Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
FreeBSD/Linux Kernel Cross Reference
sys/kern/sysv_shm.c
Version:
- FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
SearchContext: - none - 3 - 10
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 #include "opt_mac.h" 68 69 #include <sys/param.h> 70 #include <sys/systm.h> 71 #include <sys/kernel.h> 72 #include <sys/limits.h> 73 #include <sys/lock.h> 74 #include <sys/sysctl.h> 75 #include <sys/shm.h> 76 #include <sys/proc.h> 77 #include <sys/malloc.h> 78 #include <sys/mman.h> 79 #include <sys/module.h> 80 #include <sys/mutex.h> 81 #include <sys/resourcevar.h> 82 #include <sys/stat.h> 83 #include <sys/syscall.h> 84 #include <sys/syscallsubr.h> 85 #include <sys/sysent.h> 86 #include <sys/sysproto.h> 87 #include <sys/jail.h> 88 89 #include <security/mac/mac_framework.h> 90 91 #include <vm/vm.h> 92 #include <vm/vm_param.h> 93 #include <vm/pmap.h> 94 #include <vm/vm_object.h> 95 #include <vm/vm_map.h> 96 #include <vm/vm_page.h> 97 #include <vm/vm_pager.h> 98 99 static MALLOC_DEFINE(M_SHM, "shm", "SVID compatible shared memory segments"); 100 101 static int shmget_allocate_segment(struct thread *td, 102 struct shmget_args *uap, int mode); 103 static int shmget_existing(struct thread *td, struct shmget_args *uap, 104 int mode, int segnum); 105 106 #define SHMSEG_FREE 0x0200 107 #define SHMSEG_REMOVED 0x0400 108 #define SHMSEG_ALLOCATED 0x0800 109 #define SHMSEG_WANTED 0x1000 110 111 static int shm_last_free, shm_nused, shmalloced; 112 vm_size_t shm_committed; 113 static struct shmid_kernel *shmsegs; 114 115 struct shmmap_state { 116 vm_offset_t va; 117 int shmid; 118 }; 119 120 static void shm_deallocate_segment(struct shmid_kernel *); 121 static int shm_find_segment_by_key(key_t); 122 static struct shmid_kernel *shm_find_segment_by_shmid(int); 123 static struct shmid_kernel *shm_find_segment_by_shmidx(int); 124 static int shm_delete_mapping(struct vmspace *vm, struct shmmap_state *); 125 static void shmrealloc(void); 126 static void shminit(void); 127 static int sysvshm_modload(struct module *, int, void *); 128 static int shmunload(void); 129 static void shmexit_myhook(struct vmspace *vm); 130 static void shmfork_myhook(struct proc *p1, struct proc *p2); 131 static int sysctl_shmsegs(SYSCTL_HANDLER_ARGS); 132 133 /* 134 * Tuneable values. 135 */ 136 #ifndef SHMMAXPGS 137 #define SHMMAXPGS 8192 /* Note: sysv shared memory is swap backed. */ 138 #endif 139 #ifndef SHMMAX 140 #define SHMMAX (SHMMAXPGS*PAGE_SIZE) 141 #endif 142 #ifndef SHMMIN 143 #define SHMMIN 1 144 #endif 145 #ifndef SHMMNI 146 #define SHMMNI 192 147 #endif 148 #ifndef SHMSEG 149 #define SHMSEG 128 150 #endif 151 #ifndef SHMALL 152 #define SHMALL (SHMMAXPGS) 153 #endif 154 155 struct shminfo shminfo = { 156 SHMMAX, 157 SHMMIN, 158 SHMMNI, 159 SHMSEG, 160 SHMALL 161 }; 162 163 static int shm_use_phys; 164 static int shm_allow_removed; 165 166 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmax, CTLFLAG_RW, &shminfo.shmmax, 0, 167 "Maximum shared memory segment size"); 168 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmin, CTLFLAG_RW, &shminfo.shmmin, 0, 169 "Minimum shared memory segment size"); 170 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmni, CTLFLAG_RDTUN, &shminfo.shmmni, 0, 171 "Number of shared memory identifiers"); 172 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmseg, CTLFLAG_RDTUN, &shminfo.shmseg, 0, 173 "Number of segments per process"); 174 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmall, CTLFLAG_RW, &shminfo.shmall, 0, 175 "Maximum number of pages available for shared memory"); 176 