| 1 | /* |
| 2 | * Copyright (c) 2000-2007 Apple Inc. All rights reserved. |
| 3 | * |
| 4 | * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ |
| 5 | * |
| 6 | * This file contains Original Code and/or Modifications of Original Code |
| 7 | * as defined in and that are subject to the Apple Public Source License |
| 8 | * Version 2.0 (the 'License'). You may not use this file except in |
| 9 | * compliance with the License. The rights granted to you under the License |
| 10 | * may not be used to create, or enable the creation or redistribution of, |
| 11 | * unlawful or unlicensed copies of an Apple operating system, or to |
| 12 | * circumvent, violate, or enable the circumvention or violation of, any |
| 13 | * terms of an Apple operating system software license agreement. |
| 14 | * |
| 15 | * Please obtain a copy of the License at |
| 16 | * http://www.opensource.apple.com/apsl/ and read it before using this file. |
| 17 | * |
| 18 | * The Original Code and all software distributed under the License are |
| 19 | * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER |
| 20 | * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, |
| 21 | * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, |
| 22 | * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. |
| 23 | * Please see the License for the specific language governing rights and |
| 24 | * limitations under the License. |
| 25 | * |
| 26 | * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ |
| 27 | */ |
| 28 | /* $NetBSD: sysv_shm.c,v 1.23 1994/07/04 23:25:12 glass Exp $ */ |
| 29 | |
| 30 | /* |
| 31 | * Copyright (c) 1994 Adam Glass and Charles Hannum. All rights reserved. |
| 32 | * |
| 33 | * Redistribution and use in source and binary forms, with or without |
| 34 | * modification, are permitted provided that the following conditions |
| 35 | * are met: |
| 36 | * 1. Redistributions of source code must retain the above copyright |
| 37 | * notice, this list of conditions and the following disclaimer. |
| 38 | * 2. Redistributions in binary form must reproduce the above copyright |
| 39 | * notice, this list of conditions and the following disclaimer in the |
| 40 | * documentation and/or other materials provided with the distribution. |
| 41 | * 3. All advertising materials mentioning features or use of this software |
| 42 | * must display the following acknowledgement: |
| 43 | * This product includes software developed by Adam Glass and Charles |
| 44 | * Hannum. |
| 45 | * 4. The names of the authors may not be used to endorse or promote products |
| 46 | * derived from this software without specific prior written permission. |
| 47 | * |
| 48 | * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR |
| 49 | * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES |
| 50 | * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. |
| 51 | * IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT, INDIRECT, |
| 52 | * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT |
| 53 | * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
| 54 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
| 55 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| 56 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF |
| 57 | * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 58 | */ |
| 59 | /* |
| 60 | * NOTICE: This file was modified by McAfee Research in 2004 to introduce |
| 61 | * support for mandatory and extensible security protections. This notice |
| 62 | * is included in support of clause 2.2 (b) of the Apple Public License, |
| 63 | * Version 2.0. |
| 64 | * Copyright (c) 2005-2006 SPARTA, Inc. |
| 65 | */ |
| 66 | |
| 67 | |
| 68 | #include <sys/appleapiopts.h> |
| 69 | #include <sys/param.h> |
| 70 | #include <sys/systm.h> |
| 71 | #include <sys/kernel.h> |
| 72 | #include <sys/shm_internal.h> |
