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  * Copyright (c) 2003-2005 McAfee, Inc.
  33  * All rights reserved.
  34  *
  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
  38  * program.
  39  *
  40  * Redistribution and use in source and binary forms, with or without
  41  * modification, are permitted provided that the following conditions
  42  * are met:
  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.
  48  *
  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
  59  * SUCH DAMAGE.
  60  */
  61
  62 #include <sys/cdefs.h>
  63 __FBSDID("$FreeBSD$");
  64
  65 #include "opt_compat.h"
  66 #include "opt_sysvipc.h"
  67
  68 #include <sys/param.h>
  69 #include <sys/systm.h>
  70 #include <sys/kernel.h>
  71 #include <sys/limits.h>
  72 #include <sys/lock.h>
  73 #include <sys/sysctl.h>
  74 #include <sys/shm.h>
  75 #include <sys/proc.h>
  76 #include <sys/malloc.h>
  77 #include <sys/mman.h>
  78 #include <sys/module.h>
  79 #include <sys/mutex.h>
  80 #include <sys/resourcevar.h>
  81 #include <sys/stat.h>
  82 #include <sys/syscall.h>
  83 #include <sys/syscallsubr.h>
  84 #include <sys/sysent.h>
  85 #include <sys/sysproto.h>
  86 #include <sys/jail.h>
  87
  88 #include <security/mac/mac_framework.h>
  89
  90 #include <vm/vm.h>
  91 #include <vm/vm_param.h>
  92 #include <vm/pmap.h>
  93 #include <vm/vm_object.h>
  94 #include <vm/vm_map.h>
  95 #include <vm/vm_page.h>
  96 #include <vm/vm_pager.h>
  97
  98 static MALLOC_DEFINE(M_SHM, "shm", "SVID compatible shared memory segments");
  99
 100 static int shmget_allocate_segment(struct thread *td,
 101     struct shmget_args *uap, int mode);
 102 static int shmget_existing(struct thread *td, struct shmget_args *uap,
 103     int mode, int segnum);
 104
 105 #define SHMSEG_FREE             0x0200
 106 #define SHMSEG_REMOVED          0x0400
 107 #define SHMSEG_ALLOCATED        0x0800
 108 #define SHMSEG_WANTED           0x1000
 109
 110 static int shm_last_free, shm_nused, shmalloced;
 111 vm_size_t shm_committed;
 112 static struct shmid_kernel      *shmsegs;
 113
 114 struct shmmap_state {
 115         vm_offset_t va;
 116         int shmid;
 117 };
 118
 119 static void shm_deallocate_segment(struct shmid_kernel *);
 120 static int shm_find_segment_by_key(key_t);
 121 static struct shmid_kernel *shm_find_segment_by_shmid(int);
 122 static struct shmid_kernel *shm_find_segment_by_shmidx(int);
 123 static int shm_delete_mapping(struct vmspace *vm, struct shmmap_state *);
 124 static void shmrealloc(void);
 125 static int shminit(void);
 126 static int sysvshm_modload(struct module *, int, void *);
 127 static int shmunload(void);
 128 static void shmexit_myhook(struct vmspace *vm);
 129 static void shmfork_myhook(struct proc *p1, struct proc *p2);
 130 static int sysctl_shmsegs(SYSCTL_HANDLER_ARGS);
 131
 132 /*
 133  * Tuneable values.
 134  */
 135 #ifndef SHMMAXPGS
 136 #define SHMMAXPGS       131072  /* Note: sysv shared memory is swap backed. */
 137 #endif
 138 #ifndef SHMMAX
 139 #define SHMMAX  (SHMMAXPGS*PAGE_SIZE)
 140 #endif
 141 #ifndef SHMMIN
 142 #define SHMMIN  1
 143 #endif
 144 #ifndef SHMMNI
 145 #define SHMMNI  192
 146 #endif
 147 #ifndef SHMSEG
 148 #define SHMSEG  128
 149 #endif
 150 #ifndef SHMALL
 151 #define SHMALL  (SHMMAXPGS)
 152 #endif
 153
 154 struct  shminfo shminfo = {
 155         SHMMAX,
 156         SHMMIN,
 157         SHMMNI,
 158         SHMSEG,
 159         SHMALL
 160 };
 161
 162 static int shm_use_phys;
 163 static int shm_allow_removed;
 164
 165 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmax, CTLFLAG_RW, &shminfo.shmmax, 0,
 166     "Maximum shared memory segment size");
 167 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmin, CTLFLAG_RW, &shminfo.shmmin, 0,
 168     "Minimum shared memory segment size");
 169 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmni, CTLFLAG_RDTUN, &shminfo.shmmni, 0,
 170     "Number of shared memory identifiers");
 171 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmseg, CTLFLAG_RDTUN, &shminfo.shmseg, 0,
 172     "Number of segments per process");
 173 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmall, CTLFLAG_RW, &shminfo.shmall, 0,
 174     "Maximum number of pages available for shared memory");
 175 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_use_phys, CTLFLAG_RW,
 176     &shm_use_phys, 0, "Enable/Disable locking of shared memory pages in core");
 177 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_allow_removed, CTLFLAG_RW,
 178     &shm_allow_removed, 0,
 179     "Enable/Disable attachment to attached segments marked for removal");
 180 SYSCTL_PROC(_kern_ipc, OID_AUTO, shmsegs, CTLFLAG_RD,
 181     NULL, 0, sysctl_shmsegs, "",
 182     "Current number of shared memory segments allocated");
 183
 184 static int
 185 shm_find_segment_by_key(key)
 186         key_t key;
 187 {
 188         int i;
 189
 190         for (i = 0; i < shmalloced; i++)
 191                 if ((shmsegs[i].u.shm_perm.mode & SHMSEG_ALLOCATED) &&
 192                     shmsegs[i].u.shm_perm.key == key)
 193                         return (i);
 194         return (-1);
 195 }
 196
 197 static struct shmid_kernel *
 198 shm_find_segment_by_shmid(int shmid)
 199 {
 200         int segnum;
 201         struct shmid_kernel *shmseg;
 202
 203         segnum = IPCID_TO_IX(shmid);
 204         if (segnum < 0 || segnum >= shmalloced)
 205                 return (NULL);
 206         shmseg = &shmsegs[segnum];
 207         if ((shmseg->u.shm_perm.mode & SHMSEG_ALLOCATED) == 0 ||
 208             (!shm_allow_removed &&
 209              (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) != 0) ||
 210             shmseg->u.shm_perm.seq != IPCID_TO_SEQ(shmid))
 211                 return (NULL);
 212         return (shmseg);
 213 }
 214
 215 static struct shmid_kernel *
 216 shm_find_segment_by_shmidx(int segnum)
 217 {
 218         struct shmid_kernel *shmseg;
 219
 220         if (segnum < 0 || segnum >= shmalloced)
 221                 return (NULL);