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_use_phys, CTLFLAG_RW, 177 &shm_use_phys, 0, "Enable/Disable locking of shared memory pages in core"); 178 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_allow_removed, CTLFLAG_RW, 179 &shm_allow_removed, 0, 180 "Enable/Disable attachment to attached segments marked for removal"); 181 SYSCTL_PROC(_kern_ipc, OID_AUTO, shmsegs, CTLFLAG_RD, 182 NULL, 0, sysctl_shmsegs, "", 183 "Current number of shared memory segments allocated"); 184 185 static int 186 shm_find_segment_by_key(key) 187 key_t key; 188 { 189 int i; 190 191 for (i = 0; i < shmalloced; i++) 192 if ((shmsegs[i].u.shm_perm.mode & SHMSEG_ALLOCATED) && 193 shmsegs[i].u.shm_perm.key == key) 194 return (i); 195 return (-1); 196 } 197 198 static struct shmid_kernel * 199 shm_find_segment_by_shmid(int shmid) 200 { 201 int segnum; 202 struct shmid_kernel *shmseg; 203 204 segnum = IPCID_TO_IX(shmid); 205 if (segnum < 0 || segnum >= shmalloced) 206 return (NULL); 207 shmseg = &shmsegs[segnum]; 208 if ((shmseg->u.shm_perm.mode & SHMSEG_ALLOCATED) == 0 || 209 (!shm_allow_removed && 210 (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) != 0) || 211 shmseg->u.shm_perm.seq != IPCID_TO_SEQ(shmid)) 212 return (NULL); 213 return (shmseg); 214 } 215 216 static struct shmid_kernel * 217 shm_find_segment_by_shmidx(int segnum) 218 { 219 struct shmid_kernel *shmseg; 220 221 if (segnum < 0 || segnum >= shmalloced) 222 return (NULL); 223 shmseg = &shmsegs[segnum]; 224 if ((shmseg->u.shm_perm.mode & SHMSEG_ALLOCATED) == 0 || 225 (!shm_allow_removed && 226 (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) != 0)) 227 return (NULL); 228 return (shmseg); 229 } 230 231 static void 232 shm_deallocate_segment(shmseg) 233 struct shmid_kernel *shmseg; 234 { 235 vm_size_t size; 236 237 GIANT_REQUIRED; 238 239 vm_object_deallocate(shmseg->object); 240 shmseg->object = NULL; 241 size = round_page(shmseg->u.shm_segsz); 242 shm_committed -= btoc(size); 243 shm_nused--; 244 shmseg->u.shm_perm.mode = SHMSEG_FREE; 245 #ifdef MAC 246 mac_cleanup_sysv_shm(shmseg); 247 #endif 248 } 249 250 static int 251 shm_delete_mapping(struct vmspace *vm, struct shmmap_state *shmmap_s) 252 { 253 struct shmid_kernel *shmseg; 254 int segnum, result; 255 vm_size_t size; 256 257 GIANT_REQUIRED; 258 259 segnum = IPCID_TO_IX(shmmap_s->shmid); 260 shmseg = &shmsegs[segnum]; 261 size = round_page(shmseg->u.shm_segsz); 262 result = vm_map_remove(&vm->vm_map, shmmap_s->va, shmmap_s->va + size); 263 if (result != KERN_SUCCESS) 264 return (EINVAL); 265 shmmap_s->shmid = -1; 266 shmseg->u.shm_dtime = time_second; 267 if ((--shmseg->u.shm_nattch <= 0) && 268 (shmseg->u.shm_perm.mode & SHMSEG_REMOVED)) { 269 shm_deallocate_segment(shmseg); 270 shm_last_free = segnum; 271 } 272 return (0); 273 } 274 275 #ifndef _SYS_SYSPROTO_H_ 276 struct shmdt_args { 277 const void *shmaddr; 278 }; 279 #endif 280 int 281 shmdt(td, uap) 282 struct thread *td; 283 struct shmdt_args *uap; 284 { 285 struct proc *p = td->td_proc; 286 struct shmmap_state *shmmap_s; 287 #ifdef MAC 288 struct shmid_kernel *shmsegptr; 289 #endif 290 int i; 291 int error = 0; 292 293 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 294 return (ENOSYS); 295 mtx_lock(&Giant); 296 shmmap_s = p->p_vmspace->vm_shm; 297 if (shmmap_s == NULL) { 298 error = EINVAL; 299 goto done2; 300 } 301 for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) { 302 if (shmmap_s->shmid != -1 && 303 shmmap_s->va == (vm_offset_t)uap->shmaddr) { 304 break; 305 } 306 } 307 if (i == shminfo.shmseg) { 308 error = EINVAL; 309 goto done2; 310 } 311 #ifdef MAC 312 shmsegptr = &shmsegs[IPCID_TO_IX(shmmap_s->shmid)]; 