| 73 | #include <sys/proc_internal.h> |
| 74 | #include <sys/kauth.h> |
| 75 | #include <sys/malloc.h> |
| 76 | #include <sys/mman.h> |
| 77 | #include <sys/stat.h> |
| 78 | #include <sys/sysctl.h> |
| 79 | #include <sys/ipcs.h> |
| 80 | #include <sys/sysent.h> |
| 81 | #include <sys/sysproto.h> |
| 82 | #if CONFIG_MACF |
| 83 | #include <security/mac_framework.h> |
| 84 | #endif |
| 85 | |
| 86 | #include <security/audit/audit.h> |
| 87 | |
| 88 | #include <mach/mach_types.h> |
| 89 | #include <mach/vm_inherit.h> |
| 90 | #include <mach/vm_map.h> |
| 91 | |
| 92 | #include <mach/mach_vm.h> |
| 93 | |
| 94 | #include <vm/vm_map.h> |
| 95 | #include <vm/vm_protos.h> |
| 96 | #include <vm/vm_kern.h> |
| 97 | |
| 98 | #include <kern/locks.h> |
| 99 | #include <os/overflow.h> |
| 100 | |
| 101 | /* Uncomment this line to see MAC debugging output. */ |
| 102 | /* #define MAC_DEBUG */ |
| 103 | #if CONFIG_MACF_DEBUG |
| 104 | #define MPRINTF(a) printf a |
| 105 | #else |
| 106 | #define MPRINTF(a) |
| 107 | #endif |
| 108 | |
| 109 | #if SYSV_SHM |
| 110 | static int shminit(void); |
| 111 | |
| 112 | static lck_grp_t *sysv_shm_subsys_lck_grp; |
| 113 | static lck_grp_attr_t *sysv_shm_subsys_lck_grp_attr; |
| 114 | static lck_attr_t *sysv_shm_subsys_lck_attr; |
| 115 | static lck_mtx_t sysv_shm_subsys_mutex; |
| 116 | |
| 117 | #define SYSV_SHM_SUBSYS_LOCK() lck_mtx_lock(&sysv_shm_subsys_mutex) |
| 118 | #define SYSV_SHM_SUBSYS_UNLOCK() lck_mtx_unlock(&sysv_shm_subsys_mutex) |
| 119 | |
| 120 | static int oshmctl(void *p, void *uap, void *retval); |
| 121 | static int shmget_allocate_segment(struct proc *p, struct shmget_args *uap, int mode, int * retval); |
| 122 | static int shmget_existing(struct shmget_args *uap, int mode, int segnum, int * retval); |
| 123 | static void shmid_ds_64to32(struct user_shmid_ds *in, struct user32_shmid_ds *out); |
| 124 | static void shmid_ds_32to64(struct user32_shmid_ds *in, struct user_shmid_ds *out); |
| 125 | |
| 126 | /* XXX casting to (sy_call_t *) is bogus, as usual. */ |
| 127 | static sy_call_t *shmcalls[] = { |
| 128 | (sy_call_t *)shmat, (sy_call_t *)oshmctl, |
| 129 | (sy_call_t *)shmdt, (sy_call_t *)shmget, |
| 130 | (sy_call_t *)shmctl |
| 131 | }; |
| 132 | |
| 133 | #define SHMSEG_FREE 0x0200 |
| 134 | #define SHMSEG_REMOVED 0x0400 |
| 135 | #define SHMSEG_ALLOCATED 0x0800 |
| 136 | #define SHMSEG_WANTED 0x1000 |
| 137 | |
| 138 | static int shm_last_free, shm_nused, shm_committed; |
| 139 | struct shmid_kernel *shmsegs; /* 64 bit version */ |
| 140 | static int shm_inited = 0; |
| 141 | |
| 142 | /* |
| 143 | * Since anonymous memory chunks are limited to ANON_MAX_SIZE bytes, |
| 144 | * we have to keep a list of chunks when we want to handle a shared memory |
| 145 | * segment bigger than ANON_MAX_SIZE. |
| 146 | * Each chunk points to a VM named entry of up to ANON_MAX_SIZE bytes |
| 147 | * of anonymous memory. |
| 148 | */ |
| 149 | struct shm_handle { |
| 150 | void * shm_object; /* named entry for this chunk*/ |
| 151 | memory_object_size_t shm_handle_size; /* size of this chunk */ |
| 152 | struct shm_handle *shm_handle_next; /* next chunk */ |
| 153 | }; |
| 154 | |
| 155 | struct shmmap_state { |
| 156 | mach_vm_address_t va; /* user address */ |
| 157 | int shmid; /* segment id */ |
| 158 | }; |
| 159 | |
| 160 | static void shm_deallocate_segment(struct shmid_kernel *); |
| 161 | static int shm_find_segment_by_key(key_t); |
| 162 | static struct shmid_kernel *shm_find_segment_by_shmid(int); |
| 163 | static int shm_delete_mapping(struct proc *, struct shmmap_state *, int); |
| 164 | |
| 165 | #ifdef __APPLE_API_PRIVATE |
| 166 | #define DEFAULT_SHMMAX (4 * 1024 * 1024) |
| 167 | #define DEFAULT_SHMMIN 1 |
| 168 | #define DEFAULT_SHMMNI 32 |
| 169 | #define DEFAULT_SHMSEG 8 |
| 170 | #define DEFAULT_SHMALL 1024 |
| 171 | |
| 172 | struct shminfo shminfo = { |
