_segment_by_shmidx(shmid);
527 else
528 shmseg = shm_find_segment_by_shmid(shmid);
529 if (shmseg == NULL) {
530 error = EINVAL;
531 goto done2;
532 }
533 switch (cmd) {
534 case SHM_STAT:
535 case IPC_STAT:
536 error = ipcperm(td, &shmseg->shm_perm, IPC_R);
537 if (error)
538 goto done2;
539 memcpy(buf, shmseg, sizeof(struct shmid_ds));
540 if (bufsz)
541 *bufsz = sizeof(struct shmid_ds);
542 if (cmd == SHM_STAT)
543 td->td_retval[0] = IXSEQ_TO_IPCID(shmid, shmseg->shm_perm);
544 break;
545 case IPC_SET: {
546 struct shmid_ds *shmid;
547
548 shmid = (struct shmid_ds *)buf;
549 error = ipcperm(td, &shmseg->shm_perm, IPC_M);
550 if (error)
551 goto done2;
552 shmseg->shm_perm.uid = shmid->shm_perm.uid;
553 shmseg->shm_perm.gid = shmid->shm_perm.gid;
554 shmseg->shm_perm.mode =
555 (shmseg->shm_perm.mode & ~ACCESSPERMS) |
556 (shmid->shm_perm.mode & ACCESSPERMS);
557 shmseg->shm_ctime = time_second;
558 break;
559 }
560 case IPC_RMID:
561 error = ipcperm(td, &shmseg->shm_perm, IPC_M);
562 if (error)
563 goto done2;
564 shmseg->shm_perm.key = IPC_PRIVATE;
565 shmseg->shm_perm.mode |= SHMSEG_REMOVED;
566 if (shmseg->shm_nattch <= 0) {
567 shm_deallocate_segment(shmseg);
568 shm_last_free = IPCID_TO_IX(shmid);
569 }
570 break;
571 #if 0
572 case SHM_LOCK:
573 case SHM_UNLOCK:
574 #endif
575 default:
576 error = EINVAL;
577 break;
578 }
579 done2:
580 mtx_unlock(&Giant);
581 return (error);
582 }
583
584 int
585 shmctl(td, uap)
586 struct thread *td;
587 struct shmctl_args *uap;
588 {
589 int error = 0;
590 struct shmid_ds buf;
591 size_t bufsz;
592
593 /* IPC_SET needs to copyin the buffer before calling kern_shmctl */
594 if (uap->cmd == IPC_SET) {
595 if ((error = copyin(uap->buf, &buf, sizeof(struct shmid_ds))))
596 goto done;
597 }
598
599 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz);
600 if (error)
601 goto done;
602
603 /* Cases in which we need to copyout */
604 switch (uap->cmd) {
605 case IPC_INFO:
606 case SHM_INFO:
607 case SHM_STAT:
608 case IPC_STAT:
609 error = copyout(&buf, uap->buf, bufsz);
610 break;
611 }
612
613 done:
614 if (error) {
615 /* Invalidate the return value */
616 td->td_retval[0] = -1;
617 }
618 return (error);
619 }
620
621
622 #ifndef _SYS_SYSPROTO_H_
623 struct shmget_args {
624 key_t key;
625 size_t size;
626 int shmflg;
627 };
628 #endif
629
630 static int
631 shmget_existing(td, uap, mode, segnum)
632 struct thread *td;
633 struct shmget_args *uap;
634 int mode;
635 int segnum;
636 {
637 struct shmid_ds *shmseg;
638 int error;
639
640 shmseg = &shmsegs[segnum];
641 if (shmseg->shm_perm.mode & SHMSEG_REMOVED) {
642 /*
643 * This segment is in the process of being allocated. Wait
644 * until it's done, and look the key up again (in case the
645 * allocation failed or it was freed).
646 */
647 shmseg->shm_perm.mode |= SHMSEG_WANTED;
648 error = tsleep(shmseg, PLOCK | PCATCH, "shmget", 0);
649 if (error)
650 return (error);
651 return (EAGAIN);
652 }
653 if ((uap->shmflg & (IPC_CREAT | IPC_EXCL)) == (IPC_CREAT | IPC_EXCL))
654 return (EEXIST);
655 error = ipcperm(td, &shmseg->shm_perm, mode);
656 if (error)
657 return (error);
658 if (uap->size && uap->size > shmseg->shm_segsz)
659 return (EINVAL);
660 td->td_retval[0] = IXSEQ_TO_IPCID(segnum, shmseg->shm_perm);
661 return (0);
662 }
663
664 static int
665 shmget_allocate_segment(td, uap, mode)
666 struct thread *td;
667 struct shmget_args *uap;
668 int mode;
669 {
670 int i, segnum, shmid, size;
671 struct ucred *cred = td->td_ucred;
672 struct shmid_ds *shmseg;
673 vm_object_t shm_object;
674
675 GIANT_REQUIRED;
676
677 if (uap->size < shminfo.shmmin || uap->size > shminfo.shmmax)
678 return (EINVAL);
679 if (shm_nused >= shminfo.shmmni) /* Any shmids left? */
680 return (ENOSPC);
681 size = round_page(uap->size);
682 if (shm_committed + btoc(size) > shminfo.shmall)
683 return (ENOMEM);
684 if (shm_last_free < 0) {
685 shmrealloc(); /* Maybe expand the shmsegs[] array. */
686 for (i = 0; i < shmalloced; i++)
687 if (shmsegs[i].shm_perm.mode & SHMSEG_FREE)
688 break;
689 if (i == shmalloced)
690 return (ENOSPC);
691 segnum = i;
692 } else {
693 segnum = shm_last_free;
694 shm_last_free = -1;
695 }
696 shmseg = &shmsegs[segnum];
697 /*
698 * In case we sleep in malloc(), mark the segment present but deleted
699 * so that noone else tries to create the same key.