313 error = mac_check_sysv_shmdt(td->td_ucred, shmsegptr); 314 if (error != 0) 315 goto done2; 316 #endif 317 error = shm_delete_mapping(p->p_vmspace, shmmap_s); 318 done2: 319 mtx_unlock(&Giant); 320 return (error); 321 } 322 323 #ifndef _SYS_SYSPROTO_H_ 324 struct shmat_args { 325 int shmid; 326 const void *shmaddr; 327 int shmflg; 328 }; 329 #endif 330 int 331 kern_shmat(td, shmid, shmaddr, shmflg) 332 struct thread *td; 333 int shmid; 334 const void *shmaddr; 335 int shmflg; 336 { 337 struct proc *p = td->td_proc; 338 int i, flags; 339 struct shmid_kernel *shmseg; 340 struct shmmap_state *shmmap_s = NULL; 341 vm_offset_t attach_va; 342 vm_prot_t prot; 343 vm_size_t size; 344 int rv; 345 int error = 0; 346 347 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 348 return (ENOSYS); 349 mtx_lock(&Giant); 350 shmmap_s = p->p_vmspace->vm_shm; 351 if (shmmap_s == NULL) { 352 shmmap_s = malloc(shminfo.shmseg * sizeof(struct shmmap_state), 353 M_SHM, M_WAITOK); 354 for (i = 0; i < shminfo.shmseg; i++) 355 shmmap_s[i].shmid = -1; 356 p->p_vmspace->vm_shm = shmmap_s; 357 } 358 shmseg = shm_find_segment_by_shmid(shmid); 359 if (shmseg == NULL) { 360 error = EINVAL; 361 goto done2; 362 } 363 error = ipcperm(td, &shmseg->u.shm_perm, 364 (shmflg & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W); 365 if (error) 366 goto done2; 367 #ifdef MAC 368 error = mac_check_sysv_shmat(td->td_ucred, shmseg, shmflg); 369 if (error != 0) 370 goto done2; 371 #endif 372 for (i = 0; i < shminfo.shmseg; i++) { 373 if (shmmap_s->shmid == -1) 374 break; 375 shmmap_s++; 376 } 377 if (i >= shminfo.shmseg) { 378 error = EMFILE; 379 goto done2; 380 } 381 size = round_page(shmseg->u.shm_segsz); 382 #ifdef VM_PROT_READ_IS_EXEC 383 prot = VM_PROT_READ | VM_PROT_EXECUTE; 384 #else 385 prot = VM_PROT_READ; 386 #endif 387 if ((shmflg & SHM_RDONLY) == 0) 388 prot |= VM_PROT_WRITE; 389 flags = MAP_ANON | MAP_SHARED; 390 if (shmaddr) { 391 flags |= MAP_FIXED; 392 if (shmflg & SHM_RND) { 393 attach_va = (vm_offset_t)shmaddr & ~(SHMLBA-1); 394 } else if (((vm_offset_t)shmaddr & (SHMLBA-1)) == 0) { 395 attach_va = (vm_offset_t)shmaddr; 396 } else { 397 error = EINVAL; 398 goto done2; 399 } 400 } else { 401 /* 402 * This is just a hint to vm_map_find() about where to 403 * put it. 404 */ 405 PROC_LOCK(p); 406 attach_va = round_page((vm_offset_t)p->p_vmspace->vm_daddr + 407 lim_max(p, RLIMIT_DATA)); 408 PROC_UNLOCK(p); 409 } 410 411 vm_object_reference(shmseg->object); 412 rv = vm_map_find(&p->p_vmspace->vm_map, shmseg->object, 413 0, &attach_va, size, (flags & MAP_FIXED) ? VMFS_NO_SPACE : 414 VMFS_ANY_SPACE, prot, prot, 0); 415 if (rv != KERN_SUCCESS) { 416 vm_object_deallocate(shmseg->object); 417 error = ENOMEM; 418 goto done2; 419 } 420 vm_map_inherit(&p->p_vmspace->vm_map, 421 attach_va, attach_va + size, VM_INHERIT_SHARE); 422 423 shmmap_s->va = attach_va; 424 shmmap_s->shmid = shmid; 425 shmseg->u.shm_lpid = p->p_pid; 426 shmseg->u.shm_atime = time_second; 427 shmseg->u.shm_nattch++; 428 td->td_retval[0] = attach_va; 429 done2: 430 mtx_unlock(&Giant); 431 return (error); 432 } 433 434 int 435 shmat(td, uap) 436 struct thread *td; 437 struct shmat_args *uap; 438 { 439 return kern_shmat(td, uap->shmid, uap->shmaddr, uap->shmflg); 440 } 441 442 int 443 kern_shmctl(td, shmid, cmd, buf, bufsz) 444 struct thread *td; 445 int shmid; 446 int cmd; 447 void *buf; 448 size_t *bufsz; 449 { 450 int error = 0; 451 struct shmid_kernel *shmseg; 452 453 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 454 return (ENOSYS); 455 456 mtx_lock(&Giant); 457 switch (cmd) { 458 case IPC_INFO: 459 memcpy(buf, &shminfo, sizeof(shminfo)); 