| 173 | DEFAULT_SHMMAX, |
| 174 | DEFAULT_SHMMIN, |
| 175 | DEFAULT_SHMMNI, |
| 176 | DEFAULT_SHMSEG, |
| 177 | DEFAULT_SHMALL |
| 178 | }; |
| 179 | |
| 180 | #define SHMID_IS_VALID(x) ((x) >= 0) |
| 181 | #define SHMID_UNALLOCATED (-1) |
| 182 | #define SHMID_SENTINEL (-2) |
| 183 | |
| 184 | #endif /* __APPLE_API_PRIVATE */ |
| 185 | |
| 186 | void sysv_shm_lock_init(void); |
| 187 | |
| 188 | static __inline__ time_t |
| 189 | sysv_shmtime(void) |
| 190 | { |
| 191 | struct timeval tv; |
| 192 | microtime(&tv); |
| 193 | return (tv.tv_sec); |
| 194 | } |
| 195 | |
| 196 | /* |
| 197 | * This conversion is safe, since if we are converting for a 32 bit process, |
| 198 | * then it's value of (struct shmid_ds)->shm_segsz will never exceed 4G. |
| 199 | * |
| 200 | * NOTE: Source and target may *NOT* overlap! (target is smaller) |
| 201 | */ |
| 202 | static void |
| 203 | shmid_ds_64to32(struct user_shmid_ds *in, struct user32_shmid_ds *out) |
| 204 | { |
| 205 | out->shm_perm = in->shm_perm; |
| 206 | out->shm_segsz = in->shm_segsz; |
| 207 | out->shm_lpid = in->shm_lpid; |
| 208 | out->shm_cpid = in->shm_cpid; |
| 209 | out->shm_nattch = in->shm_nattch; |
| 210 | out->shm_atime = in->shm_atime; |
| 211 | out->shm_dtime = in->shm_dtime; |
| 212 | out->shm_ctime = in->shm_ctime; |
| 213 | out->shm_internal = CAST_DOWN_EXPLICIT(int,in->shm_internal); |
| 214 | } |
| 215 | |
| 216 | /* |
| 217 | * NOTE: Source and target may are permitted to overlap! (source is smaller); |
| 218 | * this works because we copy fields in order from the end of the struct to |
| 219 | * the beginning. |
| 220 | */ |
| 221 | static void |
| 222 | shmid_ds_32to64(struct user32_shmid_ds *in, struct user_shmid_ds *out) |
| 223 | { |
| 224 | out->shm_internal = in->shm_internal; |
| 225 | out->shm_ctime = in->shm_ctime; |
| 226 | out->shm_dtime = in->shm_dtime; |
| 227 | out->shm_atime = in->shm_atime; |
| 228 | out->shm_nattch = in->shm_nattch; |
| 229 | out->shm_cpid = in->shm_cpid; |
| 230 | out->shm_lpid = in->shm_lpid; |
| 231 | out->shm_segsz = in->shm_segsz; |
| 232 | out->shm_perm = in->shm_perm; |
| 233 | } |
| 234 | |
| 235 | |
| 236 | static int |
| 237 | shm_find_segment_by_key(key_t key) |
| 238 | { |
| 239 | int i; |
| 240 | |
| 241 | for (i = 0; i < shminfo.shmmni; i++) |
| 242 | if ((shmsegs[i].u.shm_perm.mode & SHMSEG_ALLOCATED) && |
| 243 | shmsegs[i].u.shm_perm._key == key) |
| 244 | return i; |
| 245 | return -1; |
| 246 | } |
| 247 | |
| 248 | static struct shmid_kernel * |
| 249 | shm_find_segment_by_shmid(int shmid) |
| 250 | { |
| 251 | int segnum; |
| 252 | struct shmid_kernel *shmseg; |
| 253 | |
| 254 | segnum = IPCID_TO_IX(shmid); |
| 255 | if (segnum < 0 || segnum >= shminfo.shmmni) |
| 256 | return NULL; |
| 257 | shmseg = &shmsegs[segnum]; |
| 258 | if ((shmseg->u.shm_perm.mode & (SHMSEG_ALLOCATED | SHMSEG_REMOVED)) |
| 259 | != SHMSEG_ALLOCATED || |
| 260 | shmseg->u.shm_perm._seq != IPCID_TO_SEQ(shmid)) |
| 261 | return NULL; |
| 262 | return shmseg; |
| 263 | } |
| 264 | |
| 265 | static void |
| 266 | shm_deallocate_segment(struct shmid_kernel *shmseg) |
| 267 | { |
| 268 | struct shm_handle *shm_handle, *shm_handle_next; |
| 269 | mach_vm_size_t size; |
| 270 | |
| 271 | for (shm_handle = CAST_DOWN(void *,shmseg->u.shm_internal); /* tunnel */ |
| 272 | shm_handle != NULL; |
| 273 | shm_handle = shm_handle_next) { |
| 274 | shm_handle_next = shm_handle->shm_handle_next; |
| 275 | mach_memory_entry_port_release(shm_handle->shm_object); |
| 276 | FREE((caddr_t) shm_handle, M_SHM); |
| 277 | } |
| 278 | shmseg->u.shm_internal = USER_ADDR_NULL; /* tunnel */ |
| 279 | size = mach_vm_round_page(shmseg->u.shm_segsz); |
| 280 | shm_committed -= btoc(size); |
| 281 | shm_nused--; |
| 282 | shmseg->u.shm_perm.mode = SHMSEG_FREE; |
| 283 | #if CONFIG_MACF |
| 284 | /* Reset the MAC label */ |