 222         shmseg = &shmsegs[segnum];
 223         if ((shmseg->u.shm_perm.mode & SHMSEG_ALLOCATED) == 0 ||
 224             (!shm_allow_removed &&
 225              (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) != 0))
 226                 return (NULL);
 227         return (shmseg);
 228 }
 229
 230 static void
 231 shm_deallocate_segment(shmseg)
 232         struct shmid_kernel *shmseg;
 233 {
 234         vm_size_t size;
 235
 236         GIANT_REQUIRED;
 237
 238         vm_object_deallocate(shmseg->object);
 239         shmseg->object = NULL;
 240         size = round_page(shmseg->u.shm_segsz);
 241         shm_committed -= btoc(size);
 242         shm_nused--;
 243         shmseg->u.shm_perm.mode = SHMSEG_FREE;
 244 #ifdef MAC
 245         mac_sysvshm_cleanup(shmseg);
 246 #endif
 247 }
 248
 249 static int
 250 shm_delete_mapping(struct vmspace *vm, struct shmmap_state *shmmap_s)
 251 {
 252         struct shmid_kernel *shmseg;
 253         int segnum, result;
 254         vm_size_t size;
 255
 256         GIANT_REQUIRED;
 257
 258         segnum = IPCID_TO_IX(shmmap_s->shmid);
 259         shmseg = &shmsegs[segnum];
 260         size = round_page(shmseg->u.shm_segsz);
 261         result = vm_map_remove(&vm->vm_map, shmmap_s->va, shmmap_s->va + size);
 262         if (result != KERN_SUCCESS)
 263                 return (EINVAL);
 264         shmmap_s->shmid = -1;
 265         shmseg->u.shm_dtime = time_second;
 266         if ((--shmseg->u.shm_nattch <= 0) &&
 267             (shmseg->u.shm_perm.mode & SHMSEG_REMOVED)) {
 268                 shm_deallocate_segment(shmseg);
 269                 shm_last_free = segnum;
 270         }
 271         return (0);
 272 }
 273
 274 #ifndef _SYS_SYSPROTO_H_
 275 struct shmdt_args {
 276         const void *shmaddr;
 277 };
 278 #endif
 279 int
 280 shmdt(td, uap)
 281         struct thread *td;
 282         struct shmdt_args *uap;
 283 {
 284         struct proc *p = td->td_proc;
 285         struct shmmap_state *shmmap_s;
 286 #ifdef MAC
 287         struct shmid_kernel *shmsegptr;
 288 #endif
 289         int i;
 290         int error = 0;
 291
 292         if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
 293                 return (ENOSYS);
 294         mtx_lock(&Giant);
 295         shmmap_s = p->p_vmspace->vm_shm;
 296         if (shmmap_s == NULL) {
 297                 error = EINVAL;
 298                 goto done2;
 299         }
 300         for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) {
 301                 if (shmmap_s->shmid != -1 &&
 302                     shmmap_s->va == (vm_offset_t)uap->shmaddr) {
 303                         break;
 304                 }
 305         }
 306         if (i == shminfo.shmseg) {
 307                 error = EINVAL;
 308                 goto done2;
 309         }
 310 #ifdef MAC
 311         shmsegptr = &shmsegs[IPCID_TO_IX(shmmap_s->shmid)];
 312         error = mac_sysvshm_check_shmdt(td->td_ucred, shmsegptr);
 313         if (error != 0)
 314                 goto done2;
 315 #endif
 316         error = shm_delete_mapping(p->p_vmspace, shmmap_s);
 317 done2:
 318         mtx_unlock(&Giant);
 319         return (error);
 320 }
 321
 322 #ifndef _SYS_SYSPROTO_H_
 323 struct shmat_args {
 324         int shmid;
 325         const void *shmaddr;
 326         int shmflg;
 327 };
 328 #endif
 329 int
 330 kern_shmat(td, shmid, shmaddr, shmflg)
 331         struct thread *td;
 332         int shmid;
 333         const void *shmaddr;
 334         int shmflg;
 335 {
 336         struct proc *p = td->td_proc;
 337         int i, flags;
 338         struct shmid_kernel *shmseg;
 339         struct shmmap_state *shmmap_s = NULL;
 340         vm_offset_t attach_va;
 341         vm_prot_t prot;
 342         vm_size_t size;
 343         int rv;
 344         int error = 0;
 345
 346         if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
 347                 return (ENOSYS);
 348         mtx_lock(&Giant);
 349         shmmap_s = p->p_vmspace->vm_shm;
 350         if (shmmap_s == NULL) {
 351                 shmmap_s = malloc(shminfo.shmseg * sizeof(struct shmmap_state),
 352                     M_SHM, M_WAITOK);
 353                 for (i = 0; i < shminfo.shmseg; i++)
 354                         shmmap_s[i].shmid = -1;
 355                 p->p_vmspace->vm_shm = shmmap_s;
 356         }
 357         shmseg = shm_find_segment_by_shmid(shmid);
 358         if (shmseg == NULL) {
 359                 error = EINVAL;
 360                 goto done2;
 361         }
 362         error = ipcperm(td, &shmseg->u.shm_perm,
 363             (shmflg & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W);
 364         if (error)
 365                 goto done2;
 366 #ifdef MAC
 367         error = mac_sysvshm_check_shmat(td->td_ucred, shmseg, shmflg);
 368         if (error != 0)
 369                 goto done2;
 370 #endif
 371         for (i = 0; i < shminfo.shmseg; i++) {
 372                 if (shmmap_s->shmid == -1)
 373                         break;
 374                 shmmap_s++;
 375         }
 376         if (i >= shminfo.shmseg) {
 377                 error = EMFILE;
 378                 goto done2;
 379         }
 380         size = round_page(shmseg->u.shm_segsz);
 381         prot = VM_PROT_READ;
 382         if ((shmflg & SHM_RDONLY) == 0)
 383                 prot |= VM_PROT_WRITE;
 384         flags = MAP_ANON | MAP_SHARED;
 385         if (shmaddr) {
 386                 flags |= MAP_FIXED;
 387                 if (shmflg & SHM_RND) {
 388                         attach_va = (vm_offset_t)shmaddr & ~(SHMLBA-1);
 389                 } else if (((vm_offset_t)shmaddr & (SHMLBA-1)) == 0) {
 390                         attach_va = (vm_offset_t)shmaddr;
 391                 } else {
 392                         error = EINVAL;
 393                         goto done2;
 394                 }
 395         } else {
 396                 /*
 397                  * This is just a hint to vm_map_find() about where to
 398                  * put it.