460 if (bufsz) 461 *bufsz = sizeof(shminfo); 462 td->td_retval[0] = shmalloced; 463 goto done2; 464 case SHM_INFO: { 465 struct shm_info shm_info; 466 shm_info.used_ids = shm_nused; 467 shm_info.shm_rss = 0; /*XXX where to get from ? */ 468 shm_info.shm_tot = 0; /*XXX where to get from ? */ 469 shm_info.shm_swp = 0; /*XXX where to get from ? */ 470 shm_info.swap_attempts = 0; /*XXX where to get from ? */ 471 shm_info.swap_successes = 0; /*XXX where to get from ? */ 472 memcpy(buf, &shm_info, sizeof(shm_info)); 473 if (bufsz) 474 *bufsz = sizeof(shm_info); 475 td->td_retval[0] = shmalloced; 476 goto done2; 477 } 478 } 479 if (cmd == SHM_STAT) 480 shmseg = shm_find_segment_by_shmidx(shmid); 481 else 482 shmseg = shm_find_segment_by_shmid(shmid); 483 if (shmseg == NULL) { 484 error = EINVAL; 485 goto done2; 486 } 487 #ifdef MAC 488 error = mac_check_sysv_shmctl(td->td_ucred, shmseg, cmd); 489 if (error != 0) 490 goto done2; 491 #endif 492 switch (cmd) { 493 case SHM_STAT: 494 case IPC_STAT: 495 error = ipcperm(td, &shmseg->u.shm_perm, IPC_R); 496 if (error) 497 goto done2; 498 memcpy(buf, &shmseg->u, sizeof(struct shmid_ds)); 499 if (bufsz) 500 *bufsz = sizeof(struct shmid_ds); 501 if (cmd == SHM_STAT) 502 td->td_retval[0] = IXSEQ_TO_IPCID(shmid, shmseg->u.shm_perm); 503 break; 504 case IPC_SET: { 505 struct shmid_ds *shmid; 506 507 shmid = (struct shmid_ds *)buf; 508 error = ipcperm(td, &shmseg->u.shm_perm, IPC_M); 509 if (error) 510 goto done2; 511 shmseg->u.shm_perm.uid = shmid->shm_perm.uid; 512 shmseg->u.shm_perm.gid = shmid->shm_perm.gid; 513 shmseg->u.shm_perm.mode = 514 (shmseg->u.shm_perm.mode & ~ACCESSPERMS) | 515 (shmid->shm_perm.mode & ACCESSPERMS); 516 shmseg->u.shm_ctime = time_second; 517 break; 518 } 519 case IPC_RMID: 520 error = ipcperm(td, &shmseg->u.shm_perm, IPC_M); 521 if (error) 522 goto done2; 523 shmseg->u.shm_perm.key = IPC_PRIVATE; 524 shmseg->u.shm_perm.mode |= SHMSEG_REMOVED; 525 if (shmseg->u.shm_nattch <= 0) { 526 shm_deallocate_segment(shmseg); 527 shm_last_free = IPCID_TO_IX(shmid); 528 } 529 break; 530 #if 0 531 case SHM_LOCK: 532 case SHM_UNLOCK: 533 #endif 534 default: 535 error = EINVAL; 536 break; 537 } 538 done2: 539 mtx_unlock(&Giant); 540 return (error); 541 } 542 543 #ifndef _SYS_SYSPROTO_H_ 544 struct shmctl_args { 545 int shmid; 546 int cmd; 547 struct shmid_ds *buf; 548 }; 549 #endif 550 int 551 shmctl(td, uap) 552 struct thread *td; 553 struct shmctl_args *uap; 554 { 555 int error = 0; 556 struct shmid_ds buf; 557 size_t bufsz; 558 559 /* IPC_SET needs to copyin the buffer before calling kern_shmctl */ 560 if (uap->cmd == IPC_SET) { 561 if ((error = copyin(uap->buf, &buf, sizeof(struct shmid_ds)))) 562 goto done; 563 } 564 565 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz); 566 if (error) 567 goto done; 568 569 /* Cases in which we need to copyout */ 570 switch (uap->cmd) { 571 case IPC_INFO: 572 case SHM_INFO: 573 case SHM_STAT: 574 case IPC_STAT: 575 error = copyout(&buf, uap->buf, bufsz); 576 break; 577 } 578 579 done: 580 if (error) { 581 /* Invalidate the return value */ 582 td->td_retval[0] = -1; 583 } 584 return (error); 585 } 586 587 588 static int 589 shmget_existing(td, uap, mode, segnum) 590 struct thread *td; 591 struct shmget_args *uap; 592 int mode; 593 int segnum; 594 { 595 struct shmid_kernel *shmseg; 596 int error; 597 598 shmseg = &shmsegs[segnum]; 599 if (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) { 600 /* 601 * This segment is in the process of being allocated. Wait 602 * until it's done, and look the key up again (in case the 603 * allocation failed or it was freed). 