| 285 | mac_sysvshm_label_recycle(shmseg); |
| 286 | #endif |
| 287 | } |
| 288 | |
| 289 | static int |
| 290 | shm_delete_mapping(__unused struct proc *p, struct shmmap_state *shmmap_s, |
| 291 | int deallocate) |
| 292 | { |
| 293 | struct shmid_kernel *shmseg; |
| 294 | int segnum, result; |
| 295 | mach_vm_size_t size; |
| 296 | |
| 297 | segnum = IPCID_TO_IX(shmmap_s->shmid); |
| 298 | shmseg = &shmsegs[segnum]; |
| 299 | size = mach_vm_round_page(shmseg->u.shm_segsz); /* XXX done for us? */ |
| 300 | if (deallocate) { |
| 301 | result = mach_vm_deallocate(current_map(), shmmap_s->va, size); |
| 302 | if (result != KERN_SUCCESS) |
| 303 | return EINVAL; |
| 304 | } |
| 305 | shmmap_s->shmid = SHMID_UNALLOCATED; |
| 306 | shmseg->u.shm_dtime = sysv_shmtime(); |
| 307 | if ((--shmseg->u.shm_nattch <= 0) && |
| 308 | (shmseg->u.shm_perm.mode & SHMSEG_REMOVED)) { |
| 309 | shm_deallocate_segment(shmseg); |
| 310 | shm_last_free = segnum; |
| 311 | } |
| 312 | return 0; |
| 313 | } |
| 314 | |
| 315 | int |
| 316 | shmdt(struct proc *p, struct shmdt_args *uap, int32_t *retval) |
| 317 | { |
| 318 | #if CONFIG_MACF |
| 319 | struct shmid_kernel *shmsegptr; |
| 320 | #endif |
| 321 | struct shmmap_state *shmmap_s; |
| 322 | int i; |
| 323 | int shmdtret = 0; |
| 324 | |
| 325 | AUDIT_ARG(svipc_addr, uap->shmaddr); |
| 326 | |
| 327 | SYSV_SHM_SUBSYS_LOCK(); |
| 328 | |
| 329 | if ((shmdtret = shminit())) { |
| 330 | goto shmdt_out; |
| 331 | } |
| 332 | |
| 333 | shmmap_s = (struct shmmap_state *)p->vm_shm; |
| 334 | if (shmmap_s == NULL) { |
| 335 | shmdtret = EINVAL; |
| 336 | goto shmdt_out; |
| 337 | } |
| 338 | |
| 339 | for (; shmmap_s->shmid != SHMID_SENTINEL; shmmap_s++) { |
| 340 | if (SHMID_IS_VALID(shmmap_s->shmid) && |
| 341 | shmmap_s->va == (mach_vm_offset_t)uap->shmaddr) { |
| 342 | break; |
| 343 | } |
| 344 | } |
| 345 | |
| 346 | if (!SHMID_IS_VALID(shmmap_s->shmid)) { |
| 347 | shmdtret = EINVAL; |
| 348 | goto shmdt_out; |
| 349 | } |
| 350 | |
| 351 | #if CONFIG_MACF |
| 352 | /* |
| 353 | * XXX: It might be useful to move this into the shm_delete_mapping |
| 354 | * function |
| 355 | */ |
| 356 | shmsegptr = &shmsegs[IPCID_TO_IX(shmmap_s->shmid)]; |
| 357 | shmdtret = mac_sysvshm_check_shmdt(kauth_cred_get(), shmsegptr); |
| 358 | if (shmdtret) |
| 359 | goto shmdt_out; |
| 360 | #endif |
| 361 | i = shm_delete_mapping(p, shmmap_s, 1); |
| 362 | |
| 363 | if (i == 0) |
| 364 | *retval = 0; |
| 365 | shmdtret = i; |
| 366 | shmdt_out: |
| 367 | SYSV_SHM_SUBSYS_UNLOCK(); |
| 368 | return shmdtret; |
| 369 | } |
| 370 | |
| 371 | int |
| 372 | shmat(struct proc *p, struct shmat_args *uap, user_addr_t *retval) |
| 373 | { |
| 374 | int error, i, flags; |
| 375 | struct shmid_kernel *shmseg; |
| 376 | struct shmmap_state *shmmap_s = NULL; |
| 377 | struct shm_handle *shm_handle; |
| 378 | mach_vm_address_t attach_va; /* attach address in/out */ |
| 379 | mach_vm_size_t map_size; /* size of map entry */ |
| 380 | mach_vm_size_t mapped_size; |
| 381 | vm_prot_t prot; |
| 382 | size_t size; |
| 383 | kern_return_t rv; |
| 384 | int shmat_ret; |
| 385 | int vm_flags; |
| 386 | |
| 387 | shmat_ret = 0; |
| 388 | |
| 389 | AUDIT_ARG(svipc_id, uap->shmid); |
| 390 | AUDIT_ARG(svipc_addr, uap->shmaddr); |
| 391 | |
| 392 | SYSV_SHM_SUBSYS_LOCK(); |
| 393 | |
| 394 | if ((shmat_ret = shminit())) { |
| 395 | goto shmat_out; |
| 396 | } |
| 397 | |
| 398 | shmmap_s = (struct shmmap_state *)p->vm_shm; |
| 399 | if (shmmap_s == NULL) { |
| 400 | /* lazily allocate the shm map */ |
| 401 | |
| 402 | int nsegs = shminfo.shmseg; |
| 403 | if (nsegs <= 0) { |
| 404 | shmat_ret = EMFILE; |
| 405 | goto shmat_out; |
| 406 | } |
| 407 | |
| 408 | /* +1 for the sentinel */ |
| 409 | if (os_add_and_mul_overflow(nsegs, 1, sizeof(struct shmmap_state), &size)) { |
| 410 | shmat_ret = ENOMEM; |