 399                  */
 400                 PROC_LOCK(p);
 401                 attach_va = round_page((vm_offset_t)p->p_vmspace->vm_daddr +
 402                     lim_max(p, RLIMIT_DATA));
 403                 PROC_UNLOCK(p);
 404         }
 405
 406         vm_object_reference(shmseg->object);
 407         rv = vm_map_find(&p->p_vmspace->vm_map, shmseg->object,
 408             0, &attach_va, size, (flags & MAP_FIXED) ? VMFS_NO_SPACE :
 409             VMFS_ANY_SPACE, prot, prot, 0);
 410         if (rv != KERN_SUCCESS) {
 411                 vm_object_deallocate(shmseg->object);
 412                 error = ENOMEM;
 413                 goto done2;
 414         }
 415         vm_map_inherit(&p->p_vmspace->vm_map,
 416                 attach_va, attach_va + size, VM_INHERIT_SHARE);
 417
 418         shmmap_s->va = attach_va;
 419         shmmap_s->shmid = shmid;
 420         shmseg->u.shm_lpid = p->p_pid;
 421         shmseg->u.shm_atime = time_second;
 422         shmseg->u.shm_nattch++;
 423         td->td_retval[0] = attach_va;
 424 done2:
 425         mtx_unlock(&Giant);
 426         return (error);
 427 }
 428
 429 int
 430 shmat(td, uap)
 431         struct thread *td;
 432         struct shmat_args *uap;
 433 {
 434         return kern_shmat(td, uap->shmid, uap->shmaddr, uap->shmflg);
 435 }
 436
 437 int
 438 kern_shmctl(td, shmid, cmd, buf, bufsz)
 439         struct thread *td;
 440         int shmid;
 441         int cmd;
 442         void *buf;
 443         size_t *bufsz;
 444 {
 445         int error = 0;
 446         struct shmid_kernel *shmseg;
 447
 448         if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
 449                 return (ENOSYS);
 450
 451         mtx_lock(&Giant);
 452         switch (cmd) {
 453         /*
 454          * It is possible that kern_shmctl is being called from the Linux ABI
 455          * layer, in which case, we will need to implement IPC_INFO.  It should
 456          * be noted that other shmctl calls will be funneled through here for
 457          * Linix binaries as well.
 458          *
 459          * NB: The Linux ABI layer will convert this data to structure(s) more
 460          * consistent with the Linux ABI.
 461          */
 462         case IPC_INFO:
 463                 memcpy(buf, &shminfo, sizeof(shminfo));
 464                 if (bufsz)
 465                         *bufsz = sizeof(shminfo);
 466                 td->td_retval[0] = shmalloced;
 467                 goto done2;
 468         case SHM_INFO: {
 469                 struct shm_info shm_info;
 470                 shm_info.used_ids = shm_nused;
 471                 shm_info.shm_rss = 0;   /*XXX where to get from ? */
 472                 shm_info.shm_tot = 0;   /*XXX where to get from ? */
 473                 shm_info.shm_swp = 0;   /*XXX where to get from ? */
 474                 shm_info.swap_attempts = 0;     /*XXX where to get from ? */
 475                 shm_info.swap_successes = 0;    /*XXX where to get from ? */
 476                 memcpy(buf, &shm_info, sizeof(shm_info));
 477                 if (bufsz)
 478                         *bufsz = sizeof(shm_info);
 479                 td->td_retval[0] = shmalloced;
 480                 goto done2;
 481         }
 482         }
 483         if (cmd == SHM_STAT)
 484                 shmseg = shm_find_segment_by_shmidx(shmid);
 485         else
 486                 shmseg = shm_find_segment_by_shmid(shmid);
 487         if (shmseg == NULL) {
 488                 error = EINVAL;
 489                 goto done2;
 490         }
 491 #ifdef MAC
 492         error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, cmd);
 493         if (error != 0)
 494                 goto done2;
 495 #endif
 496         switch (cmd) {
 497         case SHM_STAT:
 498         case IPC_STAT:
 499                 error = ipcperm(td, &shmseg->u.shm_perm, IPC_R);
 500                 if (error)
 501                         goto done2;
 502                 memcpy(buf, &shmseg->u, sizeof(struct shmid_ds));
 503                 if (bufsz)
 504                         *bufsz = sizeof(struct shmid_ds);
 505                 if (cmd == SHM_STAT)
 506                         td->td_retval[0] = IXSEQ_TO_IPCID(shmid, shmseg->u.shm_perm);
 507                 break;
 508         case IPC_SET: {
 509                 struct shmid_ds *shmid;
 510
 511                 shmid = (struct shmid_ds *)buf;
 512                 error = ipcperm(td, &shmseg->u.shm_perm, IPC_M);
 513                 if (error)
 514                         goto done2;
 515                 shmseg->u.shm_perm.uid = shmid->shm_perm.uid;
 516                 shmseg->u.shm_perm.gid = shmid->shm_perm.gid;
 517                 shmseg->u.shm_perm.mode =
 518                     (shmseg->u.shm_perm.mode & ~ACCESSPERMS) |
 519                     (shmid->shm_perm.mode & ACCESSPERMS);
 520                 shmseg->u.shm_ctime = time_second;
 521                 break;
 522         }
 523         case IPC_RMID:
 524                 error = ipcperm(td, &shmseg->u.shm_perm, IPC_M);
 525                 if (error)
 526                         goto done2;
 527                 shmseg->u.shm_perm.key = IPC_PRIVATE;
 528                 shmseg->u.shm_perm.mode |= SHMSEG_REMOVED;
 529                 if (shmseg->u.shm_nattch <= 0) {
 530                         shm_deallocate_segment(shmseg);
 531                         shm_last_free = IPCID_TO_IX(shmid);
 532                 }
 533                 break;
 534 #if 0
 535         case SHM_LOCK:
 536         case SHM_UNLOCK:
 537 #endif
 538         default:
 539                 error = EINVAL;
 540                 break;
 541         }
 542 done2:
 543         mtx_unlock(&Giant);
 544         return (error);
 545 }
 546
 547 #ifndef _SYS_SYSPROTO_H_
 548 struct shmctl_args {
 549         int shmid;
 550         int cmd;
 551         struct shmid_ds *buf;
 552 };
 553 #endif
 554 int
 555 shmctl(td, uap)
 556         struct thread *td;
 557         struct shmctl_args *uap;
 558 {
 559         int error = 0;
 560         struct shmid_ds buf;
 561         size_t bufsz;
 562
 563         /*
 564          * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support
 565          * Linux binaries.  If we see the call come through the FreeBSD ABI,
 566          * return an error back to the user since we do not to support this.