604 */ 605 shmseg->u.shm_perm.mode |= SHMSEG_WANTED; 606 error = tsleep(shmseg, PLOCK | PCATCH, "shmget", 0); 607 if (error) 608 return (error); 609 return (EAGAIN); 610 } 611 if ((uap->shmflg & (IPC_CREAT | IPC_EXCL)) == (IPC_CREAT | IPC_EXCL)) 612 return (EEXIST); 613 #ifdef MAC 614 error = mac_check_sysv_shmget(td->td_ucred, shmseg, uap->shmflg); 615 if (error != 0) 616 return (error); 617 #endif 618 if (uap->size != 0 && uap->size > shmseg->u.shm_segsz) 619 return (EINVAL); 620 td->td_retval[0] = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm); 621 return (0); 622 } 623 624 static int 625 shmget_allocate_segment(td, uap, mode) 626 struct thread *td; 627 struct shmget_args *uap; 628 int mode; 629 { 630 int i, segnum, shmid; 631 size_t size; 632 struct ucred *cred = td->td_ucred; 633 struct shmid_kernel *shmseg; 634 vm_object_t shm_object; 635 636 GIANT_REQUIRED; 637 638 if (uap->size < shminfo.shmmin || uap->size > shminfo.shmmax) 639 return (EINVAL); 640 if (shm_nused >= shminfo.shmmni) /* Any shmids left? */ 641 return (ENOSPC); 642 size = round_page(uap->size); 643 if (shm_committed + btoc(size) > shminfo.shmall) 644 return (ENOMEM); 645 if (shm_last_free < 0) { 646 shmrealloc(); /* Maybe expand the shmsegs[] array. */ 647 for (i = 0; i < shmalloced; i++) 648 if (shmsegs[i].u.shm_perm.mode & SHMSEG_FREE) 649 break; 650 if (i == shmalloced) 651 return (ENOSPC); 652 segnum = i; 653 } else { 654 segnum = shm_last_free; 655 shm_last_free = -1; 656 } 657 shmseg = &shmsegs[segnum]; 658 /* 659 * In case we sleep in malloc(), mark the segment present but deleted 660 * so that noone else tries to create the same key. 661 */ 662 shmseg->u.shm_perm.mode = SHMSEG_ALLOCATED | SHMSEG_REMOVED; 663 shmseg->u.shm_perm.key = uap->key; 664 shmseg->u.shm_perm.seq = (shmseg->u.shm_perm.seq + 1) & 0x7fff; 665 shmid = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm); 666 667 /* 668 * We make sure that we have allocated a pager before we need 669 * to. 670 */ 671 if (shm_use_phys) { 672 shm_object = 673 vm_pager_allocate(OBJT_PHYS, 0, size, VM_PROT_DEFAULT, 0); 674 } else { 675 shm_object = 676 vm_pager_allocate(OBJT_SWAP, 0, size, VM_PROT_DEFAULT, 0); 677 } 678 VM_OBJECT_LOCK(shm_object); 679 vm_object_clear_flag(shm_object, OBJ_ONEMAPPING); 680 vm_object_set_flag(shm_object, OBJ_NOSPLIT); 681 VM_OBJECT_UNLOCK(shm_object); 682 683 shmseg->object = shm_object; 684 shmseg->u.shm_perm.cuid = shmseg->u.shm_perm.uid = cred->cr_uid; 685 shmseg->u.shm_perm.cgid = shmseg->u.shm_perm.gid = cred->cr_gid; 686 shmseg->u.shm_perm.mode = (shmseg->u.shm_perm.mode & SHMSEG_WANTED) | 687 (mode & ACCESSPERMS) | SHMSEG_ALLOCATED; 688 shmseg->u.shm_segsz = uap->size; 689 shmseg->u.shm_cpid = td->td_proc->p_pid; 690 shmseg->u.shm_lpid = shmseg->u.shm_nattch = 0; 691 shmseg->u.shm_atime = shmseg->u.shm_dtime = 0; 692 #ifdef MAC 693 mac_create_sysv_shm(cred, shmseg); 694 #endif 695 shmseg->u.shm_ctime = time_second; 696 shm_committed += btoc(size); 697 shm_nused++; 698 if (shmseg->u.shm_perm.mode & SHMSEG_WANTED) { 699 /* 700 * Somebody else wanted this key while we were asleep. Wake 701 * them up now. 702 */ 703 shmseg->u.shm_perm.mode &= ~SHMSEG_WANTED; 704 wakeup(shmseg); 705 } 706 td->td_retval[0] = shmid; 707 return (0); 708 } 709 710 #ifndef _SYS_SYSPROTO_H_ 711 struct shmget_args { 712 key_t key; 713 size_t size; 714 int shmflg; 715 }; 716 #endif 717 int 718 shmget(td, uap) 719 struct thread *td; 720 struct shmget_args *uap; 721 { 722 int segnum, mode; 723 int error; 724 725 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 726 return (ENOSYS); 727 mtx_lock(&Giant); 728 mode = uap->shmflg & ACCESSPERMS; 729 