| 411 | goto shmat_out; |
| 412 | } |
| 413 | |
| 414 | MALLOC(shmmap_s, struct shmmap_state *, size, M_SHM, M_WAITOK | M_NULL); |
| 415 | if (shmmap_s == NULL) { |
| 416 | shmat_ret = ENOMEM; |
| 417 | goto shmat_out; |
| 418 | } |
| 419 | |
| 420 | /* initialize the entries */ |
| 421 | for (i = 0; i < nsegs; i++) { |
| 422 | shmmap_s[i].shmid = SHMID_UNALLOCATED; |
| 423 | } |
| 424 | shmmap_s[i].shmid = SHMID_SENTINEL; |
| 425 | |
| 426 | p->vm_shm = (caddr_t)shmmap_s; |
| 427 | } |
| 428 | |
| 429 | shmseg = shm_find_segment_by_shmid(uap->shmid); |
| 430 | if (shmseg == NULL) { |
| 431 | shmat_ret = EINVAL; |
| 432 | goto shmat_out; |
| 433 | } |
| 434 | |
| 435 | AUDIT_ARG(svipc_perm, &shmseg->u.shm_perm); |
| 436 | error = ipcperm(kauth_cred_get(), &shmseg->u.shm_perm, |
| 437 | (uap->shmflg & SHM_RDONLY) ? IPC_R : IPC_R|IPC_W); |
| 438 | if (error) { |
| 439 | shmat_ret = error; |
| 440 | goto shmat_out; |
| 441 | } |
| 442 | |
| 443 | #if CONFIG_MACF |
| 444 | error = mac_sysvshm_check_shmat(kauth_cred_get(), shmseg, uap->shmflg); |
| 445 | if (error) { |
| 446 | shmat_ret = error; |
| 447 | goto shmat_out; |
| 448 | } |
| 449 | #endif |
| 450 | |
| 451 | /* find a free shmid */ |
| 452 | while (SHMID_IS_VALID(shmmap_s->shmid)) { |
| 453 | shmmap_s++; |
| 454 | } |
| 455 | if (shmmap_s->shmid != SHMID_UNALLOCATED) { |
| 456 | /* no free shmids */ |
| 457 | shmat_ret = EMFILE; |
| 458 | goto shmat_out; |
| 459 | } |
| 460 | |
| 461 | map_size = mach_vm_round_page(shmseg->u.shm_segsz); |
| 462 | prot = VM_PROT_READ; |
| 463 | if ((uap->shmflg & SHM_RDONLY) == 0) |
| 464 | prot |= VM_PROT_WRITE; |
| 465 | flags = MAP_ANON | MAP_SHARED; |
| 466 | if (uap->shmaddr) |
| 467 | flags |= MAP_FIXED; |
| 468 | |
| 469 | attach_va = (mach_vm_address_t)uap->shmaddr; |
| 470 | if (uap->shmflg & SHM_RND) |
| 471 | attach_va &= ~(SHMLBA-1); |
| 472 | else if ((attach_va & (SHMLBA-1)) != 0) { |
| 473 | shmat_ret = EINVAL; |
| 474 | goto shmat_out; |
| 475 | } |
| 476 | |
| 477 | if (flags & MAP_FIXED) { |
| 478 | vm_flags = VM_FLAGS_FIXED; |
| 479 | } else { |
| 480 | vm_flags = VM_FLAGS_ANYWHERE; |
| 481 | } |
| 482 | |
| 483 | mapped_size = 0; |
| 484 | |
| 485 | /* first reserve enough space... */ |
| 486 | rv = mach_vm_map_kernel(current_map(), |
| 487 | &attach_va, |
| 488 | map_size, |
| 489 | 0, |
| 490 | vm_flags, |
| 491 | VM_MAP_KERNEL_FLAGS_NONE, |
| 492 | VM_KERN_MEMORY_NONE, |
| 493 | IPC_PORT_NULL, |
| 494 | 0, |
| 495 | FALSE, |
| 496 | VM_PROT_NONE, |
| 497 | VM_PROT_NONE, |
| 498 | VM_INHERIT_NONE); |
| 499 | if (rv != KERN_SUCCESS) { |
| 500 | goto out; |
| 501 | } |
| 502 | |
| 503 | shmmap_s->va = attach_va; |
| 504 | |
| 505 | /* ... then map the shared memory over the reserved space */ |
| 506 | for (shm_handle = CAST_DOWN(void *, shmseg->u.shm_internal);/* tunnel */ |
| 507 | shm_handle != NULL; |
| 508 | shm_handle = shm_handle->shm_handle_next) { |
| 509 | |
| 510 | rv = vm_map_enter_mem_object( |
| 511 | current_map(), /* process map */ |
| 512 | &attach_va, /* attach address */ |
| 513 | shm_handle->shm_handle_size, /* segment size */ |
| 514 | (mach_vm_offset_t)0, /* alignment mask */ |
| 515 | VM_FLAGS_FIXED | VM_FLAGS_OVERWRITE, |
| 516 | VM_MAP_KERNEL_FLAGS_NONE, |
| 517 | VM_KERN_MEMORY_NONE, |
| 518 | shm_handle->shm_object, |
| 519 | (mach_vm_offset_t)0, |
| 520 | FALSE, |
| 521 | prot, |
| 522 | prot, |
| 523 | VM_INHERIT_SHARE); |
| 524 | if (rv != KERN_SUCCESS) |
| 525 | goto out; |
| 526 | |
| 527 | mapped_size += shm_handle->shm_handle_size; |
| 528 | attach_va = attach_va + shm_handle->shm_handle_size; |
| 529 | } |
| 530 | |
| 531 | shmmap_s->shmid = uap->shmid; |
| 532 | shmseg->u.shm_lpid = p->p_pid; |
| 533 | shmseg->u.shm_atime = sysv_shmtime(); |
| 534 | shmseg->u.shm_nattch++; |
| 535 | *retval = shmmap_s->va; /* XXX return -1 on error */ |
| 536 | shmat_ret = 0; |