 567          */
 568         if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO ||
 569             uap->cmd == SHM_STAT)
 570                 return (EINVAL);
 571
 572         /* IPC_SET needs to copyin the buffer before calling kern_shmctl */
 573         if (uap->cmd == IPC_SET) {
 574                 if ((error = copyin(uap->buf, &buf, sizeof(struct shmid_ds))))
 575                         goto done;
 576         }
 577
 578         error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz);
 579         if (error)
 580                 goto done;
 581
 582         /* Cases in which we need to copyout */
 583         switch (uap->cmd) {
 584         case IPC_STAT:
 585                 error = copyout(&buf, uap->buf, bufsz);
 586                 break;
 587         }
 588
 589 done:
 590         if (error) {
 591                 /* Invalidate the return value */
 592                 td->td_retval[0] = -1;
 593         }
 594         return (error);
 595 }
 596
 597
 598 static int
 599 shmget_existing(td, uap, mode, segnum)
 600         struct thread *td;
 601         struct shmget_args *uap;
 602         int mode;
 603         int segnum;
 604 {
 605         struct shmid_kernel *shmseg;
 606         int error;
 607
 608         shmseg = &shmsegs[segnum];
 609         if (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) {
 610                 /*
 611                  * This segment is in the process of being allocated.  Wait
 612                  * until it's done, and look the key up again (in case the
 613                  * allocation failed or it was freed).
 614                  */
 615                 shmseg->u.shm_perm.mode |= SHMSEG_WANTED;
 616                 error = tsleep(shmseg, PLOCK | PCATCH, "shmget", 0);
 617                 if (error)
 618                         return (error);
 619                 return (EAGAIN);
 620         }
 621         if ((uap->shmflg & (IPC_CREAT | IPC_EXCL)) == (IPC_CREAT | IPC_EXCL))
 622                 return (EEXIST);
 623 #ifdef MAC
 624         error = mac_sysvshm_check_shmget(td->td_ucred, shmseg, uap->shmflg);
 625         if (error != 0)
 626                 return (error);
 627 #endif
 628         if (uap->size != 0 && uap->size > shmseg->u.shm_segsz)
 629                 return (EINVAL);
 630         td->td_retval[0] = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm);
 631         return (0);
 632 }
 633
 634 static int
 635 shmget_allocate_segment(td, uap, mode)
 636         struct thread *td;
 637         struct shmget_args *uap;
 638         int mode;
 639 {
 640         int i, segnum, shmid;
 641         size_t size;
 642         struct ucred *cred = td->td_ucred;
 643         struct shmid_kernel *shmseg;
 644         vm_object_t shm_object;
 645
 646         GIANT_REQUIRED;
 647
 648         if (uap->size < shminfo.shmmin || uap->size > shminfo.shmmax)
 649                 return (EINVAL);
 650         if (shm_nused >= shminfo.shmmni) /* Any shmids left? */
 651                 return (ENOSPC);
 652         size = round_page(uap->size);
 653         if (shm_committed + btoc(size) > shminfo.shmall)
 654                 return (ENOMEM);
 655         if (shm_last_free < 0) {
 656                 shmrealloc();   /* Maybe expand the shmsegs[] array. */
 657                 for (i = 0; i < shmalloced; i++)
 658                         if (shmsegs[i].u.shm_perm.mode & SHMSEG_FREE)
 659                                 break;
 660                 if (i == shmalloced)
 661                         return (ENOSPC);
 662                 segnum = i;
 663         } else  {
 664                 segnum = shm_last_free;
 665                 shm_last_free = -1;
 666         }
 667         shmseg = &shmsegs[segnum];
 668         /*
 669          * In case we sleep in malloc(), mark the segment present but deleted
 670          * so that noone else tries to create the same key.
 671          */
 672         shmseg->u.shm_perm.mode = SHMSEG_ALLOCATED | SHMSEG_REMOVED;
 673         shmseg->u.shm_perm.key = uap->key;
 674         shmseg->u.shm_perm.seq = (shmseg->u.shm_perm.seq + 1) & 0x7fff;
 675         shmid = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm);
 676
 677         /*
 678          * We make sure that we have allocated a pager before we need
 679          * to.
 680          */
 681         shm_object = vm_pager_allocate(shm_use_phys ? OBJT_PHYS : OBJT_SWAP,
 682             0, size, VM_PROT_DEFAULT, 0, cred);
 683         if (shm_object == NULL)
 684                 return (ENOMEM);
 685         VM_OBJECT_LOCK(shm_object);
 686         vm_object_clear_flag(shm_object, OBJ_ONEMAPPING);
 687         vm_object_set_flag(shm_object, OBJ_NOSPLIT);
 688         VM_OBJECT_UNLOCK(shm_object);
 689
 690         shmseg->object = shm_object;
 691         shmseg->u.shm_perm.cuid = shmseg->u.shm_perm.uid = cred->cr_uid;
 692         shmseg->u.shm_perm.cgid = shmseg->u.shm_perm.gid = cred->cr_gid;
 693         shmseg->u.shm_perm.mode = (shmseg->u.shm_perm.mode & SHMSEG_WANTED) |
 694             (mode & ACCESSPERMS) | SHMSEG_ALLOCATED;
 695         shmseg->u.shm_segsz = uap->size;
 696         shmseg->u.shm_cpid = td->td_proc->p_pid;
 697         shmseg->u.shm_lpid = shmseg->u.shm_nattch = 0;
 698         shmseg->u.shm_atime = shmseg->u.shm_dtime = 0;
 699 #ifdef MAC
 700         mac_sysvshm_create(cred, shmseg);
 701 #endif
 702         shmseg->u.shm_ctime = time_second;
 703         shm_committed += btoc(size);
 704         shm_nused++;
 705         if (shmseg->u.shm_perm.mode & SHMSEG_WANTED) {
 706                 /*
 707                  * Somebody else wanted this key while we were asleep.  Wake
 708                  * them up now.