if (uap->key != IPC_PRIVATE) { 730 again: 731 segnum = shm_find_segment_by_key(uap->key); 732 if (segnum >= 0) { 733 error = shmget_existing(td, uap, mode, segnum); 734 if (error == EAGAIN) 735 goto again; 736 goto done2; 737 } 738 if ((uap->shmflg & IPC_CREAT) == 0) { 739 error = ENOENT; 740 goto done2; 741 } 742 } 743 error = shmget_allocate_segment(td, uap, mode); 744 done2: 745 mtx_unlock(&Giant); 746 return (error); 747 } 748 749 static void 750 shmfork_myhook(p1, p2) 751 struct proc *p1, *p2; 752 { 753 struct shmmap_state *shmmap_s; 754 size_t size; 755 int i; 756 757 mtx_lock(&Giant); 758 size = shminfo.shmseg * sizeof(struct shmmap_state); 759 shmmap_s = malloc(size, M_SHM, M_WAITOK); 760 bcopy(p1->p_vmspace->vm_shm, shmmap_s, size); 761 p2->p_vmspace->vm_shm = shmmap_s; 762 for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) 763 if (shmmap_s->shmid != -1) 764 shmsegs[IPCID_TO_IX(shmmap_s->shmid)].u.shm_nattch++; 765 mtx_unlock(&Giant); 766 } 767 768 static void 769 shmexit_myhook(struct vmspace *vm) 770 { 771 struct shmmap_state *base, *shm; 772 int i; 773 774 if ((base = vm->vm_shm) != NULL) { 775 vm->vm_shm = NULL; 776 mtx_lock(&Giant); 777 for (i = 0, shm = base; i < shminfo.shmseg; i++, shm++) { 778 if (shm->shmid != -1) 779 shm_delete_mapping(vm, shm); 780 } 781 mtx_unlock(&Giant); 782 free(base, M_SHM); 783 } 784 } 785 786 static void 787 shmrealloc(void) 788 { 789 int i; 790 struct shmid_kernel *newsegs; 791 792 if (shmalloced >= shminfo.shmmni) 793 return; 794 795 newsegs = malloc(shminfo.shmmni * sizeof(*newsegs), M_SHM, M_WAITOK); 796 if (newsegs == NULL) 797 return; 798 for (i = 0; i < shmalloced; i++) 799 bcopy(&shmsegs[i], &newsegs[i], sizeof(newsegs[0])); 800 for (; i < shminfo.shmmni; i++) { 801 shmsegs[i].u.shm_perm.mode = SHMSEG_FREE; 802 shmsegs[i].u.shm_perm.seq = 0; 803 #ifdef MAC 804 mac_init_sysv_shm(&shmsegs[i]); 805 #endif 806 } 807 free(shmsegs, M_SHM); 808 shmsegs = newsegs; 809 shmalloced = shminfo.shmmni; 810 } 811 812 static void 813 shminit() 814 { 815 int i; 816 817 TUNABLE_ULONG_FETCH("kern.ipc.shmmaxpgs", &shminfo.shmall); 818 for (i = PAGE_SIZE; i > 0; i--) { 819 shminfo.shmmax = shminfo.shmall * i; 820 if (shminfo.shmmax >= shminfo.shmall) 821 break; 822 } 823 TUNABLE_ULONG_FETCH("kern.ipc.shmmin", &shminfo.shmmin); 824 TUNABLE_ULONG_FETCH("kern.ipc.shmmni", &shminfo.shmmni); 825 TUNABLE_ULONG_FETCH("kern.ipc.shmseg", &shminfo.shmseg); 826 TUNABLE_INT_FETCH("kern.ipc.shm_use_phys", &shm_use_phys); 827 828 shmalloced = shminfo.shmmni; 829 shmsegs = malloc(shmalloced * sizeof(shmsegs[0]), M_SHM, M_WAITOK); 830 if (shmsegs == NULL) 831 panic("cannot allocate initial memory for sysvshm"); 832 for (i = 0; i < shmalloced; i++) { 833 shmsegs[i].u.shm_perm.mode = SHMSEG_FREE; 834 shmsegs[i].u.shm_perm.seq = 0; 835 #ifdef MAC 836 mac_init_sysv_shm(&shmsegs[i]); 837 #endif 838 } 839 shm_last_free = 0; 840 shm_nused = 0; 841 shm_committed = 0; 842 shmexit_hook = &shmexit_myhook; 843 shmfork_hook = &shmfork_myhook; 844 } 845 846 static int 847 shmunload() 848 { 849 #ifdef MAC 850 int i; 851 #endif 852 853 if (shm_nused > 0) 854 return (EBUSY); 855 856 #ifdef MAC 857 for (i = 0; i < shmalloced; i++) 858 mac_destroy_sysv_shm(&shmsegs[i]); 859 #endif 860 free(shmsegs, M_SHM); 861 shmexit_hook = NULL; 862 shmfork_hook = NULL; 863 return (0); 864 } 865 866 static int 867 sysctl_shmsegs(SYSCTL_HANDLER_ARGS) 868 { 869 870 return (SYSCTL_OUT(req, shmsegs, shmalloced * sizeof(shmsegs[0]))); 871 } 872 873 #if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43)) 874 struct oshmid_ds { 875 struct ipc_perm_old shm_perm; /* operation perms */ 876 int