| 537 | goto shmat_out; |
| 538 | out: |
| 539 | if (mapped_size > 0) { |
| 540 | (void) mach_vm_deallocate(current_map(), |
| 541 | shmmap_s->va, |
| 542 | mapped_size); |
| 543 | } |
| 544 | switch (rv) { |
| 545 | case KERN_INVALID_ADDRESS: |
| 546 | case KERN_NO_SPACE: |
| 547 | shmat_ret = ENOMEM; |
| 548 | break; |
| 549 | case KERN_PROTECTION_FAILURE: |
| 550 | shmat_ret = EACCES; |
| 551 | break; |
| 552 | default: |
| 553 | shmat_ret = EINVAL; |
| 554 | break; |
| 555 | } |
| 556 | shmat_out: |
| 557 | SYSV_SHM_SUBSYS_UNLOCK(); |
| 558 | return shmat_ret; |
| 559 | } |
| 560 | |
| 561 | static int |
| 562 | oshmctl(__unused void *p, __unused void *uap, __unused void *retval) |
| 563 | { |
| 564 | return EINVAL; |
| 565 | } |
| 566 | |
| 567 | /* |
| 568 | * Returns: 0 Success |
| 569 | * EINVAL |
| 570 | * copyout:EFAULT |
| 571 | * copyin:EFAULT |
| 572 | * ipcperm:EPERM |
| 573 | * ipcperm:EACCES |
| 574 | */ |
| 575 | int |
| 576 | shmctl(__unused struct proc *p, struct shmctl_args *uap, int32_t *retval) |
| 577 | { |
| 578 | int error; |
| 579 | kauth_cred_t cred = kauth_cred_get(); |
| 580 | struct user_shmid_ds inbuf; |
| 581 | struct shmid_kernel *shmseg; |
| 582 | |
| 583 | int shmctl_ret = 0; |
| 584 | |
| 585 | AUDIT_ARG(svipc_cmd, uap->cmd); |
| 586 | AUDIT_ARG(svipc_id, uap->shmid); |
| 587 | |
| 588 | SYSV_SHM_SUBSYS_LOCK(); |
| 589 | |
| 590 | if ((shmctl_ret = shminit())) { |
| 591 | goto shmctl_out; |
| 592 | } |
| 593 | |
| 594 | shmseg = shm_find_segment_by_shmid(uap->shmid); |
| 595 | if (shmseg == NULL) { |
| 596 | shmctl_ret = EINVAL; |
| 597 | goto shmctl_out; |
| 598 | } |
| 599 | |
| 600 | /* XXAUDIT: This is the perms BEFORE any change by this call. This |
| 601 | * may not be what is desired. |
| 602 | */ |
| 603 | AUDIT_ARG(svipc_perm, &shmseg->u.shm_perm); |
| 604 | |
| 605 | #if CONFIG_MACF |
| 606 | error = mac_sysvshm_check_shmctl(cred, shmseg, uap->cmd); |
| 607 | if (error) { |
| 608 | shmctl_ret = error; |
| 609 | goto shmctl_out; |
| 610 | } |
| 611 | #endif |
| 612 | switch (uap->cmd) { |
| 613 | case IPC_STAT: |
| 614 | error = ipcperm(cred, &shmseg->u.shm_perm, IPC_R); |
| 615 | if (error) { |
| 616 | shmctl_ret = error; |
| 617 | goto shmctl_out; |
| 618 | } |
| 619 | |
| 620 | if (IS_64BIT_PROCESS(p)) { |
| 621 | struct user_shmid_ds shmid_ds; |
| 622 | memcpy(&shmid_ds, &shmseg->u, sizeof(struct user_shmid_ds)); |
| 623 | |
| 624 | /* Clear kernel reserved pointer before copying to user space */ |
| 625 | shmid_ds.shm_internal = USER_ADDR_NULL; |
| 626 | |
| 627 | error = copyout(&shmid_ds, uap->buf, sizeof(shmid_ds)); |
| 628 | } else { |
| 629 | struct user32_shmid_ds shmid_ds32 = {}; |
| 630 | shmid_ds_64to32(&shmseg->u, &shmid_ds32); |
| 631 | |
| 632 | /* Clear kernel reserved pointer before copying to user space */ |
| 633 | shmid_ds32.shm_internal = (user32_addr_t)0; |
| 634 | |
| 635 | error = copyout(&shmid_ds32, uap->buf, sizeof(shmid_ds32)); |
| 636 | } |
| 637 | if (error) { |
| 638 | shmctl_ret = error; |
| 639 | goto shmctl_out; |
| 640 | } |
| 641 | break; |
| 642 | case IPC_SET: |
| 643 | error = ipcperm(cred, &shmseg->u.shm_perm, IPC_M); |
| 644 | if (error) { |
| 645 | shmctl_ret = error; |
| 646 | goto shmctl_out; |
| 647 | } |
| 648 | if (IS_64BIT_PROCESS(p)) { |
| 649 | error = copyin(uap->buf, &inbuf, sizeof(struct user_shmid_ds)); |
| 650 | } else { |
| 651 | struct user32_shmid_ds shmid_ds32; |
| 652 | error = copyin(uap->buf, &shmid_ds32, sizeof(shmid_ds32)); |
| 653 | /* convert in place; ugly, but safe */ |
| 654 | shmid_ds_32to64(&shmid_ds32, &inbuf); |
| 655 | } |
| 656 | if (error) { |
| 657 | shmctl_ret = error; |
| 658 | goto shmctl_out; |
| 659 | } |
| 660 | shmseg->u.shm_perm.uid = inbuf.shm_perm.uid; |
| 661 | shmseg->u.shm_perm.gid = inbuf.shm_perm.gid; |
| 662 | shmseg->u.shm_perm.mode = |