 709                  */
 710                 shmseg->u.shm_perm.mode &= ~SHMSEG_WANTED;
 711                 wakeup(shmseg);
 712         }
 713         td->td_retval[0] = shmid;
 714         return (0);
 715 }
 716
 717 #ifndef _SYS_SYSPROTO_H_
 718 struct shmget_args {
 719         key_t key;
 720         size_t size;
 721         int shmflg;
 722 };
 723 #endif
 724 int
 725 shmget(td, uap)
 726         struct thread *td;
 727         struct shmget_args *uap;
 728 {
 729         int segnum, mode;
 730         int error;
 731
 732         if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
 733                 return (ENOSYS);
 734         mtx_lock(&Giant);
 735         mode = uap->shmflg & ACCESSPERMS;
 736         if (uap->key != IPC_PRIVATE) {
 737         again:
 738                 segnum = shm_find_segment_by_key(uap->key);
 739                 if (segnum >= 0) {
 740                         error = shmget_existing(td, uap, mode, segnum);
 741                         if (error == EAGAIN)
 742                                 goto again;
 743                         goto done2;
 744                 }
 745                 if ((uap->shmflg & IPC_CREAT) == 0) {
 746                         error = ENOENT;
 747                         goto done2;
 748                 }
 749         }
 750         error = shmget_allocate_segment(td, uap, mode);
 751 done2:
 752         mtx_unlock(&Giant);
 753         return (error);
 754 }
 755
 756 static void
 757 shmfork_myhook(p1, p2)
 758         struct proc *p1, *p2;
 759 {
 760         struct shmmap_state *shmmap_s;
 761         size_t size;
 762         int i;
 763
 764         mtx_lock(&Giant);
 765         size = shminfo.shmseg * sizeof(struct shmmap_state);
 766         shmmap_s = malloc(size, M_SHM, M_WAITOK);
 767         bcopy(p1->p_vmspace->vm_shm, shmmap_s, size);
 768         p2->p_vmspace->vm_shm = shmmap_s;
 769         for (i = 0; i < shminfo.shmseg; i++, shmmap_s++)
 770                 if (shmmap_s->shmid != -1)
 771                         shmsegs[IPCID_TO_IX(shmmap_s->shmid)].u.shm_nattch++;
 772         mtx_unlock(&Giant);
 773 }
 774
 775 static void
 776 shmexit_myhook(struct vmspace *vm)
 777 {
 778         struct shmmap_state *base, *shm;
 779         int i;
 780
 781         if ((base = vm->vm_shm) != NULL) {
 782                 vm->vm_shm = NULL;
 783                 mtx_lock(&Giant);
 784                 for (i = 0, shm = base; i < shminfo.shmseg; i++, shm++) {
 785                         if (shm->shmid != -1)
 786                                 shm_delete_mapping(vm, shm);
 787                 }
 788                 mtx_unlock(&Giant);
 789                 free(base, M_SHM);
 790         }
 791 }
 792
 793 static void
 794 shmrealloc(void)
 795 {
 796         int i;
 797         struct shmid_kernel *newsegs;
 798
 799         if (shmalloced >= shminfo.shmmni)
 800                 return;
 801
 802         newsegs = malloc(shminfo.shmmni * sizeof(*newsegs), M_SHM, M_WAITOK);
 803         if (newsegs == NULL)
 804                 return;
 805         for (i = 0; i < shmalloced; i++)
 806                 bcopy(&shmsegs[i], &newsegs[i], sizeof(newsegs[0]));
 807         for (; i < shminfo.shmmni; i++) {
 808                 shmsegs[i].u.shm_perm.mode = SHMSEG_FREE;
 809                 shmsegs[i].u.shm_perm.seq = 0;
 810 #ifdef MAC
 811                 mac_sysvshm_init(&shmsegs[i]);
 812 #endif
 813         }
 814         free(shmsegs, M_SHM);
 815         shmsegs = newsegs;
 816         shmalloced = shminfo.shmmni;
 817 }
 818
 819 static struct syscall_helper_data shm_syscalls[] = {
 820         SYSCALL_INIT_HELPER(shmat),
 821         SYSCALL_INIT_HELPER(shmctl),
 822         SYSCALL_INIT_HELPER(shmdt),
 823         SYSCALL_INIT_HELPER(shmget),
 824 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
 825     defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
 826         SYSCALL_INIT_HELPER(freebsd7_shmctl),
 827 #endif
 828 #if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43))
 829         SYSCALL_INIT_HELPER(shmsys),
 830 #endif
 831         SYSCALL_INIT_LAST
 832 };
 833
 834 #ifdef COMPAT_FREEBSD32
 835 #include <compat/freebsd32/freebsd32.h>
 836 #include <compat/freebsd32/freebsd32_ipc.h>
 837 #include <compat/freebsd32/freebsd32_proto.h>
 838 #include <compat/freebsd32/freebsd32_signal.h>
 839 #include <compat/freebsd32/freebsd32_syscall.h>
 840 #include <compat/freebsd32/freebsd32_util.h>
 841
 842 static struct syscall_helper_data shm32_syscalls[] = {
 843         SYSCALL32_INIT_HELPER(shmat),
 844         SYSCALL32_INIT_HELPER(shmdt),
 845         SYSCALL32_INIT_HELPER(shmget),
 846         SYSCALL32_INIT_HELPER(freebsd32_shmsys),
 847         SYSCALL32_INIT_HELPER(freebsd32_shmctl),
 848 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
 849     defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
 850         SYSCALL32_INIT_HELPER(freebsd7_freebsd32_shmctl),
 851 #endif
 852         SYSCALL_INIT_LAST
 853 };
 854 #endif
 855
 856 static int
 857 shminit()
 858 {
 859         int i, error;
 860
 861 #ifndef BURN_BRIDGES
 862         if (TUNABLE_ULONG_FETCH("kern.ipc.shmmaxpgs", &shminfo.shmall) != 0)
 863                 printf("kern.ipc.shmmaxpgs is now called kern.ipc.shmall!\n");
 864 #endif
 865         TUNABLE_ULONG_FETCH("kern.ipc.shmall", &shminfo.shmall);
 866
 867         /* Initialize shmmax dealing with possible overflow. */
 868         for (i = PAGE_SIZE; i > 0; i--) {
 869                 shminfo.shmmax = shminfo.shmall * i;
 870                 if (shminfo.shmmax >= shminfo.shmall)
 871                         break;
 872         }
 873
 874         TUNABLE_ULONG_FETCH("kern.ipc.shmmin", &shminfo.shmmin);
 875         TUNABLE_ULONG_FETCH("kern.ipc.shmmni", &shminfo.shmmni);
 876         TUNABLE_ULONG_FETCH("kern.ipc.shmseg", &shminfo.shmseg);
 877         TUNABLE_INT_FETCH("kern.ipc.shm_use_phys", &shm_use_phys);
 878
 879         shmalloced = shminfo.shmmni;
 880         shmsegs = malloc(shmalloced * sizeof(shmsegs[0]), M_SHM, M_WAITOK);
 881         if (shmsegs == NULL)
 882                 panic("cannot allocate initial memory for sysvshm");
 883         for (i = 0; i < shmalloced; i++) {
 884                 shmsegs[i].u.shm_perm.mode = SHMSEG_FREE;
 885                 shmsegs[i].u.shm_perm.seq = 0;
 886 #ifdef MAC
 887                 mac_sysvshm_init(&shmsegs[i]);
 888 #endif
 889         }
 890         shm_last_free = 0;
 891         shm_nused = 0;
 892         shm_committed = 0;
 893         shmexit_hook = &shmexit_myhook;
 894         shmfork_hook = &shmfork_myhook;
 895
 896         error = syscall_helper_register(shm_syscalls);
 897         if (error != 0)
 898                 return (error);
 899 #ifdef COMPAT_FREEBSD32
 900         error = syscall32_helper_register(shm32_syscalls);
 901         if (error != 0)
 902                 return (error);
 903 #endif
 904         return (0);
 905 }
 906
 907 static int
 908 shmunload()
 909 {
 910         int i;
 911
 912         if (shm_nused > 0)
 913                 return (EBUSY);
 914
 915 #ifdef COMPAT_FREEBSD32
 916         syscall32_helper_unregister(shm32_syscalls);
 917 #endif
 918         syscall_helper_unregister(shm_syscalls);
 919
 920         for (i = 0; i < shmalloced; i++) {
 921 #ifdef MAC
 922                 mac_sysvshm_destroy(&shmsegs[i]);
 923 #endif
 924                 /*
 925                  * Objects might be still mapped into the processes
 926                  * address spaces.  Actual free would happen on the
 927                  * last mapping destruction.