shm_segsz; /* size of segment (bytes) */ 877 u_short shm_cpid; /* pid, creator */ 878 u_short shm_lpid; /* pid, last operation */ 879 short shm_nattch; /* no. of current attaches */ 880 time_t shm_atime; /* last attach time */ 881 time_t shm_dtime; /* last detach time */ 882 time_t shm_ctime; /* last change time */ 883 void *shm_handle; /* internal handle for shm segment */ 884 }; 885 886 struct oshmctl_args { 887 int shmid; 888 int cmd; 889 struct oshmid_ds *ubuf; 890 }; 891 892 static int 893 oshmctl(struct thread *td, struct oshmctl_args *uap) 894 { 895 #ifdef COMPAT_43 896 int error = 0; 897 struct shmid_kernel *shmseg; 898 struct oshmid_ds outbuf; 899 900 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 901 return (ENOSYS); 902 mtx_lock(&Giant); 903 shmseg = shm_find_segment_by_shmid(uap->shmid); 904 if (shmseg == NULL) { 905 error = EINVAL; 906 goto done2; 907 } 908 switch (uap->cmd) { 909 case IPC_STAT: 910 error = ipcperm(td, &shmseg->u.shm_perm, IPC_R); 911 if (error) 912 goto done2; 913 #ifdef MAC 914 error = mac_check_sysv_shmctl(td->td_ucred, shmseg, uap->cmd); 915 if (error != 0) 916 goto done2; 917 #endif 918 ipcperm_new2old(&shmseg->u.shm_perm, &outbuf.shm_perm); 919 outbuf.shm_segsz = shmseg->u.shm_segsz; 920 outbuf.shm_cpid = shmseg->u.shm_cpid; 921 outbuf.shm_lpid = shmseg->u.shm_lpid; 922 outbuf.shm_nattch = shmseg->u.shm_nattch; 923 outbuf.shm_atime = shmseg->u.shm_atime; 924 outbuf.shm_dtime = shmseg->u.shm_dtime; 925 outbuf.shm_ctime = shmseg->u.shm_ctime; 926 outbuf.shm_handle = shmseg->object; 927 error = copyout(&outbuf, uap->ubuf, sizeof(outbuf)); 928 if (error) 929 goto done2; 930 break; 931 default: 932 error = freebsd7_shmctl(td, (struct freebsd7_shmctl_args *)uap); 933 break; 934 } 935 done2: 936 mtx_unlock(&Giant); 937 return (error); 938 #else 939 return (EINVAL); 940 #endif 941 } 942 943 /* XXX casting to (sy_call_t *) is bogus, as usual. */ 944 static sy_call_t *shmcalls[] = { 945 (sy_call_t *)shmat, (sy_call_t *)oshmctl, 946 (sy_call_t *)shmdt, (sy_call_t *)shmget, 947 (sy_call_t *)freebsd7_shmctl 948 }; 949 950 int 951 shmsys(td, uap) 952 struct thread *td; 953 /* XXX actually varargs. */ 954 struct shmsys_args /* { 955 int which; 956 int a2; 957 int a3; 958 int a4; 959 } */ *uap; 960 { 961 int error; 962 963 if (!jail_sysvipc_allowed && jailed(td->td_ucred)) 964 return (ENOSYS); 965 if (uap->which < 0 || 966 uap->which >= sizeof(shmcalls)/sizeof(shmcalls[0])) 967 return (EINVAL); 968 mtx_lock(&Giant); 969 error = (*shmcalls[uap->which])(td, &uap->a2); 970 mtx_unlock(&Giant); 971 return (error); 972 } 973 974 SYSCALL_MODULE_HELPER(shmsys); 975 #endif /* i386 && (COMPAT_FREEBSD4 || COMPAT_43) */ 976 977 #define CP(src, dst, fld) do { (dst).fld = (src).fld; } while (0) 978 979 980 #ifndef _SYS_SYSPROTO_H_ 981 struct freebsd7_shmctl_args { 982 int shmid; 983 int cmd; 984 struct shmid_ds_old *buf; 985 }; 986 #endif 987 int 988 freebsd7_shmctl(td, uap) 989 struct thread *td; 990 struct freebsd7_shmctl_args *uap; 991 { 992 int error = 0; 993 struct shmid_ds_old old; 994 struct shmid_ds buf; 995 size_t bufsz; 996 997 /* 998 * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support 999 * Linux binaries. If we see the call come through the FreeBSD ABI, 1000 * return an error back to the user since we do not to support this. 1001 */ 1002 if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO || 1003 uap->cmd == SHM_STAT) 1004 return (EINVAL); 1005 1006 /* IPC_SET needs to copyin the buffer before calling kern_shmctl */ 1007 if (uap->cmd == IPC_SET) { 1008 if ((error = copyin(uap->buf, &old, sizeof(old)))) 1009 goto done; 1010 