1 /*- 2 * SPDX-License-Identifier: BSD-4-Clause AND BSD-2-Clause 3 * 4 * Copyright (c) 1994 Adam Glass and Charles Hannum. All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. All advertising materials mentioning features or use of this software 15 * must display the following acknowledgement: 16 * This product includes software developed by Adam Glass and Charles 17 * Hannum. 18 * 4. The names of the authors may not be used to endorse or promote products 19 * derived from this software without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR 22 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 23 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 24 * IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT, INDIRECT, 25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 26 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 27 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 28 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 29 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 30 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 31 * 32 * $NetBSD: sysv_shm.c,v 1.39 1997/10/07 10:02:03 drochner Exp $ 33 */ 34 /*- 35 * Copyright (c) 2003-2005 McAfee, Inc. 36 * Copyright (c) 2016-2017 Robert N. M. Watson 37 * All rights reserved. 38 * 39 * This software was developed for the FreeBSD Project in part by McAfee 40 * Research, the Security Research Division of McAfee, Inc under DARPA/SPAWAR 41 * contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS research 42 * program. 43 * 44 * Portions of this software were developed by BAE Systems, the University of 45 * Cambridge Computer Laboratory, and Memorial University under DARPA/AFRL 46 * contract FA8650-15-C-7558 ("CADETS"), as part of the DARPA Transparent 47 * Computing (TC) research program. 48 * 49 * Redistribution and use in source and binary forms, with or without 50 * modification, are permitted provided that the following conditions 51 * are met: 52 * 1. Redistributions of source code must retain the above copyright 53 * notice, this list of conditions and the following disclaimer. 54 * 2. Redistributions in binary form must reproduce the above copyright 55 * notice, this list of conditions and the following disclaimer in the 56 * documentation and/or other materials provided with the distribution. 57 * 58 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 59 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 60 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 61 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 62 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 63 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 64 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 65 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 66 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 67 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 68 * SUCH DAMAGE. 69 */ 70 71 #include <sys/cdefs.h> 72 #include "opt_sysvipc.h" 73 74 #include <sys/param.h> 75 #include <sys/systm.h> 76 #include <sys/abi_compat.h> 77 #include <sys/kernel.h> 78 #include <sys/limits.h> 79 #include <sys/lock.h> 80 #include <sys/sysctl.h> 81 #include <sys/shm.h> 82 #include <sys/proc.h> 83 #include <sys/malloc.h> 84 #include <sys/mman.h> 85 #include <sys/module.h> 86 #include <sys/mutex.h> 87 #include <sys/racct.h> 88 #include <sys/resourcevar.h> 89 #include <sys/rwlock.h> 90 #include <sys/stat.h> 91 #include <sys/syscall.h> 92 #include <sys/syscallsubr.h> 93 #include <sys/sysent.h> 94 #include <sys/sysproto.h> 95 #include <sys/jail.h> 96 97 #include <security/audit/audit.h> 98 #include <security/mac/mac_framework.h> 99 100 #include <vm/vm.h> 101 #include <vm/vm_param.h> 102 #include <vm/pmap.h> 103 #include <vm/vm_object.h> 104 #include <vm/vm_map.h> 105 #include <vm/vm_page.h> 106 #include <vm/vm_pager.h> 107 108 FEATURE(sysv_shm, "System V shared memory segments support"); 109 110 static MALLOC_DEFINE(M_SHM, "shm", "SVID compatible shared memory segments"); 111 112 static int shm_last_free, shm_nused, shmalloced; 113 vm_size_t shm_committed; 114 static struct shmid_kernel *shmsegs; 115 static unsigned shm_prison_slot; 116 117 struct shmmap_state { 118 vm_offset_t va; 119 int shmid; 120 }; 121 122 static void shm_deallocate_segment(struct shmid_kernel *); 123 static int shm_find_segment_by_key(struct prison *, key_t); 124 static struct shmid_kernel *shm_find_segment(struct prison *, int, bool); 125 static int shm_delete_mapping(struct vmspace *vm, struct shmmap_state *); 126 static int shmget_allocate_segment(struct thread *td, key_t key, size_t size, 127 int mode); 128 static int shmget_existing(struct thread *td, size_t size, int shmflg, 129 int mode, int segnum); 130 static void shmrealloc(void); 131 static int shminit(void); 132 static int sysvshm_modload(struct module *, int, void *); 133 static int shmunload(void); 134 #ifndef SYSVSHM 135 static void shmexit_myhook(struct vmspace *vm); 136 static void shmfork_myhook(struct proc *p1, struct proc *p2); 137 #endif 138 static int sysctl_shmsegs(SYSCTL_HANDLER_ARGS); 139 static void shm_remove(struct shmid_kernel *, int); 140 static struct prison *shm_find_prison(struct ucred *); 141 static int shm_prison_cansee(struct prison *, struct shmid_kernel *); 142 static int shm_prison_check(void *, void *); 143 static int shm_prison_set(void *, void *); 144 static int shm_prison_get(void *, void *); 145 static int shm_prison_remove(void *, void *); 146 static void shm_prison_cleanup(struct prison *); 147 148 /* 149 * Tuneable values. 