| 663 | (shmseg->u.shm_perm.mode & ~ACCESSPERMS) | |
| 664 | (inbuf.shm_perm.mode & ACCESSPERMS); |
| 665 | shmseg->u.shm_ctime = sysv_shmtime(); |
| 666 | break; |
| 667 | case IPC_RMID: |
| 668 | error = ipcperm(cred, &shmseg->u.shm_perm, IPC_M); |
| 669 | if (error) { |
| 670 | shmctl_ret = error; |
| 671 | goto shmctl_out; |
| 672 | } |
| 673 | shmseg->u.shm_perm._key = IPC_PRIVATE; |
| 674 | shmseg->u.shm_perm.mode |= SHMSEG_REMOVED; |
| 675 | if (shmseg->u.shm_nattch <= 0) { |
| 676 | shm_deallocate_segment(shmseg); |
| 677 | shm_last_free = IPCID_TO_IX(uap->shmid); |
| 678 | } |
| 679 | break; |
| 680 | #if 0 |
| 681 | case SHM_LOCK: |
| 682 | case SHM_UNLOCK: |
| 683 | #endif |
| 684 | default: |
| 685 | shmctl_ret = EINVAL; |
| 686 | goto shmctl_out; |
| 687 | } |
| 688 | *retval = 0; |
| 689 | shmctl_ret = 0; |
| 690 | shmctl_out: |
| 691 | SYSV_SHM_SUBSYS_UNLOCK(); |
| 692 | return shmctl_ret; |
| 693 | } |
| 694 | |
| 695 | static int |
| 696 | shmget_existing(struct shmget_args *uap, int mode, int segnum, int *retval) |
| 697 | { |
| 698 | struct shmid_kernel *shmseg; |
| 699 | int error = 0; |
| 700 | |
| 701 | shmseg = &shmsegs[segnum]; |
| 702 | if (shmseg->u.shm_perm.mode & SHMSEG_REMOVED) { |
| 703 | /* |
| 704 | * This segment is in the process of being allocated. Wait |
| 705 | * until it's done, and look the key up again (in case the |
| 706 | * allocation failed or it was freed). |
| 707 | */ |
| 708 | shmseg->u.shm_perm.mode |= SHMSEG_WANTED; |
| 709 | error = tsleep((caddr_t)shmseg, PLOCK | PCATCH, "shmget" , 0); |
| 710 | if (error) |
| 711 | return error; |
| 712 | return EAGAIN; |
| 713 | } |
| 714 | |
| 715 | /* |
| 716 | * The low 9 bits of shmflag are the mode bits being requested, which |
| 717 | * are the actual mode bits desired on the segment, and not in IPC_R |
| 718 | * form; therefore it would be incorrect to call ipcperm() to validate |
| 719 | * them; instead, we AND the existing mode with the requested mode, and |
| 720 | * verify that it matches the requested mode; otherwise, we fail with |
| 721 | * EACCES (access denied). |
| 722 | */ |
| 723 | if ((shmseg->u.shm_perm.mode & mode) != mode) |
| 724 | return EACCES; |
| 725 | |
| 726 | #if CONFIG_MACF |
| 727 | error = mac_sysvshm_check_shmget(kauth_cred_get(), shmseg, uap->shmflg); |
| 728 | if (error) |
| 729 | return (error); |
| 730 | #endif |
| 731 | |
| 732 | if (uap->size && uap->size > shmseg->u.shm_segsz) |
| 733 | return EINVAL; |
| 734 | |
| 735 | if ((uap->shmflg & (IPC_CREAT | IPC_EXCL)) == (IPC_CREAT | IPC_EXCL)) |
| 736 | return EEXIST; |
| 737 | |
| 738 | *retval = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm); |
| 739 | return 0; |
| 740 | } |
| 741 | |
| 742 | static int |
| 743 | shmget_allocate_segment(struct proc *p, struct shmget_args *uap, int mode, |
| 744 | int *retval) |
| 745 | { |
| 746 | int i, segnum, shmid; |
| 747 | kauth_cred_t cred = kauth_cred_get(); |
| 748 | struct shmid_kernel *shmseg; |
| 749 | struct shm_handle *shm_handle; |
| 750 | kern_return_t kret; |
| 751 | mach_vm_size_t total_size, size, alloc_size; |
| 752 | void * mem_object; |
| 753 | struct shm_handle *shm_handle_next, **shm_handle_next_p; |
| 754 | |
| 755 | if (uap->size <= 0 || |
| 756 | uap->size < (user_size_t)shminfo.shmmin || |
| 757 | uap->size > (user_size_t)shminfo.shmmax) { |
| 758 | return EINVAL; |
| 759 | } |
| 760 | if (shm_nused >= shminfo.shmmni) /* any shmids left? */ |
| 761 | return ENOSPC; |
| 762 | if (mach_vm_round_page_overflow(uap->size, &total_size)) { |
| 763 | return EINVAL; |
| 764 | } |
| 765 | if ((user_ssize_t)(shm_committed + btoc(total_size)) > shminfo.shmall) |
| 766 | return ENOMEM; |
| 767 | if (shm_last_free < 0) { |
| 768 | for (i = 0; i < shminfo.shmmni; i++) |
| 769 | if (shmsegs[i].u.shm_perm.mode & SHMSEG_FREE) |
| 770 | break; |
| 771 | if (i == shminfo.shmmni) |