 928                  */
 929                 if (shmsegs[i].u.shm_perm.mode != SHMSEG_FREE)
 930                         vm_object_deallocate(shmsegs[i].object);
 931         }
 932         free(shmsegs, M_SHM);
 933         shmexit_hook = NULL;
 934         shmfork_hook = NULL;
 935         return (0);
 936 }
 937
 938 static int
 939 sysctl_shmsegs(SYSCTL_HANDLER_ARGS)
 940 {
 941
 942         return (SYSCTL_OUT(req, shmsegs, shmalloced * sizeof(shmsegs[0])));
 943 }
 944
 945 #if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43))
 946 struct oshmid_ds {
 947         struct  ipc_perm_old shm_perm;  /* operation perms */
 948         int     shm_segsz;              /* size of segment (bytes) */
 949         u_short shm_cpid;               /* pid, creator */
 950         u_short shm_lpid;               /* pid, last operation */
 951         short   shm_nattch;             /* no. of current attaches */
 952         time_t  shm_atime;              /* last attach time */
 953         time_t  shm_dtime;              /* last detach time */
 954         time_t  shm_ctime;              /* last change time */
 955         void    *shm_handle;            /* internal handle for shm segment */
 956 };
 957
 958 struct oshmctl_args {
 959         int shmid;
 960         int cmd;
 961         struct oshmid_ds *ubuf;
 962 };
 963
 964 static int
 965 oshmctl(struct thread *td, struct oshmctl_args *uap)
 966 {
 967 #ifdef COMPAT_43
 968         int error = 0;
 969         struct shmid_kernel *shmseg;
 970         struct oshmid_ds outbuf;
 971
 972         if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
 973                 return (ENOSYS);
 974         mtx_lock(&Giant);
 975         shmseg = shm_find_segment_by_shmid(uap->shmid);
 976         if (shmseg == NULL) {
 977                 error = EINVAL;
 978                 goto done2;
 979         }
 980         switch (uap->cmd) {
 981         case IPC_STAT:
 982                 error = ipcperm(td, &shmseg->u.shm_perm, IPC_R);
 983                 if (error)
 984                         goto done2;
 985 #ifdef MAC
 986                 error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, uap->cmd);
 987                 if (error != 0)
 988                         goto done2;
 989 #endif
 990                 ipcperm_new2old(&shmseg->u.shm_perm, &outbuf.shm_perm);
 991                 outbuf.shm_segsz = shmseg->u.shm_segsz;
 992                 outbuf.shm_cpid = shmseg->u.shm_cpid;
 993                 outbuf.shm_lpid = shmseg->u.shm_lpid;
 994                 outbuf.shm_nattch = shmseg->u.shm_nattch;
 995                 outbuf.shm_atime = shmseg->u.shm_atime;
 996                 outbuf.shm_dtime = shmseg->u.shm_dtime;
 997                 outbuf.shm_ctime = shmseg->u.shm_ctime;
 998                 outbuf.shm_handle = shmseg->object;
 999                 error = copyout(&outbuf, uap->ubuf, sizeof(outbuf));
1000                 if (error)
1001                         goto done2;
1002                 break;
1003         default:
1004                 error = freebsd7_shmctl(td, (struct freebsd7_shmctl_args *)uap);
1005                 break;
1006         }
1007 done2:
1008         mtx_unlock(&Giant);
1009         return (error);
1010 #else
1011         return (EINVAL);
1012 #endif
1013 }
1014
1015 /* XXX casting to (sy_call_t *) is bogus, as usual. */
1016 static sy_call_t *shmcalls[] = {
1017         (sy_call_t *)shmat, (sy_call_t *)oshmctl,
1018         (sy_call_t *)shmdt, (sy_call_t *)shmget,
1019         (sy_call_t *)freebsd7_shmctl
1020 };
1021
1022 int
1023 shmsys(td, uap)
1024         struct thread *td;
1025         /* XXX actually varargs. */
1026         struct shmsys_args /* {
1027                 int     which;
1028                 int     a2;
1029                 int     a3;
1030                 int     a4;
1031         } */ *uap;
1032 {
1033         int error;
1034
1035         if (!prison_allow(td->td_ucred, PR_ALLOW_SYSVIPC))
1036                 return (ENOSYS);
1037         if (uap->which < 0 ||
1038             uap->which >= sizeof(shmcalls)/sizeof(shmcalls[0]))
1039                 return (EINVAL);
1040         mtx_lock(&Giant);
1041         error = (*shmcalls[uap->which])(td, &uap->a2);
1042         mtx_unlock(&Giant);
1043         return (error);
1044 }
1045
1046 #endif  /* i386 && (COMPAT_FREEBSD4 || COMPAT_43) */
1047
1048 #ifdef COMPAT_FREEBSD32
1049
1050 int
1051 freebsd32_shmsys(struct thread *td, struct freebsd32_shmsys_args *uap)
1052 {
1053
1054 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1055     defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1056         switch (uap->which) {
1057         case 0: {       /* shmat */
1058                 struct shmat_args ap;
1059
1060                 ap.shmid = uap->a2;
1061                 ap.shmaddr = PTRIN(uap->a3);
1062                 ap.shmflg = uap->a4;
1063                 return (sysent[SYS_shmat].sy_call(td, &ap));
1064         }
1065         case 2: {       /* shmdt */
1066                 struct shmdt_args ap;
1067
1068                 ap.shmaddr = PTRIN(uap->a2);
1069                 return (sysent[SYS_shmdt].sy_call(td, &ap));
1070         }
1071         case 3: {       /* shmget */
1072                 struct shmget_args ap;
1073
1074                 ap.key = uap->a2;
1075                 ap.size = uap->a3;
1076                 ap.shmflg = uap->a4;
1077                 return (sysent[SYS_shmget].sy_call(td, &ap));
1078         }
1079         case 4: {       /* shmctl */
1080                 struct freebsd7_freebsd32_shmctl_args ap;
1081
1082                 ap.shmid = uap->a2;
1083                 ap.cmd = uap->a3;
1084                 ap.buf = PTRIN(uap->a4);
1085                 return (freebsd7_freebsd32_shmctl(td, &ap));
1086         }
1087         case 1:         /* oshmctl */
1088         default:
1089                 return (EINVAL);
1090         }
1091 #else
1092         return (nosys(td, NULL));
1093 #endif
1094 }
1095
1096 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1097     defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1098 int
1099 freebsd7_freebsd32_shmctl(struct thread *td,
1100     struct freebsd7_freebsd32_shmctl_args *uap)
1101 {
1102         int error = 0;
1103         union {
1104                 struct shmid_ds shmid_ds;
1105                 struct shm_info shm_info;
1106                 struct shminfo shminfo;
1107         } u;
1108         union {
1109                 struct shmid_ds32_old shmid_ds32;
1110                 struct shm_info32 shm_info32;
1111                 struct shminfo32 shminfo32;
1112         } u32;
1113         size_t sz;
1114
1115         if (uap->cmd == IPC_SET) {
1116                 if ((error = copyin(uap->buf, &u32.shmid_ds32,
1117                     sizeof(u32.shmid_ds32))))
1118                         goto done;
1119                 freebsd32_ipcperm_old_in(&u32.shmid_ds32.shm_perm,
1120                     &u.shmid_ds.shm_perm);
1121                 CP(u32.shmid_ds32, u.shmid_ds, shm_segsz);
1122                 CP(u32.shmid_ds32, u.shmid_ds, shm_lpid);
1123                 CP(u32.shmid_ds32, u.shmid_ds, shm_cpid);
1124                 CP(u32.shmid_ds32, u.shmid_ds, shm_nattch);
1125                 CP(u32.shmid_ds32, u.shmid_ds, shm_atime);
1126                 CP(u32.shmid_ds32, u.shmid_ds, shm_dtime);
1127                 CP(u32.shmid_ds32, u.shmid_ds, shm_ctime);
1128         }
1129
1130         error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz);
1131         if (error)
1132                 goto done;
1133
1134         /* Cases in which we need to copyout */
1135         switch (uap->cmd) {
1136         case IPC_INFO:
1137                 CP(u.shminfo, u32.shminfo32, shmmax);
1138                 CP(u.shminfo, u32.shminfo32, shmmin);
1139                 CP(u.shminfo, u32.shminfo32, shmmni);
1140                 CP(u.shminfo, u32.shminfo32, shmseg);
1141                 CP(u.shminfo, u32.shminfo32, shmall);
1142                 error = copyout(&u32.shminfo32, uap->buf,
1143                     sizeof(u32.shminfo32));
1144                 break;
1145         case SHM_INFO:
1146                 CP(u.shm_info, u32.shm_info32, used_ids);
1147                 CP(u.shm_info, u32.shm_info32, shm_rss);
1148                 CP(u.shm_info, u32.shm_info32, shm_tot);
1149                 CP(u.shm_info, u32.shm_info32, shm_swp);
1150                 CP(u.shm_info, u32.shm_info32, swap_attempts);
1151                 CP(u.shm_info, u32.shm_info32, swap_successes);
1152                 error = copyout(&u32.shm_info32, uap->buf,
1153                     sizeof(u32.shm_info32));
1154                 break;
1155         case SHM_STAT:
1156         case IPC_STAT:
1157                 freebsd32_ipcperm_old_out(&u.shmid_ds.shm_perm,
1158                     &u32.shmid_ds32.shm_perm);
1159                 if (u.shmid_ds.shm_segsz > INT32_MAX)
1160                         u32.shmid_ds32.shm_segsz = INT32_MAX;
1161                 else
1162                         CP(u.shmid_ds, u32.shmid_ds32, shm_segsz);
1163                 CP(u.shmid_ds, u32.shmid_ds32, shm_lpid);
1164                 CP(u.shmid_ds, u32.shmid_ds32, shm_cpid);
1165                 CP(u.shmid_ds, u32.shmid_ds32, shm_nattch);
1166                 CP(u.shmid_ds, u32.shmid_ds32, shm_atime);
1167                 CP(u.shmid_ds, u32.shmid_ds32, shm_dtime);
1168                 CP(u.shmid_ds, u32.shmid_ds32, shm_ctime);
1169                 u32.shmid_ds32.shm_internal = 0;
1170                 error = copyout(&u32.shmid_ds32, uap->buf,
1171                     sizeof(u32.shmid_ds32));
1172                 break;
1173         }
1174
1175 done:
1176         if (error) {
1177                 /* Invalidate the return value */
1178                 td->td_retval[0] = -1;
1179         }
1180         return (error);
1181 }
1182 #endif
1183
1184 int
1185 freebsd32_shmctl(struct thread *td, struct freebsd32_shmctl_args *uap)
1186 {
1187         int error = 0;
1188         union {
1189                 struct shmid_ds shmid_ds;
1190                 struct shm_info shm_info;
1191                 struct shminfo shminfo;
1192         } u;
1193         union {
1194                 struct shmid_ds32 shmid_ds32;
1195                 struct shm_info32 shm_info32;
1196                 struct shminfo32 shminfo32;
1197         } u32;
1198         size_t sz;
1199
1200         if (uap->cmd == IPC_SET) {
1201                 if ((error = copyin(uap->buf, &u32.shmid_ds32,
1202                     sizeof(u32.shmid_ds32))))
1203                         goto done;
1204                 freebsd32_ipcperm_in(&u32.shmid_ds32.shm_perm,
1205                     &u.shmid_ds.shm_perm);
1206                 CP(u32.shmid_ds32, u.shmid_ds, shm_segsz);
1207                 CP(u32.shmid_ds32, u.shmid_ds, shm_lpid);
1208                 CP(u32.shmid_ds32, u.shmid_ds, shm_cpid);
1209                 CP(u32.shmid_ds32, u.shmid_ds, shm_nattch);
1210                 CP(u32.shmid_ds32, u.shmid_ds, shm_atime);
1211                 CP(u32.shmid_ds32, u.shmid_ds, shm_dtime);
1212                 CP(u32.shmid_ds32, u.shmid_ds, shm_ctime);
1213         }
1214
1215         error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz);
1216         if (error)
1217                 goto done;
1218
1219         /* Cases in which we need to copyout */
1220         switch (uap->cmd) {
1221         case IPC_INFO:
1222                 CP(u.shminfo, u32.shminfo32, shmmax);
1223                 CP(u.shminfo, u32.shminfo32, shmmin);
1224                 CP(u.shminfo, u32.shminfo32, shmmni);
1225                 CP(u.shminfo, u32.shminfo32, shmseg);
1226                 CP(u.shminfo, u32.shminfo32, shmall);