ipcperm_old2new(&old.shm_perm, &buf.shm_perm); 1011 CP(old, buf, shm_segsz); 1012 CP(old, buf, shm_lpid); 1013 CP(old, buf, shm_cpid); 1014 CP(old, buf, shm_nattch); 1015 CP(old, buf, shm_atime); 1016 CP(old, buf, shm_dtime); 1017 CP(old, buf, shm_ctime); 1018 } 1019 1020 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz); 1021 if (error) 1022 goto done; 1023 1024 /* Cases in which we need to copyout */ 1025 switch (uap->cmd) { 1026 case IPC_STAT: 1027 ipcperm_new2old(&buf.shm_perm, &old.shm_perm); 1028 if (buf.shm_segsz > INT_MAX) 1029 old.shm_segsz = INT_MAX; 1030 else 1031 CP(buf, old, shm_segsz); 1032 CP(buf, old, shm_lpid); 1033 CP(buf, old, shm_cpid); 1034 if (buf.shm_nattch > SHRT_MAX) 1035 old.shm_nattch = SHRT_MAX; 1036 else 1037 CP(buf, old, shm_nattch); 1038 CP(buf, old, shm_atime); 1039 CP(buf, old, shm_dtime); 1040 CP(buf, old, shm_ctime); 1041 old.shm_internal = NULL; 1042 error = copyout(&old, uap->buf, sizeof(old)); 1043 break; 1044 } 1045 1046 done: 1047 if (error) { 1048 /* Invalidate the return value */ 1049 td->td_retval[0] = -1; 1050 } 1051 return (error); 1052 } 1053 1054 SYSCALL_MODULE_HELPER(freebsd7_shmctl); 1055 1056 /* ABI compat shim for older kernel modules. */ 1057 #undef kern_shmctl 1058 1059 int kern_shmctl(struct thread *td, int shmid, int cmd, void *bufp, 1060 size_t *bufszp); 1061 1062 int 1063 kern_shmctl(td, shmid, cmd, bufp, bufszp) 1064 struct thread *td; 1065 int shmid; 1066 int cmd; 1067 void *bufp; 1068 size_t *bufszp; 1069 { 1070 struct shmid_ds_old *old; 1071 struct shmid_ds shm; 1072 void *buf; 1073 size_t *bufsz; 1074 int error; 1075 1076 switch (cmd) { 1077 case IPC_SET: 1078 old = bufp; 1079 ipcperm_old2new(&old->shm_perm, &shm.shm_perm); 1080 CP(*old, shm, shm_segsz); 1081 CP(*old, shm, shm_lpid); 1082 CP(*old, shm, shm_cpid); 1083 CP(*old, shm, shm_nattch); 1084 CP(*old, shm, shm_atime); 1085 CP(*old, shm, shm_dtime); 1086 CP(*old, shm, shm_ctime); 1087 /* FALLTHROUGH */ 1088 case IPC_STAT: 1089 buf = &shm; 1090 bufsz = NULL; 1091 break; 1092 default: 1093 buf = bufp; 1094 bufsz = bufszp; 1095 break; 1096 } 1097 1098 error = kern_new_shmctl(td, shmid, cmd, buf, bufsz); 1099 if (error) 1100 return (error); 1101 1102 switch (cmd) { 1103 case IPC_STAT: 1104 old = bufp; 1105 if (bufszp) 1106 *bufszp = sizeof(*old); 1107 ipcperm_new2old(&shm.shm_perm, &old->shm_perm); 1108 if (shm.shm_segsz > INT_MAX) 1109 old->shm_segsz = INT_MAX; 1110 else 1111 CP(shm, *old, shm_segsz); 1112 CP(shm, *old, shm_lpid); 1113 CP(shm, *old, shm_cpid); 1114 if (shm.shm_nattch > SHRT_MAX) 1115 old->shm_nattch = SHRT_MAX; 1116 else 1117 CP(shm, *old, shm_nattch); 1118 CP(shm, *old, shm_atime); 1119 CP(shm, *old, shm_dtime); 1120 CP(shm, *old, shm_ctime); 1121 old->shm_internal = NULL; 1122 break; 1123 } 1124 1125 return (error); 1126 } 1127 1128 #undef CP 1129 1130 static int 1131 sysvshm_modload(struct module *module, int cmd, void *arg) 1132 { 1133 int error = 0; 1134 1135 switch (cmd) { 1136 case MOD_LOAD: 1137 shminit(); 1138 break; 1139 case MOD_UNLOAD: 1140 error = shmunload(); 1141 break; 1142 case MOD_SHUTDOWN: 1143 break; 1144 default: 1145 error = EINVAL; 1146 break; 1147 } 1148 return (error); 1149 } 1150 1151 static moduledata_t sysvshm_mod = { 1152 "sysvshm", 1153 &sysvshm_modload, 1154 NULL 1155 }; 1156 1157 SYSCALL_MODULE_HELPER(shmat); 1158 SYSCALL_MODULE_HELPER(shmctl); 1159 SYSCALL_MODULE_HELPER(shmdt); 1160 SYSCALL_MODULE_HELPER(shmget); 1161 1162 DECLARE_MODULE(sysvshm, sysvshm_mod, SI_SUB_SYSV_SHM, SI_ORDER_FIRST); 1163 MODULE_VERSION(sysvshm, 1);
Cache object: c834b30b69bae9ed790251d873095644
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]