150 */ 151 #ifndef SHMMAXPGS 152 #define SHMMAXPGS 131072ul /* Note: sysv shared memory is swap backed. */ 153 #endif 154 #ifndef SHMMAX 155 #define SHMMAX (SHMMAXPGS*PAGE_SIZE) 156 #endif 157 #ifndef SHMMIN 158 #define SHMMIN 1 159 #endif 160 #ifndef SHMMNI 161 #define SHMMNI 192 162 #endif 163 #ifndef SHMSEG 164 #define SHMSEG 128 165 #endif 166 #ifndef SHMALL 167 #define SHMALL (SHMMAXPGS) 168 #endif 169 170 struct shminfo shminfo = { 171 .shmmax = SHMMAX, 172 .shmmin = SHMMIN, 173 .shmmni = SHMMNI, 174 .shmseg = SHMSEG, 175 .shmall = SHMALL 176 }; 177 178 static int shm_use_phys; 179 static int shm_allow_removed = 1; 180 181 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmax, CTLFLAG_RWTUN, &shminfo.shmmax, 0, 182 "Maximum shared memory segment size"); 183 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmin, CTLFLAG_RWTUN, &shminfo.shmmin, 0, 184 "Minimum shared memory segment size"); 185 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmmni, CTLFLAG_RDTUN, &shminfo.shmmni, 0, 186 "Number of shared memory identifiers"); 187 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmseg, CTLFLAG_RDTUN, &shminfo.shmseg, 0, 188 "Number of segments per process"); 189 SYSCTL_ULONG(_kern_ipc, OID_AUTO, shmall, CTLFLAG_RWTUN, &shminfo.shmall, 0, 190 "Maximum number of pages available for shared memory"); 191 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_use_phys, CTLFLAG_RWTUN, 192 &shm_use_phys, 0, "Enable/Disable locking of shared memory pages in core"); 193 SYSCTL_INT(_kern_ipc, OID_AUTO, shm_allow_removed, CTLFLAG_RWTUN, 194 &shm_allow_removed, 0, 195 "Enable/Disable attachment to attached segments marked for removal"); 196 SYSCTL_PROC(_kern_ipc, OID_AUTO, shmsegs, CTLTYPE_OPAQUE | CTLFLAG_RD | 197 CTLFLAG_MPSAFE, NULL, 0, sysctl_shmsegs, "", 198 "Array of struct shmid_kernel for each potential shared memory segment"); 199 200 static struct sx sysvshmsx; 201 #define SYSVSHM_LOCK() sx_xlock(&sysvshmsx) 202 #define SYSVSHM_UNLOCK() sx_xunlock(&sysvshmsx) 203 #define SYSVSHM_ASSERT_LOCKED() sx_assert(&sysvshmsx, SA_XLOCKED) 204 205 static int 206 shm_find_segment_by_key(struct prison *pr, key_t key) 207 { 208 int i; 209 210 for (i = 0; i < shmalloced; i++) 211 if ((shmsegs[i].u.shm_perm.mode & SHMSEG_ALLOCATED) && 212 shmsegs[i].cred != NULL && 213 shmsegs[i].cred->cr_prison == pr && 214 shmsegs[i].u.shm_perm.key == key) 215 return (i); 216 return (-1); 217 } 218 219 /* 220 * Finds segment either by shmid if is_shmid is true, or by segnum if 221 * is_shmid is false. 222 */ 223 static struct shmid_kernel * 224 shm_find_segment(struct prison *rpr, int arg, bool is_shmid) 225 { 226 struct shmid_kernel *shmseg; 227 int segnum; 228 229 segnum = is_shmid ? IPCID_TO_IX(arg) : arg; 230 if (segnum < 0 || segnum >= shmalloced) 231 return (NULL); 232 shmseg = &shmsegs[segnum]; 233 if ((shmseg->u.shm_perm.mode & SHMSEG_ALLOCATED) == 0 || 234 (!shm_allow_removed && 235 (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) != 0) || 236 (is_shmid && shmseg->u.shm_perm.seq != IPCID_TO_SEQ(arg)) || 237 shm_prison_cansee(rpr, shmseg) != 0) 238 return (NULL); 239 return (shmseg); 240 } 241 242 static void 243 shm_deallocate_segment(struct shmid_kernel *shmseg) 244 { 245 vm_size_t size; 246 247 SYSVSHM_ASSERT_LOCKED(); 248 249 vm_object_deallocate(shmseg->object); 250 shmseg->object = NULL; 251 size = round_page(shmseg->u.shm_segsz); 252 shm_committed -= btoc(size); 253 shm_nused--; 254 shmseg->u.shm_perm.mode = SHMSEG_FREE; 255 #ifdef MAC 256 mac_sysvshm_cleanup(shmseg); 257 #endif 258 racct_sub_cred(shmseg->cred, RACCT_NSHM, 1); 259 racct_sub_cred(shmseg->cred, RACCT_SHMSIZE, size); 260 crfree(shmseg->cred); 261 shmseg->cred = NULL; 262 } 263 264 static int 265 shm_delete_mapping(struct vmspace *vm, struct shmmap_state *shmmap_s) 266 { 267 struct shmid_kernel *shmseg; 268 int segnum, result; 269 vm_size_t size; 270 271 SYSVSHM_ASSERT_LOCKED(); 272 segnum = IPCID_TO_IX(shmmap_s->shmid); 273 KASSERT(segnum >= 0 && segnum < shmalloced, 274 ("segnum %d shmalloced %d", segnum, shmalloced)); 275 276 shmseg = &shmsegs[segnum]; 277 size = round_page(shmseg->u.shm_segsz); 278 result = vm_map_remove(&vm->vm_map, shmmap_s->va, shmmap_s->va + size); 279 if (result != KERN_SUCCESS) 280 return (EINVAL); 281 shmmap_s->shmid = -1; 282 shmseg->u.shm_dtime = time_second; 283 if (--shmseg->u.shm_nattch == 0 && 284 (shmseg->u.shm_perm.mode & SHMSEG_REMOVED)) { 285 shm_deallocate_segment(shmseg); 286 shm_last_free = segnum; 287 } 288 return (0); 289 } 290 291 static void 292 shm_remove(struct shmid_kernel *shmseg, int segnum) 293 { 294 295 shmseg->u.shm_perm.key = IPC_PRIVATE; 296 shmseg->u.shm_perm.mode |= SHMSEG_REMOVED; 297 if (shmseg->u.shm_nattch == 0) { 298 shm_deallocate_segment(shmseg); 299 shm_last_free = segnum; 300 } 301 } 302 303 static struct prison * 304 shm_find_prison(struct ucred *cred) 305 { 306 struct prison *pr, *rpr; 307 308 pr = cred->cr_prison; 309 prison_lock(pr); 310 rpr = osd_jail_get(pr, shm_prison_slot); 311 prison_unlock(pr); 312 return rpr; 313 } 314 315 static int 316 shm_prison_cansee(struct prison *rpr, struct shmid_kernel *shmseg) 317 { 318 319 if (shmseg->cred == NULL || 320 !(rpr == shmseg->cred->cr_prison || 321 prison_ischild(rpr, shmseg->cred->cr_prison))) 322 return (EINVAL); 323 return (0); 324 } 325 326 static int 327 kern_shmdt_locked(struct thread *td, const void *shmaddr) 328 { 329 struct proc *p = td->td_proc; 330 struct shmmap_state *shmmap_s; 331 #ifdef MAC 332 int error; 333 #endif 334 int i; 335 336 SYSVSHM_ASSERT_LOCKED(); 337 if (shm_find_prison(td->td_ucred) == NULL) 338 return (ENOSYS); 339 shmmap_s = p->p_vmspace->vm_shm; 340 if (shmmap_s == NULL) 341 return (EINVAL); 342 AUDIT_ARG_SVIPC_ID(shmmap_s->shmid); 343 for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) { 344 if (shmmap_s->shmid != -1 && 345 shmmap_s->va == (vm_offset_t)shmaddr) { 346 break; 347 } 348 } 349 if (i == shminfo.shmseg) 350 return (EINVAL); 351 #ifdef MAC 352 error = mac_sysvshm_check_shmdt(td->td_ucred, 353 &shmsegs[IPCID_TO_IX(shmmap_s->shmid)]); 354 if (error != 0) 355 return (error); 356 #endif 357 return (shm_delete_mapping(p->p_vmspace, shmmap_s)); 358 } 359 360 #ifndef _SYS_SYSPROTO_H_ 361 struct shmdt_args { 362 const void *shmaddr; 363 }; 364 #endif 365 int 366 sys_shmdt(struct thread *td, struct shmdt_args *uap) 367 { 368 int error; 369 370 SYSVSHM_LOCK(); 371 error = kern_shmdt_locked(td, uap->shmaddr); 372 SYSVSHM_UNLOCK(); 373 return (error); 374 } 375 376 static int 377 kern_shmat_locked(struct thread *td, int shmid, const void *shmaddr, 378 int shmflg) 379 { 380 struct prison *rpr; 381 struct proc *p = td->td_proc; 382 struct shmid_kernel *shmseg; 383 struct shmmap_state *shmmap_s; 384 vm_offset_t attach_va; 385 vm_prot_t prot; 386 vm_size_t size; 387 int cow, error, find_space, i, rv; 388 389 AUDIT_ARG_SVIPC_ID(shmid); 390 AUDIT_ARG_VALUE(shmflg); 391 392 SYSVSHM_ASSERT_LOCKED(); 