1227                 error = copyout(&u32.shminfo32, uap->buf,
1228                     sizeof(u32.shminfo32));
1229                 break;
1230         case SHM_INFO:
1231                 CP(u.shm_info, u32.shm_info32, used_ids);
1232                 CP(u.shm_info, u32.shm_info32, shm_rss);
1233                 CP(u.shm_info, u32.shm_info32, shm_tot);
1234                 CP(u.shm_info, u32.shm_info32, shm_swp);
1235                 CP(u.shm_info, u32.shm_info32, swap_attempts);
1236                 CP(u.shm_info, u32.shm_info32, swap_successes);
1237                 error = copyout(&u32.shm_info32, uap->buf,
1238                     sizeof(u32.shm_info32));
1239                 break;
1240         case SHM_STAT:
1241         case IPC_STAT:
1242                 freebsd32_ipcperm_out(&u.shmid_ds.shm_perm,
1243                     &u32.shmid_ds32.shm_perm);
1244                 if (u.shmid_ds.shm_segsz > INT32_MAX)
1245                         u32.shmid_ds32.shm_segsz = INT32_MAX;
1246                 else
1247                         CP(u.shmid_ds, u32.shmid_ds32, shm_segsz);
1248                 CP(u.shmid_ds, u32.shmid_ds32, shm_lpid);
1249                 CP(u.shmid_ds, u32.shmid_ds32, shm_cpid);
1250                 CP(u.shmid_ds, u32.shmid_ds32, shm_nattch);
1251                 CP(u.shmid_ds, u32.shmid_ds32, shm_atime);
1252                 CP(u.shmid_ds, u32.shmid_ds32, shm_dtime);
1253                 CP(u.shmid_ds, u32.shmid_ds32, shm_ctime);
1254                 error = copyout(&u32.shmid_ds32, uap->buf,
1255                     sizeof(u32.shmid_ds32));
1256                 break;
1257         }
1258
1259 done:
1260         if (error) {
1261                 /* Invalidate the return value */
1262                 td->td_retval[0] = -1;
1263         }
1264         return (error);
1265 }
1266 #endif
1267
1268 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \
1269     defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7)
1270
1271 #ifndef CP
1272 #define CP(src, dst, fld)       do { (dst).fld = (src).fld; } while (0)
1273 #endif
1274
1275 #ifndef _SYS_SYSPROTO_H_
1276 struct freebsd7_shmctl_args {
1277         int shmid;
1278         int cmd;
1279         struct shmid_ds_old *buf;
1280 };
1281 #endif
1282 int
1283 freebsd7_shmctl(td, uap)
1284         struct thread *td;
1285         struct freebsd7_shmctl_args *uap;
1286 {
1287         int error = 0;
1288         struct shmid_ds_old old;
1289         struct shmid_ds buf;
1290         size_t bufsz;
1291
1292         /*
1293          * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support
1294          * Linux binaries.  If we see the call come through the FreeBSD ABI,
1295          * return an error back to the user since we do not to support this.
1296          */
1297         if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO ||
1298             uap->cmd == SHM_STAT)
1299                 return (EINVAL);
1300
1301         /* IPC_SET needs to copyin the buffer before calling kern_shmctl */
1302         if (uap->cmd == IPC_SET) {
1303                 if ((error = copyin(uap->buf, &old, sizeof(old))))
1304                         goto done;
1305                 ipcperm_old2new(&old.shm_perm, &buf.shm_perm);
1306                 CP(old, buf, shm_segsz);
1307                 CP(old, buf, shm_lpid);
1308                 CP(old, buf, shm_cpid);
1309                 CP(old, buf, shm_nattch);
1310                 CP(old, buf, shm_atime);
1311                 CP(old, buf, shm_dtime);
1312                 CP(old, buf, shm_ctime);
1313         }
1314
1315         error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz);
1316         if (error)
1317                 goto done;
1318
1319         /* Cases in which we need to copyout */
1320         switch (uap->cmd) {
1321         case IPC_STAT:
1322                 ipcperm_new2old(&buf.shm_perm, &old.shm_perm);
1323                 if (buf.shm_segsz > INT_MAX)
1324                         old.shm_segsz = INT_MAX;
1325                 else
1326                         CP(buf, old, shm_segsz);
1327                 CP(buf, old, shm_lpid);
1328                 CP(buf, old, shm_cpid);
1329                 if (buf.shm_nattch > SHRT_MAX)
1330                         old.shm_nattch = SHRT_MAX;
1331                 else
1332                         CP(buf, old, shm_nattch);
1333                 CP(buf, old, shm_atime);
1334                 CP(buf, old, shm_dtime);
1335                 CP(buf, old, shm_ctime);
1336                 old.shm_internal = NULL;
1337                 error = copyout(&old, uap->buf, sizeof(old));
1338                 break;
1339         }
1340
1341 done:
1342         if (error) {
1343                 /* Invalidate the return value */
1344                 td->td_retval[0] = -1;
1345         }
1346         return (error);
1347 }
1348
1349 #endif  /* COMPAT_FREEBSD4 || COMPAT_FREEBSD5 || COMPAT_FREEBSD6 ||
1350            COMPAT_FREEBSD7 */
1351
1352 static int
1353 sysvshm_modload(struct module *module, int cmd, void *arg)
1354 {
1355         int error = 0;
1356
1357         switch (cmd) {
1358         case MOD_LOAD:
1359                 error = shminit();
1360                 if (error != 0)
1361                         shmunload();
1362                 break;
1363         case MOD_UNLOAD:
1364                 error = shmunload();
1365                 break;
1366         case MOD_SHUTDOWN:
1367                 break;
1368         default:
1369                 error = EINVAL;
1370                 break;
1371         }
1372         return (error);
1373 }
1374
1375 static moduledata_t sysvshm_mod = {
1376         "sysvshm",
1377         &sysvshm_modload,
1378         NULL
1379 };
1380
1381 DECLARE_MODULE(sysvshm, sysvshm_mod, SI_SUB_SYSV_SHM, SI_ORDER_FIRST);
1382 MODULE_VERSION(sysvshm, 1);