700 */
701 shmseg->shm_perm.mode = SHMSEG_ALLOCATED | SHMSEG_REMOVED;
702 shmseg->shm_perm.key = uap->key;
703 shmseg->shm_perm.seq = (shmseg->shm_perm.seq + 1) & 0x7fff;
704 shmid = IXSEQ_TO_IPCID(segnum, shmseg->shm_perm);
705
706 /*
707 * We make sure that we have allocated a pager before we need
708 * to.
709 */
710 if (shm_use_phys) {
711 shm_object =
712 vm_pager_allocate(OBJT_PHYS, 0, size, VM_PROT_DEFAULT, 0);
713 } else {
714 shm_object =
715 vm_pager_allocate(OBJT_SWAP, 0, size, VM_PROT_DEFAULT, 0);
716 }
717 VM_OBJECT_LOCK(shm_object);
718 vm_object_clear_flag(shm_object, OBJ_ONEMAPPING);
719 vm_object_set_flag(shm_object, OBJ_NOSPLIT);
720 VM_OBJECT_UNLOCK(shm_object);
721
722 shmseg->shm_internal = shm_object;
723 shmseg->shm_perm.cuid = shmseg->shm_perm.uid = cred->cr_uid;
724 shmseg->shm_perm.cgid = shmseg->shm_perm.gid = cred->cr_gid;
725 shmseg->shm_perm.mode = (shmseg->shm_perm.mode & SHMSEG_WANTED) |
726 (mode & ACCESSPERMS) | SHMSEG_ALLOCATED;
727 shmseg->shm_segsz = uap->size;
728 shmseg->shm_cpid = td->td_proc->p_pid;
729 shmseg->shm_lpid = shmseg->shm_nattch = 0;
730 shmseg->shm_atime = shmseg->shm_dtime = 0;
731 shmseg->shm_ctime = time_second;
732 shm_committed += btoc(size);
733 shm_nused++;
734 if (shmseg->shm_perm.mode & SHMSEG_WANTED) {
735 /*
736 * Somebody else wanted this key while we were asleep. Wake
737 * them up now.
738 */
739 shmseg->shm_perm.mode &= ~SHMSEG_WANTED;
740 wakeup(shmseg);
741 }
742 td->td_retval[0] = shmid;
743 return (0);
744 }
745
746 /*
747 * MPSAFE
748 */
749 int
750 shmget(td, uap)
751 struct thread *td;
752 struct shmget_args *uap;
753 {
754 int segnum, mode;
755 int error;
756
757 if (!jail_sysvipc_allowed && jailed(td->td_ucred))
758 return (ENOSYS);
759 mtx_lock(&Giant);
760 mode = uap->shmflg & ACCESSPERMS;
761 if (uap->key != IPC_PRIVATE) {
762 again:
763 segnum = shm_find_segment_by_key(uap->key);
764 if (segnum >= 0) {
765 error = shmget_existing(td, uap, mode, segnum);
766 if (error == EAGAIN)
767 goto again;
768 goto done2;
769 }
770 if ((uap->shmflg & IPC_CREAT) == 0) {
771 error = ENOENT;
772 goto done2;
773 }
774 }
775 error = shmget_allocate_segment(td, uap, mode);
776 done2:
777 mtx_unlock(&Giant);
778 return (error);
779 }
780
781 /*
782 * MPSAFE
783 */
784 int
785 shmsys(td, uap)
786 struct thread *td;
787 /* XXX actually varargs. */
788 struct shmsys_args /* {
789 int which;
790 int a2;
791 int a3;
792 int a4;
793 } */ *uap;
794 {
795 int error;
796
797 if (!jail_sysvipc_allowed && jailed(td->td_ucred))
798 return (ENOSYS);
799 if (uap->which < 0 ||
800 uap->which >= sizeof(shmcalls)/sizeof(shmcalls[0]))
801 return (EINVAL);
802 mtx_lock(&Giant);
803 error = (*shmcalls[uap->which])(td, &uap->a2);
804 mtx_unlock(&Giant);
805 return (error);
806 }
807
808 static void
809 shmfork_myhook(p1, p2)
810 struct proc *p1, *p2;
811 {
812 struct shmmap_state *shmmap_s;
813 size_t size;
814 int i;
815
816 mtx_lock(&Giant);
817 size = shminfo.shmseg * sizeof(struct shmmap_state);
818 shmmap_s = malloc(size, M_SHM, M_WAITOK);
819 bcopy(p1->p_vmspace->vm_shm, shmmap_s, size);
820 p2->p_vmspace->vm_shm = shmmap_s;
821 for (i = 0; i < shminfo.shmseg; i++, shmmap_s++)