393 rpr = shm_find_prison(td->td_ucred); 394 if (rpr == NULL) 395 return (ENOSYS); 396 shmmap_s = p->p_vmspace->vm_shm; 397 if (shmmap_s == NULL) { 398 shmmap_s = malloc(shminfo.shmseg * sizeof(struct shmmap_state), 399 M_SHM, M_WAITOK); 400 for (i = 0; i < shminfo.shmseg; i++) 401 shmmap_s[i].shmid = -1; 402 KASSERT(p->p_vmspace->vm_shm == NULL, ("raced")); 403 p->p_vmspace->vm_shm = shmmap_s; 404 } 405 shmseg = shm_find_segment(rpr, shmid, true); 406 if (shmseg == NULL) 407 return (EINVAL); 408 error = ipcperm(td, &shmseg->u.shm_perm, 409 (shmflg & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W); 410 if (error != 0) 411 return (error); 412 #ifdef MAC 413 error = mac_sysvshm_check_shmat(td->td_ucred, shmseg, shmflg); 414 if (error != 0) 415 return (error); 416 #endif 417 for (i = 0; i < shminfo.shmseg; i++) { 418 if (shmmap_s->shmid == -1) 419 break; 420 shmmap_s++; 421 } 422 if (i >= shminfo.shmseg) 423 return (EMFILE); 424 size = round_page(shmseg->u.shm_segsz); 425 prot = VM_PROT_READ; 426 cow = MAP_INHERIT_SHARE | MAP_PREFAULT_PARTIAL; 427 if ((shmflg & SHM_RDONLY) == 0) 428 prot |= VM_PROT_WRITE; 429 if (shmaddr != NULL) { 430 if ((shmflg & SHM_RND) != 0) 431 attach_va = rounddown2((vm_offset_t)shmaddr, SHMLBA); 432 else if (((vm_offset_t)shmaddr & (SHMLBA-1)) == 0) 433 attach_va = (vm_offset_t)shmaddr; 434 else 435 return (EINVAL); 436 if ((shmflg & SHM_REMAP) != 0) 437 cow |= MAP_REMAP; 438 find_space = VMFS_NO_SPACE; 439 } else { 440 /* 441 * This is just a hint to vm_map_find() about where to 442 * put it. 443 */ 444 attach_va = round_page((vm_offset_t)p->p_vmspace->vm_daddr + 445 lim_max(td, RLIMIT_DATA)); 446 find_space = VMFS_OPTIMAL_SPACE; 447 } 448 449 vm_object_reference(shmseg->object); 450 rv = vm_map_find(&p->p_vmspace->vm_map, shmseg->object, 0, &attach_va, 451 size, 0, find_space, prot, prot, cow); 452 if (rv != KERN_SUCCESS) { 453 vm_object_deallocate(shmseg->object); 454 return (ENOMEM); 455 } 456 457 shmmap_s->va = attach_va; 458 shmmap_s->shmid = shmid; 459 shmseg->u.shm_lpid = p->p_pid; 460 shmseg->u.shm_atime = time_second; 461 shmseg->u.shm_nattch++; 462 td->td_retval[0] = attach_va; 463 return (error); 464 } 465 466 int 467 kern_shmat(struct thread *td, int shmid, const void *shmaddr, int shmflg) 468 { 469 int error; 470 471 SYSVSHM_LOCK(); 472 error = kern_shmat_locked(td, shmid, shmaddr, shmflg); 473 SYSVSHM_UNLOCK(); 474 return (error); 475 } 476 477 #ifndef _SYS_SYSPROTO_H_ 478 struct shmat_args { 479 int shmid; 480 const void *shmaddr; 481 int shmflg; 482 }; 483 #endif 484 int 485 sys_shmat(struct thread *td, struct shmat_args *uap) 486 { 487 488 return (kern_shmat(td, uap->shmid, uap->shmaddr, uap->shmflg)); 489 } 490 491 static int 492 kern_shmctl_locked(struct thread *td, int shmid, int cmd, void *buf, 493 size_t *bufsz) 494 { 495 struct prison *rpr; 496 struct shmid_kernel *shmseg; 497 struct shmid_ds *shmidp; 498 struct shm_info shm_info; 499 int error; 500 501 SYSVSHM_ASSERT_LOCKED(); 502 503 rpr = shm_find_prison(td->td_ucred); 504 if (rpr == NULL) 505 return (ENOSYS); 506 507 AUDIT_ARG_SVIPC_ID(shmid); 508 AUDIT_ARG_SVIPC_CMD(cmd); 509 510 switch (cmd) { 511 /* 512 * It is possible that kern_shmctl is being called from the Linux ABI 513 * layer, in which case, we will need to implement IPC_INFO. It should 514 * be noted that other shmctl calls will be funneled through here for 515 * Linix binaries as well. 516 * 517 * NB: The Linux ABI layer will convert this data to structure(s) more 518 * consistent with the Linux ABI. 519 */ 520 case IPC_INFO: 521 memcpy(buf, &shminfo, sizeof(shminfo)); 522 if (bufsz) 523 *bufsz = sizeof(shminfo); 524 td->td_retval[0] = shmalloced; 525 return (0); 526 case SHM_INFO: { 527 shm_info.used_ids = shm_nused; 528 shm_info.shm_rss = 0; /*XXX where to get from ? */ 529 shm_info.shm_tot = 0; /*XXX where to get from ? */ 530 shm_info.shm_swp = 0; /*XXX where to get from ? */ 531 shm_info.swap_attempts = 0; /*XXX where to get from ? */ 532 shm_info.swap_successes = 0; /*XXX where to get from ? */ 533 memcpy(buf, &shm_info, sizeof(shm_info)); 534 if (bufsz != NULL) 535 *bufsz = sizeof(shm_info); 536 td->td_retval[0] = shmalloced; 537 return (0); 538 } 539 } 540 shmseg = shm_find_segment(rpr, shmid, cmd != SHM_STAT); 541 if (shmseg == NULL) 542 return (EINVAL); 543 #ifdef MAC 544 error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, cmd); 545 if (error != 0) 546 return (error); 547 #endif 548 switch (cmd) { 549 case SHM_STAT: 550 case IPC_STAT: 551 shmidp = (struct shmid_ds *)buf; 552 error = ipcperm(td, &shmseg->u.shm_perm, IPC_R); 553 if (error != 0) 554 return (error); 555 memcpy(shmidp, &shmseg->u, sizeof(struct shmid_ds)); 556 if (td->td_ucred->cr_prison != shmseg->cred->cr_prison) 557 shmidp->shm_perm.key = IPC_PRIVATE; 558 if (bufsz != NULL) 559 *bufsz = sizeof(struct shmid_ds); 560 if (cmd == SHM_STAT) { 561 td->td_retval[0] = IXSEQ_TO_IPCID(shmid, 562 shmseg->u.shm_perm); 563 } 564 break; 565 case IPC_SET: 566 shmidp = (struct shmid_ds *)buf; 567 AUDIT_ARG_SVIPC_PERM(&shmidp->shm_perm); 568 error = ipcperm(td, &shmseg->u.shm_perm, IPC_M); 569 if (error != 0) 570 return (error); 571 shmseg->u.shm_perm.uid = shmidp->shm_perm.uid; 572 shmseg->u.shm_perm.gid = shmidp->shm_perm.gid; 573 shmseg->u.shm_perm.mode = 574 (shmseg->u.shm_perm.mode & ~ACCESSPERMS) | 575 (shmidp->shm_perm.mode & ACCESSPERMS); 576 shmseg->u.shm_ctime = time_second; 577 break; 578 case IPC_RMID: 579 error = ipcperm(td, &shmseg->u.shm_perm, IPC_M); 580 if (error != 0) 581 return (error); 582 shm_remove(shmseg, IPCID_TO_IX(shmid)); 583 break; 584 #if 0 585 case SHM_LOCK: 586 case SHM_UNLOCK: 587 #endif 588 default: 589 error = EINVAL; 590 break; 591 } 592 return (error); 593 } 594 595 int 596 kern_shmctl(struct thread *td, int shmid, int cmd, void *buf, size_t *bufsz) 597 { 598 int error; 599 600 SYSVSHM_LOCK(); 601 error = kern_shmctl_locked(td, shmid, cmd, buf, bufsz); 602 SYSVSHM_UNLOCK(); 603 return (error); 604 } 605 606 #ifndef _SYS_SYSPROTO_H_ 607 struct shmctl_args { 608 int shmid; 609 int cmd; 610 struct shmid_ds *buf; 611 }; 612 #endif 613 int 614 sys_shmctl(struct thread *td, struct shmctl_args *uap) 615 { 616 int error; 617 struct shmid_ds buf; 618 size_t bufsz; 619 620 /* 621 * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support 622 * Linux binaries. If we see the call come through the FreeBSD ABI, 623 * return an error back to the user since we do not to support this. 624 */ 625 if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO || 626 uap->cmd == SHM_STAT) 627 return (EINVAL); 628 629 /* IPC_SET needs to copyin the buffer before calling kern_shmctl */ 630 if (uap->cmd == IPC_SET) { 631 if ((error = copyin(uap->buf, &buf, sizeof(struct shmid_ds)))) 632 goto done; 633 } 634 635 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz); 636 if (error) 637 goto done; 638 639 /* Cases in which we need to copyout */ 640 switch (uap->cmd) { 641 case IPC_STAT: 642 error = copyout(&buf, uap->buf, bufsz); 643 break; 644 } 645 646 done: 647 if (error) { 648 /* Invalidate the return value */ 649 td->td_retval[0] = -1; 650 } 651 return (error); 652 } 653 654 static int 655 shmget_existing(struct thread *td, size_t size, int shmflg, int mode, 656 int segnum) 657 { 658 struct shmid_kernel *shmseg; 659 #ifdef MAC 660 int error; 661 #endif 662 663 SYSVSHM_ASSERT_LOCKED(); 664 KASSERT(segnum >= 0 && segnum < shmalloced, 665 ("segnum %d shmalloced %d", segnum, shmalloced)); 666 shmseg = &shmsegs[segnum]; 667 if ((shmflg & (IPC_CREAT | IPC_EXCL)) == (IPC_CREAT | IPC_EXCL)) 668 return (EEXIST); 669 #ifdef MAC 670 error = mac_sysvshm_check_shmget(td->td_ucred, shmseg, shmflg); 671 if (error != 0) 672 return (error); 673 #endif 674 if (size != 0 && size > shmseg->u.shm_segsz) 675 return (EINVAL); 676 td->td_retval[0] = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm); 677 return (0); 678 } 679 680 static int 681 shmget_allocate_segment(struct thread *td, key_t key, size_t size, int mode) 682 { 683 struct ucred *cred = td->td_ucred; 684 struct shmid_kernel *shmseg; 685 vm_object_t shm_object; 686 int i, segnum; 687 688 SYSVSHM_ASSERT_LOCKED(); 689 690 if (size < shminfo.shmmin || size > shminfo.shmmax) 691 return (EINVAL); 692 if (shm_nused >= shminfo.shmmni) /* Any shmids left? */ 693 return (ENOSPC); 694 size = round_page(size); 695 if (shm_committed + btoc(size) > shminfo.shmall) 696 return (ENOMEM); 697 if (shm_last_free < 0) { 698 shmrealloc(); /* Maybe expand the shmsegs[] array. */ 699 for (i = 0; i < shmalloced; i++) 700 if (shmsegs[i].u.shm_perm.mode & SHMSEG_FREE) 701 break; 702 if (i == shmalloced) 703 return (ENOSPC); 704 segnum = i; 705 } else { 706 segnum = shm_last_free; 707 shm_last_free = -1; 708 } 709 KASSERT(segnum >= 0 && segnum < shmalloced, 710 ("segnum %d shmalloced %d", segnum, shmalloced)); 711 shmseg = &shmsegs[segnum]; 712 #ifdef RACCT 713 if (racct_enable) { 714 PROC_LOCK(td->td_proc); 715 if (racct_add(td->td_proc, RACCT_NSHM, 1)) { 716 PROC_UNLOCK(td->td_proc); 717 return (ENOSPC); 718 } 719 if (racct_add(td->td_proc, RACCT_SHMSIZE, size)) { 720 racct_sub(td->td_proc, RACCT_NSHM, 1); 721 PROC_UNLOCK(td->td_proc); 722 return (ENOMEM); 723 } 724 PROC_UNLOCK(td->td_proc); 725 } 726 #endif 727 728 /* 729 * We make sure that we have allocated a pager before we need 730 * to. 731 */ 732 shm_object = vm_pager_allocate(shm_use_phys ? OBJT_PHYS : OBJT_SWAP, 733 0, size, VM_PROT_DEFAULT, 0, cred); 734 if (shm_object == NULL) { 735 #ifdef RACCT 736 if (racct_enable) { 737 PROC_LOCK(td->td_proc); 738 racct_sub(td->td_proc, RACCT_NSHM, 1); 739 racct_sub(td->td_proc, RACCT_SHMSIZE, size); 740 PROC_UNLOCK(td->td_proc); 741 } 742 #endif 743 return (ENOMEM); 744 } 745 746 shmseg->object = shm_object; 747 shmseg->u.shm_perm.cuid = shmseg->u.shm_perm.uid = cred->cr_uid; 748 shmseg->u.shm_perm.cgid = shmseg->u.shm_perm.gid = cred->cr_gid; 749 shmseg->u.shm_perm.mode = (mode & ACCESSPERMS) | SHMSEG_ALLOCATED; 750 shmseg->u.shm_perm.key = key; 751 shmseg->u.shm_perm.seq = (shmseg->u.shm_perm.seq + 1) & 0x7fff; 752 shmseg->cred = crhold(cred); 753 shmseg->u.shm_segsz = size; 754 shmseg->u.shm_cpid = td->td_proc->p_pid; 755 shmseg->u.shm_lpid = shmseg->u.shm_nattch = 0; 756 shmseg->u.shm_atime = shmseg->u.shm_dtime = 0; 757 #ifdef MAC 758 mac_sysvshm_create(cred, shmseg); 759 #endif 760 shmseg->u.shm_ctime = time_second; 761 shm_committed += btoc(size); 762 shm_nused++; 763 td->td_retval[0] = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm); 764 765 return (0); 766 } 767 768 #ifndef _SYS_SYSPROTO_H_ 769 struct shmget_args { 770 key_t key; 771 size_t size; 772 int shmflg; 773 }; 774 #endif 775 int 776 sys_shmget(struct thread *td, struct shmget_args *uap) 777 { 778 int segnum, mode; 779 int error; 780 781 if (shm_find_prison(td->td_ucred) == NULL) 782 return (ENOSYS); 783 mode = uap->shmflg & ACCESSPERMS; 784 SYSVSHM_LOCK(); 785 if (uap->key == IPC_PRIVATE) { 786 error = shmget_allocate_segment(td, uap->key, uap->size, mode); 787 } else { 788 segnum = shm_find_segment_by_key(td->td_ucred->cr_prison, 789 uap->key); 790 if (segnum >= 0) 791 error = shmget_existing(td, uap->size, uap->shmflg, 792 mode, segnum); 793 else if ((uap->shmflg & IPC_CREAT) == 0) 794 error = ENOENT; 795 else 796 error = shmget_allocate_segment(td, uap->key, 797 uap->size, mode); 798 } 799 SYSVSHM_UNLOCK(); 800 return (error); 801 } 802 803 #ifdef SYSVSHM 804 void 805 shmfork(struct proc *p1, struct proc *p2) 806 #else 807 static void 808 shmfork_myhook(struct proc *p1, struct proc *p2) 809 #endif 810 { 811 struct shmmap_state *shmmap_s; 812 size_t size; 813 int i; 814 815 SYSVSHM_LOCK(); 816 size = shminfo.shmseg * sizeof(struct shmmap_state); 817 shmmap_s = malloc(size, M_SHM, M_WAITOK); 818 bcopy(p1->p_vmspace->vm_shm, shmmap_s, size); 819 p2->p_vmspace->vm_shm = shmmap_s; 820 for (i = 0; i < shminfo.shmseg; i++, shmmap_s++) { 821 if (shmmap_s->shmid != -1) { 822 KASSERT(IPCID_TO_IX(shmmap_s->shmid) >= 0 && 823 IPCID_TO_IX(shmmap_s->shmid) < shmalloced, 824 ("segnum %d shmalloced %d", 825 IPCID_TO_IX(shmmap_s->shmid), shmalloced)); 826 shmsegs[IPCID_TO_IX(shmmap_s->shmid)].u.shm_nattch++; 827 } 828 } 829 SYSVSHM_UNLOCK(); 830 } 831 832 #ifdef SYSVSHM 833 void 834 shmexit(struct vmspace *vm) 835 #else 836 static void 837 shmexit_myhook(struct vmspace *vm) 838 #endif 839 { 840 struct shmmap_state *base, *shm; 841 int i; 842 843 base = vm->vm_shm; 844 if (base != NULL) { 845 vm->vm_shm = NULL; 846 