822 if (shmmap_s->shmid != -1)
823 shmsegs[IPCID_TO_IX(shmmap_s->shmid)].shm_nattch++;
824 mtx_unlock(&Giant);
825 }
826
827 static void
828 shmexit_myhook(struct vmspace *vm)
829 {
830 struct shmmap_state *base, *shm;
831 int i;
832
833 if ((base = vm->vm_shm) != NULL) {
834 vm->vm_shm = NULL;
835 mtx_lock(&Giant);
836 for (i = 0, shm = base; i < shminfo.shmseg; i++, shm++) {
837 if (shm->shmid != -1)
838 shm_delete_mapping(vm, shm);
839 }
840 mtx_unlock(&Giant);
841 free(base, M_SHM);
842 }
843 }
844
845 static void
846 shmrealloc(void)
847 {
848 int i;
849 struct shmid_ds *newsegs;
850
851 if (shmalloced >= shminfo.shmmni)
852 return;
853
854 newsegs = malloc(shminfo.shmmni * sizeof(*newsegs), M_SHM, M_WAITOK);
855 if (newsegs == NULL)
856 return;
857 for (i = 0; i < shmalloced; i++)
858 bcopy(&shmsegs[i], &newsegs[i], sizeof(newsegs[0]));
859 for (; i < shminfo.shmmni; i++) {
860 shmsegs[i].shm_perm.mode = SHMSEG_FREE;
861 shmsegs[i].shm_perm.seq = 0;
862 }
863 free(shmsegs, M_SHM);
864 shmsegs = newsegs;
865 shmalloced = shminfo.shmmni;
866 }
867
868 static void
869 shminit()
870 {
871 int i;
872
873 TUNABLE_INT_FETCH("kern.ipc.shmmaxpgs", &shminfo.shmall);
874 for (i = PAGE_SIZE; i > 0; i--) {
875 shminfo.shmmax = shminfo.shmall * PAGE_SIZE;
876 if (shminfo.shmmax >= shminfo.shmall)
877 break;
878 }
879 TUNABLE_INT_FETCH("kern.ipc.shmmin", &shminfo.shmmin);
880 TUNABLE_INT_FETCH("kern.ipc.shmmni", &shminfo.shmmni);
881 TUNABLE_INT_FETCH("kern.ipc.shmseg", &shminfo.shmseg);
882 TUNABLE_INT_FETCH("kern.ipc.shm_use_phys", &shm_use_phys);
883
884 shmalloced = shminfo.shmmni;
885 shmsegs = malloc(shmalloced * sizeof(shmsegs[0]), M_SHM, M_WAITOK);
886 if (shmsegs == NULL)
887 panic("cannot allocate initial memory for sysvshm");
888 for (i = 0; i < shmalloced; i++) {
889 shmsegs[i].shm_perm.mode = SHMSEG_FREE;
890 shmsegs[i].shm_perm.seq = 0;
891 }
892 shm_last_free = 0;
893 shm_nused = 0;
894 shm_committed = 0;
895 shmexit_hook = &shmexit_myhook;
896 shmfork_hook = &shmfork_myhook;
897 }
898
899 static int
900 shmunload()
901 {
902
903 if (shm_nused > 0)
904 return (EBUSY);
905
906 free(shmsegs, M_SHM);
907 shmexit_hook = NULL;
908 shmfork_hook = NULL;
909 return (0);
910 }
911
912 static int
913 sysctl_shmsegs(SYSCTL_HANDLER_ARGS)
914 {
915
916 return (SYSCTL_OUT(req, shmsegs, shmalloced * sizeof(shmsegs[0])));
917 }
918
919 static int
920 sysvshm_modload(struct module *module, int cmd, void *arg)
921 {
922 int error = 0;
923
924 switch (cmd) {
925 case MOD_LOAD:
926 shminit();
927 break;
928 case MOD_UNLOAD:
929 error = shmunload();
930 break;
931 case MOD_SHUTDOWN:
932 break;
933 default:
934 error = EINVAL;
935 break;
936 }
937 return (error);
938 }
939
940 static moduledata_t sysvshm_mod = {
941 "sysvshm",
942 &sysvshm_modload,
943 NULL
944 };
945
946 SYSCALL_MODULE_HELPER(shmsys);
947 SYSCALL_MODULE_HELPER(shmat);
948 SYSCALL_MODULE_HELPER(shmctl);
949 SYSCALL_MODULE_HELPER(shmdt);
950 SYSCALL_MODULE_HELPER(shmget);
951
952 DECLARE_MODULE(sysvshm, sysvshm_mod,
953 SI_SUB_SYSV_SHM, SI_ORDER_FIRST);
954 MODULE_VERSION(sysvshm, 1);
Cache object: cfede77c63a20161d8496a0d8fcc474a
|