SYSVSHM_LOCK(); 847 for (i = 0, shm = base; i < shminfo.shmseg; i++, shm++) { 848 if (shm->shmid != -1) 849 shm_delete_mapping(vm, shm); 850 } 851 SYSVSHM_UNLOCK(); 852 free(base, M_SHM); 853 } 854 } 855 856 static void 857 shmrealloc(void) 858 { 859 struct shmid_kernel *newsegs; 860 int i; 861 862 SYSVSHM_ASSERT_LOCKED(); 863 864 if (shmalloced >= shminfo.shmmni) 865 return; 866 867 newsegs = malloc(shminfo.shmmni * sizeof(*newsegs), M_SHM, 868 M_WAITOK | M_ZERO); 869 for (i = 0; i < shmalloced; i++) 870 bcopy(&shmsegs[i], &newsegs[i], sizeof(newsegs[0])); 871 for (; i < shminfo.shmmni; i++) { 872 newsegs[i].u.shm_perm.mode = SHMSEG_FREE; 873 newsegs[i].u.shm_perm.seq = 0; 874 #ifdef MAC 875 mac_sysvshm_init(&newsegs[i]); 876 #endif 877 } 878 free(shmsegs, M_SHM); 879 shmsegs = newsegs; 880 shmalloced = shminfo.shmmni; 881 } 882 883 static struct syscall_helper_data shm_syscalls[] = { 884 SYSCALL_INIT_HELPER(shmat), 885 SYSCALL_INIT_HELPER(shmctl), 886 SYSCALL_INIT_HELPER(shmdt), 887 SYSCALL_INIT_HELPER(shmget), 888 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \ 889 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7) 890 SYSCALL_INIT_HELPER_COMPAT(freebsd7_shmctl), 891 #endif 892 #if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43)) 893 SYSCALL_INIT_HELPER(shmsys), 894 #endif 895 SYSCALL_INIT_LAST 896 }; 897 898 #ifdef COMPAT_FREEBSD32 899 #include <compat/freebsd32/freebsd32.h> 900 #include <compat/freebsd32/freebsd32_ipc.h> 901 #include <compat/freebsd32/freebsd32_proto.h> 902 #include <compat/freebsd32/freebsd32_signal.h> 903 #include <compat/freebsd32/freebsd32_syscall.h> 904 #include <compat/freebsd32/freebsd32_util.h> 905 906 static struct syscall_helper_data shm32_syscalls[] = { 907 SYSCALL32_INIT_HELPER_COMPAT(shmat), 908 SYSCALL32_INIT_HELPER_COMPAT(shmdt), 909 SYSCALL32_INIT_HELPER_COMPAT(shmget), 910 SYSCALL32_INIT_HELPER(freebsd32_shmsys), 911 SYSCALL32_INIT_HELPER(freebsd32_shmctl), 912 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \ 913 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7) 914 SYSCALL32_INIT_HELPER(freebsd7_freebsd32_shmctl), 915 #endif 916 SYSCALL_INIT_LAST 917 }; 918 #endif 919 920 static int 921 shminit(void) 922 { 923 struct prison *pr; 924 void **rsv; 925 int i, error; 926 osd_method_t methods[PR_MAXMETHOD] = { 927 [PR_METHOD_CHECK] = shm_prison_check, 928 [PR_METHOD_SET] = shm_prison_set, 929 [PR_METHOD_GET] = shm_prison_get, 930 [PR_METHOD_REMOVE] = shm_prison_remove, 931 }; 932 933 #ifndef BURN_BRIDGES 934 if (TUNABLE_ULONG_FETCH("kern.ipc.shmmaxpgs", &shminfo.shmall) != 0) 935 printf("kern.ipc.shmmaxpgs is now called kern.ipc.shmall!\n"); 936 #endif 937 if (shminfo.shmmax == SHMMAX) { 938 /* Initialize shmmax dealing with possible overflow. */ 939 for (i = PAGE_SIZE; i != 0; i--) { 940 shminfo.shmmax = shminfo.shmall * i; 941 if ((shminfo.shmmax / shminfo.shmall) == (u_long)i) 942 break; 943 } 944 } 945 shmalloced = shminfo.shmmni; 946 shmsegs = malloc(shmalloced * sizeof(shmsegs[0]), M_SHM, 947 M_WAITOK|M_ZERO); 948 for (i = 0; i < shmalloced; i++) { 949 shmsegs[i].u.shm_perm.mode = SHMSEG_FREE; 950 shmsegs[i].u.shm_perm.seq = 0; 951 #ifdef MAC 952 mac_sysvshm_init(&shmsegs[i]); 953 #endif 954 } 955 shm_last_free = 0; 956 shm_nused = 0; 957 shm_committed = 0; 958 sx_init(&sysvshmsx, "sysvshmsx"); 959 #ifndef SYSVSHM 960 shmexit_hook = &shmexit_myhook; 961 shmfork_hook = &shmfork_myhook; 962 #endif 963 964 /* Set current prisons according to their allow.sysvipc. */ 965 shm_prison_slot = osd_jail_register(NULL, methods); 966 rsv = osd_reserve(shm_prison_slot); 967 prison_lock(&prison0); 968 (void)osd_jail_set_reserved(&prison0, shm_prison_slot, rsv, &prison0); 969 prison_unlock(&prison0); 970 rsv = NULL; 971 sx_slock(&allprison_lock); 972 TAILQ_FOREACH(pr, &allprison, pr_list) { 973 if (rsv == NULL) 974 rsv = osd_reserve(shm_prison_slot); 975 prison_lock(pr); 976 if (pr->pr_allow & PR_ALLOW_SYSVIPC) { 977 (void)osd_jail_set_reserved(pr, shm_prison_slot, rsv, 978 &prison0); 979 rsv = NULL; 980 } 981 prison_unlock(pr); 982 } 983 if (rsv != NULL) 984 osd_free_reserved(rsv); 985 sx_sunlock(&allprison_lock); 986 987 error = syscall_helper_register(shm_syscalls, SY_THR_STATIC_KLD); 988 if (error != 0) 989 return (error); 990 #ifdef COMPAT_FREEBSD32 991 error = syscall32_helper_register(shm32_syscalls, SY_THR_STATIC_KLD); 992 if (error != 0) 993 return (error); 994 #endif 995 return (0); 996 } 997 998 static int 999 shmunload(void) 1000 { 1001 int i; 1002 1003 if (shm_nused > 0) 1004 return (EBUSY); 1005 1006 #ifdef COMPAT_FREEBSD32 1007 syscall32_helper_unregister(shm32_syscalls); 1008 #endif 1009 syscall_helper_unregister(shm_syscalls); 1010 if (shm_prison_slot != 0) 1011 osd_jail_deregister(shm_prison_slot); 1012 1013 for (i = 0; i < shmalloced; i++) { 1014 #ifdef MAC 1015 mac_sysvshm_destroy(&shmsegs[i]); 1016 #endif 1017 /* 1018 * Objects might be still mapped into the processes 1019 * address spaces. Actual free would happen on the 1020 * last mapping destruction. 1021 */ 1022 if (shmsegs[i].u.shm_perm.mode != SHMSEG_FREE) 1023 vm_object_deallocate(shmsegs[i].object); 1024 } 1025 free(shmsegs, M_SHM); 1026 #ifndef SYSVSHM 1027 shmexit_hook = NULL; 1028 shmfork_hook = NULL; 1029 #endif 1030 sx_destroy(&sysvshmsx); 1031 return (0); 1032 } 1033 1034 static int 1035 sysctl_shmsegs(SYSCTL_HANDLER_ARGS) 1036 { 1037 struct shmid_kernel tshmseg; 1038 #ifdef COMPAT_FREEBSD32 1039 struct shmid_kernel32 tshmseg32; 1040 #endif 1041 struct prison *pr, *rpr; 1042 void *outaddr; 1043 size_t outsize; 1044 int error, i; 1045 1046 SYSVSHM_LOCK(); 1047 pr = req->td->td_ucred->cr_prison; 1048 rpr = shm_find_prison(req->td->td_ucred); 1049 error = 0; 1050 for (i = 0; i < shmalloced; i++) { 1051 if ((shmsegs[i].u.shm_perm.mode & SHMSEG_ALLOCATED) == 0 || 1052 rpr == NULL || shm_prison_cansee(rpr, &shmsegs[i]) != 0) { 1053 bzero(&tshmseg, sizeof(tshmseg)); 1054 tshmseg.u.shm_perm.mode = SHMSEG_FREE; 1055 } else { 1056 tshmseg = shmsegs[i]; 1057 if (tshmseg.cred->cr_prison != pr) 1058 tshmseg.u.shm_perm.key = IPC_PRIVATE; 1059 } 1060 #ifdef COMPAT_FREEBSD32 1061 if (SV_CURPROC_FLAG(SV_ILP32)) { 1062 bzero(&tshmseg32, sizeof(tshmseg32)); 1063 freebsd32_ipcperm_out(&tshmseg.u.shm_perm, 1064 &tshmseg32.u.shm_perm); 1065 CP(tshmseg, tshmseg32, u.shm_segsz); 1066 CP(tshmseg, tshmseg32, u.shm_lpid); 1067 CP(tshmseg, tshmseg32, u.shm_cpid); 1068 CP(tshmseg, tshmseg32, u.shm_nattch); 1069 CP(tshmseg, tshmseg32, u.shm_atime); 1070 CP(tshmseg, tshmseg32, u.shm_dtime); 1071 CP(tshmseg, tshmseg32, u.shm_ctime); 1072 /* Don't copy object, label, or cred */ 1073 outaddr = &tshmseg32; 1074 outsize = sizeof(tshmseg32); 1075 } else 1076 #endif 1077 { 1078 tshmseg.object = NULL; 1079 tshmseg.label = NULL; 1080 tshmseg.cred = NULL; 1081 outaddr = &tshmseg; 1082 outsize = sizeof(tshmseg); 1083 } 1084 error = SYSCTL_OUT(req, outaddr, outsize); 1085 if (error != 0) 1086 break; 1087 } 1088 SYSVSHM_UNLOCK(); 1089 return (error); 1090 } 1091 1092 int 1093 kern_get_shmsegs(struct thread *td, struct shmid_kernel **res, size_t *sz) 1094 { 1095 struct shmid_kernel *pshmseg; 1096 struct prison *pr, *rpr; 1097 int i; 1098 1099 SYSVSHM_LOCK(); 1100 *sz = shmalloced; 1101 if (res == NULL) 1102 goto out; 1103 1104 pr = td->td_ucred->cr_prison; 1105 rpr = shm_find_prison(td->td_ucred); 1106 *res = malloc(sizeof(struct shmid_kernel) * shmalloced, M_TEMP, 1107 M_WAITOK); 1108 for (i = 0; i < shmalloced; i++) { 1109 pshmseg = &(*res)[i]; 1110 if ((shmsegs[i].u.shm_perm.mode & SHMSEG_ALLOCATED) == 0 || 1111 rpr == NULL || shm_prison_cansee(rpr, &shmsegs[i]) != 0) { 1112 bzero(pshmseg, sizeof(*pshmseg)); 1113 pshmseg->u.shm_perm.mode = SHMSEG_FREE; 1114 } else { 1115 *pshmseg = shmsegs[i]; 1116 if (pshmseg->cred->cr_prison != pr) 1117 pshmseg->u.shm_perm.key = IPC_PRIVATE; 1118 } 1119 pshmseg->object = NULL; 1120 pshmseg->label = NULL; 1121 pshmseg->cred = NULL; 1122 } 1123 out: 1124 SYSVSHM_UNLOCK(); 1125 return (0); 1126 } 1127 1128 static int 1129 shm_prison_check(void *obj, void *data) 1130 { 1131 struct prison *pr = obj; 1132 struct prison *prpr; 1133 struct vfsoptlist *opts = data; 1134 int error, jsys; 1135 1136 /* 1137 * sysvshm is a jailsys integer. 1138 * It must be "disable" if the parent jail is disabled. 1139 */ 1140 error = vfs_copyopt(opts, "sysvshm", &jsys, sizeof(jsys)); 1141 if (error != ENOENT) { 1142 if (error != 0) 1143 return (error); 1144 switch (jsys) { 1145 case JAIL_SYS_DISABLE: 1146 break; 1147 case JAIL_SYS_NEW: 1148 case JAIL_SYS_INHERIT: 1149 prison_lock(pr->pr_parent); 1150 prpr = osd_jail_get(pr->pr_parent, shm_prison_slot); 1151 prison_unlock(pr->pr_parent); 1152 if (prpr == NULL) 1153 return (EPERM); 1154 break; 1155 default: 1156 return (EINVAL); 1157 } 1158 } 1159 1160 return (0); 1161 } 1162 1163 static int 1164 shm_prison_set(void *obj, void *data) 1165 { 1166 struct prison *pr = obj; 1167 struct prison *tpr, *orpr, *nrpr, *trpr; 1168 struct vfsoptlist *opts = data; 1169 void *rsv; 1170 int jsys, descend; 1171 1172 /* 1173 * sysvshm controls which jail is the root of the associated segments 1174 * (this jail or same as the parent), or if the feature is available 1175 * at all. 1176 */ 1177 if (vfs_copyopt(opts, "sysvshm", &jsys, sizeof(jsys)) == ENOENT) 1178 jsys = vfs_flagopt(opts, "allow.sysvipc", NULL, 0) 1179 ? JAIL_SYS_INHERIT 1180 : vfs_flagopt(opts, "allow.nosysvipc", NULL, 0) 1181 ? JAIL_SYS_DISABLE 1182 : -1; 1183 if (jsys == JAIL_SYS_DISABLE) { 1184 prison_lock(pr); 1185 orpr = osd_jail_get(pr, shm_prison_slot); 1186 if (orpr != NULL) 1187 osd_jail_del(pr, shm_prison_slot); 1188 prison_unlock(pr); 1189 if (orpr != NULL) { 1190 if (orpr == pr) 1191 shm_prison_cleanup(pr); 1192 /* Disable all child jails as well. */ 1193 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) { 1194 prison_lock(tpr); 1195 trpr = osd_jail_get(tpr, shm_prison_slot); 1196 if (trpr != NULL) { 1197 osd_jail_del(tpr, shm_prison_slot); 1198 prison_unlock(tpr); 1199 if (trpr == tpr) 1200 shm_prison_cleanup(tpr); 1201 } else { 1202 prison_unlock(tpr); 1203 descend = 0; 1204 } 1205 } 1206 } 1207 } else if (jsys != -1) { 1208 if (jsys == JAIL_SYS_NEW) 1209 nrpr = pr; 1210 else { 1211 prison_lock(pr->pr_parent); 1212 nrpr = osd_jail_get(pr->pr_parent, shm_prison_slot); 1213 prison_unlock(pr->pr_parent); 1214 } 1215 rsv = osd_reserve(shm_prison_slot); 1216 prison_lock(pr); 1217 orpr = osd_jail_get(pr, shm_prison_slot); 1218 if (orpr != nrpr) 1219 (void)osd_jail_set_reserved(pr, shm_prison_slot, rsv, 1220 nrpr); 1221 else 1222 osd_free_reserved(rsv); 1223 prison_unlock(pr); 1224 if (orpr != nrpr) { 1225 if (orpr == pr) 1226 shm_prison_cleanup(pr); 1227 if (orpr != NULL) { 1228 /* Change child jails matching the old root, */ 1229 FOREACH_PRISON_DESCENDANT(pr, tpr, descend) { 1230 prison_lock(tpr); 1231 trpr = osd_jail_get(tpr, 1232 shm_prison_slot); 1233 if (trpr == orpr) { 1234 (void)osd_jail_set(tpr, 1235 shm_prison_slot, nrpr); 1236 prison_unlock(tpr); 1237 if (trpr == tpr) 1238 shm_prison_cleanup(tpr); 1239 } else { 1240 prison_unlock(tpr); 1241 descend = 0; 1242 } 1243 } 1244 } 1245 } 1246 } 1247 1248 return (0); 1249 } 1250 1251 static int 1252 shm_prison_get(void *obj, void *data) 1253 { 1254 struct prison *pr = obj; 1255 struct prison *rpr; 1256 struct vfsoptlist *opts = data; 1257 int error, jsys; 1258 1259 /* Set sysvshm based on the jail's root prison. */ 1260 prison_lock(pr); 1261 rpr = osd_jail_get(pr, shm_prison_slot); 1262 prison_unlock(pr); 1263 jsys = rpr == NULL ? JAIL_SYS_DISABLE 1264 : rpr == pr ? JAIL_SYS_NEW : JAIL_SYS_INHERIT; 1265 error = vfs_setopt(opts, "sysvshm", &jsys, sizeof(jsys)); 1266 if (error == ENOENT) 1267 error = 0; 1268 return (error); 1269 } 1270 1271 static int 1272 shm_prison_remove(void *obj, void *data __unused) 1273 { 1274 struct prison *pr = obj; 1275 struct prison *rpr; 1276 1277 SYSVSHM_LOCK(); 1278 prison_lock(pr); 1279 rpr = osd_jail_get(pr, shm_prison_slot); 1280 prison_unlock(pr); 1281 if (rpr == pr) 1282 shm_prison_cleanup(pr); 1283 SYSVSHM_UNLOCK(); 1284 return (0); 1285 } 1286 1287 static void 1288 shm_prison_cleanup(struct prison *pr) 1289 { 1290 struct shmid_kernel *shmseg; 1291 int i; 1292 1293 /* Remove any segments that belong to this jail. */ 1294 for (i = 0; i < shmalloced; i++) { 1295 shmseg = &shmsegs[i]; 1296 if ((shmseg->u.shm_perm.mode & SHMSEG_ALLOCATED) && 1297 shmseg->cred != NULL && shmseg->cred->cr_prison == pr) { 1298 shm_remove(shmseg, i); 1299 } 1300 } 1301 } 1302 1303 SYSCTL_JAIL_PARAM_SYS_NODE(sysvshm, CTLFLAG_RW, "SYSV shared memory"); 1304 1305 #if defined(__i386__) && (defined(COMPAT_FREEBSD4) || defined(COMPAT_43)) 1306 struct oshmid_ds { 1307 struct ipc_perm_old shm_perm; /* operation perms */ 1308 int shm_segsz; /* size of segment (bytes) */ 1309 u_short shm_cpid; /* pid, creator */ 1310 u_short shm_lpid; /* pid, last operation */ 1311 short shm_nattch; /* no. of current attaches */ 1312 time_t shm_atime; /* last attach time */ 1313 time_t shm_dtime; /* last detach time */ 1314 time_t shm_ctime; /* last change time */ 1315 void *shm_handle; /* internal handle for shm segment */ 1316 }; 1317 1318 struct oshmctl_args { 1319 int shmid; 1320 int cmd; 1321 struct oshmid_ds *ubuf; 1322 }; 1323 1324 static int 1325 oshmctl(struct thread *td, struct oshmctl_args *uap) 1326 { 1327 #ifdef COMPAT_43 1328 int error = 0; 1329 struct prison *rpr; 1330 struct shmid_kernel *shmseg; 1331 struct oshmid_ds outbuf; 1332 1333 rpr = shm_find_prison(td->td_ucred); 1334 if (rpr == NULL) 1335 return (ENOSYS); 1336 if (uap->cmd != IPC_STAT) { 1337 return (freebsd7_shmctl(td, 1338 (struct freebsd7_shmctl_args *)uap)); 1339 } 1340 SYSVSHM_LOCK(); 1341 shmseg = shm_find_segment(rpr, uap->shmid, true); 1342 if (shmseg == NULL) { 1343 SYSVSHM_UNLOCK(); 1344 return (EINVAL); 1345 } 1346 error = ipcperm(td, &shmseg->u.shm_perm, IPC_R); 1347 if (error != 0) { 1348 SYSVSHM_UNLOCK(); 1349 return (error); 1350 } 1351 #ifdef MAC 1352 error = mac_sysvshm_check_shmctl(td->td_ucred, shmseg, uap->cmd); 1353 if (error != 0) { 1354 SYSVSHM_UNLOCK(); 1355 return (error); 1356 } 1357 #endif 1358 ipcperm_new2old(&shmseg->u.shm_perm, &outbuf.shm_perm); 1359 outbuf.shm_segsz = shmseg->u.shm_segsz; 1360 outbuf.shm_cpid = shmseg->u.shm_cpid; 1361 outbuf.shm_lpid = shmseg->u.shm_lpid; 1362 outbuf.shm_nattch = shmseg->u.shm_nattch; 1363 outbuf.shm_atime = shmseg->u.shm_atime; 1364 outbuf.shm_dtime = shmseg->u.shm_dtime; 1365 outbuf.shm_ctime = shmseg->u.shm_ctime; 1366 outbuf.shm_handle = shmseg->object; 1367 SYSVSHM_UNLOCK(); 1368 return (copyout(&outbuf, uap->ubuf, sizeof(outbuf))); 1369 #else 1370 return (EINVAL); 1371 #endif 1372 } 1373 1374 /* XXX casting to (sy_call_t *) is bogus, as usual. */ 1375 static sy_call_t *shmcalls[] = { 1376 (sy_call_t *)sys_shmat, (sy_call_t *)oshmctl, 1377 (sy_call_t *)sys_shmdt, (sy_call_t *)sys_shmget, 1378 (sy_call_t *)freebsd7_shmctl 1379 }; 1380 1381 #ifndef _SYS_SYSPROTO_H_ 1382 /* XXX actually varargs. */ 1383 struct shmsys_args { 1384 int which; 1385 int a2; 1386 int a3; 1387 int a4; 1388 }; 1389 #endif 1390 int 1391 sys_shmsys(struct thread *td, struct shmsys_args *uap) 1392 { 1393 1394 AUDIT_ARG_SVIPC_WHICH(uap->which); 1395 if (uap->which < 0 || uap->which >= nitems(shmcalls)) 1396 return (EINVAL); 1397 return ((*shmcalls[uap->which])(td, &uap->a2)); 1398 } 1399 1400 #endif /* i386 && (COMPAT_FREEBSD4 || COMPAT_43) */ 1401 1402 #ifdef COMPAT_FREEBSD32 1403 1404 int 1405 freebsd32_shmsys(struct thread *td, struct freebsd32_shmsys_args *uap) 1406 { 1407 1408 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \ 1409 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7) 1410 AUDIT_ARG_SVIPC_WHICH(uap->which); 1411 switch (uap->which) { 1412 case 0: { /* shmat */ 1413 struct shmat_args ap; 1414 1415 ap.shmid = uap->a2; 1416 ap.shmaddr = PTRIN(uap->a3); 1417 ap.shmflg = uap->a4; 1418 return (sysent[SYS_shmat].sy_call(td, &ap)); 1419 } 1420 case 2: { /* shmdt */ 1421 struct shmdt_args ap; 1422 1423 ap.shmaddr = PTRIN(uap->a2); 1424 return (sysent[SYS_shmdt].sy_call(td, &ap)); 1425 } 1426 case 3: { /* shmget */ 1427 struct shmget_args ap; 1428 1429 ap.key = uap->a2; 1430 ap.size = uap->a3; 1431 ap.shmflg = uap->a4; 1432 return (sysent[SYS_shmget].sy_call(td, &ap)); 1433 } 1434 case 4: { /* shmctl */ 1435 struct freebsd7_freebsd32_shmctl_args ap; 1436 1437 ap.shmid = uap->a2; 1438 ap.cmd = uap->a3; 1439 ap.buf = PTRIN(uap->a4); 1440 return (freebsd7_freebsd32_shmctl(td, &ap)); 1441 } 1442 case 1: /* oshmctl */ 1443 default: 1444 return (EINVAL); 1445 } 1446 #else 1447 return (nosys(td, NULL)); 1448 #endif 1449 } 1450 1451 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \ 1452 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7) 1453 int 1454 freebsd7_freebsd32_shmctl(struct thread *td, 1455 struct freebsd7_freebsd32_shmctl_args *uap) 1456 { 1457 int error; 1458 union { 1459 struct shmid_ds shmid_ds; 1460 struct shm_info shm_info; 1461 struct shminfo shminfo; 1462 } u; 1463 union { 1464 struct shmid_ds_old32 shmid_ds32; 1465 struct shm_info32 shm_info32; 1466 struct shminfo32 shminfo32; 1467 } u32; 1468 size_t sz; 1469 1470 if (uap->cmd == IPC_SET) { 1471 if ((error = copyin(uap->buf, &u32.shmid_ds32, 1472 sizeof(u32.shmid_ds32)))) 1473 goto done; 1474 freebsd32_ipcperm_old_in(&u32.shmid_ds32.shm_perm, 1475 &u.shmid_ds.shm_perm); 1476 CP(u32.shmid_ds32, u.shmid_ds, shm_segsz); 1477 CP(u32.shmid_ds32, u.shmid_ds, shm_lpid); 1478 CP(u32.shmid_ds32, u.shmid_ds, shm_cpid); 1479 CP(u32.shmid_ds32, u.shmid_ds, shm_nattch); 1480 CP(u32.shmid_ds32, u.shmid_ds, shm_atime); 1481 CP(u32.shmid_ds32, u.shmid_ds, shm_dtime); 1482 CP(u32.shmid_ds32, u.shmid_ds, shm_ctime); 1483 } 1484 1485 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz); 1486 if (error) 1487 goto done; 1488 1489 /* Cases in which we need to copyout */ 1490 switch (uap->cmd) { 1491 case IPC_INFO: 1492 CP(u.shminfo, u32.shminfo32, shmmax); 1493 CP(u.shminfo, u32.shminfo32, shmmin); 1494 CP(u.shminfo, u32.shminfo32, shmmni); 1495 CP(u.shminfo, u32.shminfo32, shmseg); 1496 CP(u.shminfo, u32.shminfo32, shmall); 1497 error = copyout(&u32.shminfo32, uap->buf, 1498 sizeof(u32.shminfo32)); 1499 break; 1500 case SHM_INFO: 1501 CP(u.shm_info, u32.shm_info32, used_ids); 1502 CP(u.shm_info, u32.shm_info32, shm_rss); 1503 CP(u.shm_info, u32.shm_info32, shm_tot); 1504 CP(u.shm_info, u32.shm_info32, shm_swp); 1505 CP(u.shm_info, u32.shm_info32, swap_attempts); 1506 CP(u.shm_info, u32.shm_info32, swap_successes); 1507 error = copyout(&u32.shm_info32, uap->buf, 1508 sizeof(u32.shm_info32)); 1509 break; 1510 case SHM_STAT: 1511 case IPC_STAT: 1512 memset(&u32.shmid_ds32, 0, sizeof(u32.shmid_ds32)); 1513 freebsd32_ipcperm_old_out(&u.shmid_ds.shm_perm, 1514 &u32.shmid_ds32.shm_perm); 1515 if (u.shmid_ds.shm_segsz > INT32_MAX) 1516 u32.shmid_ds32.shm_segsz = INT32_MAX; 1517 else 1518 CP(u.shmid_ds, u32.shmid_ds32, shm_segsz); 1519 CP(u.shmid_ds, u32.shmid_ds32, shm_lpid); 1520 CP(u.shmid_ds, u32.shmid_ds32, shm_cpid); 1521 CP(u.shmid_ds, u32.shmid_ds32, shm_nattch); 1522 CP(u.shmid_ds, u32.shmid_ds32, shm_atime); 1523 CP(u.shmid_ds, u32.shmid_ds32, shm_dtime); 1524 CP(u.shmid_ds, u32.shmid_ds32, shm_ctime); 1525 u32.shmid_ds32.shm_internal = 0; 1526 error = copyout(&u32.shmid_ds32, uap->buf, 1527 sizeof(u32.shmid_ds32)); 1528 break; 1529 } 1530 1531 done: 1532 if (error) { 1533 /* Invalidate the return value */ 1534 td->td_retval[0] = -1; 1535 } 1536 return (error); 1537 } 1538 #endif 1539 1540 int 1541 freebsd32_shmctl(struct thread *td, struct freebsd32_shmctl_args *uap) 1542 { 1543 int error; 1544 union { 1545 struct shmid_ds shmid_ds; 1546 struct shm_info shm_info; 1547 struct shminfo shminfo; 1548 } u; 1549 union { 1550 struct shmid_ds32 shmid_ds32; 1551 struct shm_info32 shm_info32; 1552 struct shminfo32 shminfo32; 1553 } u32; 1554 size_t sz; 1555 1556 if (uap->cmd == IPC_SET) { 1557 if ((error = copyin(uap->buf, &u32.shmid_ds32, 1558 sizeof(u32.shmid_ds32)))) 1559 goto done; 1560 freebsd32_ipcperm_in(&u32.shmid_ds32.shm_perm, 1561 &u.shmid_ds.shm_perm); 1562 CP(u32.shmid_ds32, u.shmid_ds, shm_segsz); 1563 CP(u32.shmid_ds32, u.shmid_ds, shm_lpid); 1564 CP(u32.shmid_ds32, u.shmid_ds, shm_cpid); 1565 CP(u32.shmid_ds32, u.shmid_ds, shm_nattch); 1566 CP(u32.shmid_ds32, u.shmid_ds, shm_atime); 1567 CP(u32.shmid_ds32, u.shmid_ds, shm_dtime); 1568 CP(u32.shmid_ds32, u.shmid_ds, shm_ctime); 1569 } 1570 1571 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&u, &sz); 1572 if (error) 1573 goto done; 1574 1575 /* Cases in which we need to copyout */ 1576 switch (uap->cmd) { 1577 case IPC_INFO: 1578 CP(u.shminfo, u32.shminfo32, shmmax); 1579 CP(u.shminfo, u32.shminfo32, shmmin); 1580 CP(u.shminfo, u32.shminfo32, shmmni); 1581 CP(u.shminfo, u32.shminfo32, shmseg); 1582 CP(u.shminfo, u32.shminfo32, shmall); 1583 error = copyout(&u32.shminfo32, uap->buf, 1584 sizeof(u32.shminfo32)); 1585 break; 1586 case SHM_INFO: 1587 CP(u.shm_info, u32.shm_info32, used_ids); 1588 CP(u.shm_info, u32.shm_info32, shm_rss); 1589 CP(u.shm_info, u32.shm_info32, shm_tot); 1590 CP(u.shm_info, u32.shm_info32, shm_swp); 1591 CP(u.shm_info, u32.shm_info32, swap_attempts); 1592 CP(u.shm_info, u32.shm_info32, swap_successes); 1593 error = copyout(&u32.shm_info32, uap->buf, 1594 sizeof(u32.shm_info32)); 1595 break; 1596 case SHM_STAT: 1597 case IPC_STAT: 1598 freebsd32_ipcperm_out(&u.shmid_ds.shm_perm, 1599 &u32.shmid_ds32.shm_perm); 1600 if (u.shmid_ds.shm_segsz > INT32_MAX) 1601 u32.shmid_ds32.shm_segsz = INT32_MAX; 1602 else 1603 CP(u.shmid_ds, u32.shmid_ds32, shm_segsz); 1604 CP(u.shmid_ds, u32.shmid_ds32, shm_lpid); 1605 CP(u.shmid_ds, u32.shmid_ds32, shm_cpid); 1606 CP(u.shmid_ds, u32.shmid_ds32, shm_nattch); 1607 CP(u.shmid_ds, u32.shmid_ds32, shm_atime); 1608 CP(u.shmid_ds, u32.shmid_ds32, shm_dtime); 1609 CP(u.shmid_ds, u32.shmid_ds32, shm_ctime); 1610 error = copyout(&u32.shmid_ds32, uap->buf, 1611 sizeof(u32.shmid_ds32)); 1612 break; 1613 } 1614 1615 done: 1616 if (error) { 1617 /* Invalidate the return value */ 1618 td->td_retval[0] = -1; 1619 } 1620 return (error); 1621 } 1622 #endif 1623 1624 #if defined(COMPAT_FREEBSD4) || defined(COMPAT_FREEBSD5) || \ 1625 defined(COMPAT_FREEBSD6) || defined(COMPAT_FREEBSD7) 1626 1627 #ifndef _SYS_SYSPROTO_H_ 1628 struct freebsd7_shmctl_args { 1629 int shmid; 1630 int cmd; 1631 struct shmid_ds_old *buf; 1632 }; 1633 #endif 1634 int 1635 freebsd7_shmctl(struct thread *td, struct freebsd7_shmctl_args *uap) 1636 { 1637 int error; 1638 struct shmid_ds_old old; 1639 struct shmid_ds buf; 1640 size_t bufsz; 1641 1642 /* 1643 * The only reason IPC_INFO, SHM_INFO, SHM_STAT exists is to support 1644 * Linux binaries. If we see the call come through the FreeBSD ABI, 1645 * return an error back to the user since we do not to support this. 1646 */ 1647 if (uap->cmd == IPC_INFO || uap->cmd == SHM_INFO || 1648 uap->cmd == SHM_STAT) 1649 return (EINVAL); 1650 1651 /* IPC_SET needs to copyin the buffer before calling kern_shmctl */ 1652 if (uap->cmd == IPC_SET) { 1653 if ((error = copyin(uap->buf, &old, sizeof(old)))) 1654 goto done; 1655 ipcperm_old2new(&old.shm_perm, &buf.shm_perm); 1656 CP(old, buf, shm_segsz); 1657 CP(old, buf, shm_lpid); 1658 CP(old, buf, shm_cpid); 1659 CP(old, buf, shm_nattch); 1660 CP(old, buf, shm_atime); 1661 CP(old, buf, shm_dtime); 1662 CP(old, buf, shm_ctime); 1663 } 1664 1665 error = kern_shmctl(td, uap->shmid, uap->cmd, (void *)&buf, &bufsz); 1666 if (error) 1667 goto done; 1668 1669 /* Cases in which we need to copyout */ 1670 switch (uap->cmd) { 1671 case IPC_STAT: 1672 memset(&old, 0, sizeof(old)); 1673 ipcperm_new2old(&buf.shm_perm, &old.shm_perm); 1674 if (buf.shm_segsz > INT_MAX) 1675 old.shm_segsz = INT_MAX; 1676 else 1677 CP(buf, old, shm_segsz); 1678 CP(buf, old, shm_lpid); 1679 CP(buf, old, shm_cpid); 1680 if (buf.shm_nattch > SHRT_MAX) 1681 old.shm_nattch = SHRT_MAX; 1682 else 1683 CP(buf, old, shm_nattch); 1684 CP(buf, old, shm_atime); 1685 CP(buf, old, shm_dtime); 1686 CP(buf, old, shm_ctime); 1687 old.shm_internal = NULL; 1688 error = copyout(&old, uap->buf, sizeof(old)); 1689 break; 1690 } 1691 1692 done: 1693 if (error) { 1694 /* Invalidate the return value */ 1695 td->td_retval[0] = -1; 1696 } 1697 return (error); 1698 } 1699 1700 #endif /* COMPAT_FREEBSD4 || COMPAT_FREEBSD5 || COMPAT_FREEBSD6 || 1701 COMPAT_FREEBSD7 */ 1702 1703 static int 1704 sysvshm_modload(struct module *module, int cmd, void *arg) 1705 { 1706 int error = 0; 1707 1708 switch (cmd) { 1709 case MOD_LOAD: 1710 error = shminit(); 1711 if (error != 0) 1712 shmunload(); 1713 break; 1714 case MOD_UNLOAD: 1715 error = shmunload(); 1716 break; 1717 case MOD_SHUTDOWN: 1718 break; 1719 default: 1720 error = EINVAL; 1721 break; 1722 } 1723 return (error); 1724 } 1725 1726 static moduledata_t sysvshm_mod = { 1727 "sysvshm", 1728 &sysvshm_modload, 1729 NULL 1730 }; 1731 1732 DECLARE_MODULE(sysvshm, sysvshm_mod, SI_SUB_SYSV_SHM, SI_ORDER_FIRST); 1733 MODULE_VERSION(sysvshm, 1); 1734