| 772 | panic("shmseg free count inconsistent" ); |
| 773 | segnum = i; |
| 774 | } else { |
| 775 | segnum = shm_last_free; |
| 776 | shm_last_free = -1; |
| 777 | } |
| 778 | shmseg = &shmsegs[segnum]; |
| 779 | |
| 780 | /* |
| 781 | * In case we sleep in malloc(), mark the segment present but deleted |
| 782 | * so that noone else tries to create the same key. |
| 783 | * XXX but we don't release the global lock !? |
| 784 | */ |
| 785 | shmseg->u.shm_perm.mode = SHMSEG_ALLOCATED | SHMSEG_REMOVED; |
| 786 | shmseg->u.shm_perm._key = uap->key; |
| 787 | shmseg->u.shm_perm._seq = (shmseg->u.shm_perm._seq + 1) & 0x7fff; |
| 788 | |
| 789 | shm_handle_next_p = NULL; |
| 790 | for (alloc_size = 0; |
| 791 | alloc_size < total_size; |
| 792 | alloc_size += size) { |
| 793 | size = MIN(total_size - alloc_size, ANON_MAX_SIZE); |
| 794 | kret = mach_make_memory_entry_64( |
| 795 | VM_MAP_NULL, |
| 796 | (memory_object_size_t *) &size, |
| 797 | (memory_object_offset_t) 0, |
| 798 | MAP_MEM_NAMED_CREATE | VM_PROT_DEFAULT, |
| 799 | (ipc_port_t *) &mem_object, 0); |
| 800 | if (kret != KERN_SUCCESS) |
| 801 | goto out; |
| 802 | |
| 803 | MALLOC(shm_handle, struct shm_handle *, sizeof(struct shm_handle), M_SHM, M_WAITOK); |
| 804 | if (shm_handle == NULL) { |
| 805 | kret = KERN_NO_SPACE; |
| 806 | mach_memory_entry_port_release(mem_object); |
| 807 | mem_object = NULL; |
| 808 | goto out; |
| 809 | } |
| 810 | shm_handle->shm_object = mem_object; |
| 811 | shm_handle->shm_handle_size = size; |
| 812 | shm_handle->shm_handle_next = NULL; |
| 813 | if (shm_handle_next_p == NULL) { |
| 814 | shmseg->u.shm_internal = CAST_USER_ADDR_T(shm_handle);/* tunnel */ |
| 815 | } else { |
| 816 | *shm_handle_next_p = shm_handle; |
| 817 | } |
| 818 | shm_handle_next_p = &shm_handle->shm_handle_next; |
| 819 | } |
| 820 | |
| 821 | shmid = IXSEQ_TO_IPCID(segnum, shmseg->u.shm_perm); |
| 822 | |
| 823 | shmseg->u.shm_perm.cuid = shmseg->u.shm_perm.uid = kauth_cred_getuid(cred); |
| 824 | shmseg->u.shm_perm.cgid = shmseg->u.shm_perm.gid = kauth_cred_getgid(cred); |
| 825 | shmseg->u.shm_perm.mode = (shmseg->u.shm_perm.mode & SHMSEG_WANTED) | |
| 826 | (mode & ACCESSPERMS) | SHMSEG_ALLOCATED; |
| 827 | shmseg->u.shm_segsz = uap->size; |
| 828 | shmseg->u.shm_cpid = p->p_pid; |
| 829 | shmseg->u.shm_lpid = shmseg->u.shm_nattch = 0; |
| 830 | shmseg->u.shm_atime = shmseg->u.shm_dtime = 0; |
| 831 | #if CONFIG_MACF |
| 832 | mac_sysvshm_label_associate(cred, shmseg); |
| 833 | #endif |
| 834 | shmseg->u.shm_ctime = sysv_shmtime(); |
| 835 | shm_committed += btoc(size); |
| 836 | shm_nused++; |
| 837 | AUDIT_ARG(svipc_perm, &shmseg->u.shm_perm); |
| 838 | if (shmseg->u.shm_perm.mode & SHMSEG_WANTED) { |
| 839 | /* |
| 840 | * Somebody else wanted this key while we were asleep. Wake |
| 841 | * them up now. |
| 842 | */ |
| 843 | shmseg->u.shm_perm.mode &= ~SHMSEG_WANTED; |
| 844 | wakeup((caddr_t)shmseg); |
| 845 | } |
| 846 | *retval = shmid; |
| 847 | AUDIT_ARG(svipc_id, shmid); |
| 848 | return 0; |
| 849 | out: |
| 850 | if (kret != KERN_SUCCESS) { |
| 851 | for (shm_handle = CAST_DOWN(void *,shmseg->u.shm_internal); /* tunnel */ |
| 852 | shm_handle != NULL; |
| 853 | shm_handle = shm_handle_next) { |
| 854 | shm_handle_next = shm_handle->shm_handle_next; |
| 855 | mach_memory_entry_port_release(shm_handle->shm_object); |
| 856 | FREE((caddr_t) shm_handle, M_SHM); |
| 857 | } |
| 858 | shmseg->u.shm_internal = USER_ADDR_NULL; /* tunnel */ |
| 859 | } |
| 860 | |
| 861 | switch (kret) { |
| 862 | case KERN_INVALID_ADDRESS: |
| 863 | case KERN_NO_SPACE: |
| 864 | return (ENOMEM); |
| 865 | case KERN_PROTECTION_FAILURE: |
| 866 | return (EACCES); |
| 867 | default: |
| 868 | return (EINVAL); |
| 869 | } |
| 870 | |
| 871 | } |
| 872 | |
| 873 | int |
| 874 | shmget(struct proc *p, struct shmget_args *uap, int32_t *retval) |
| 875 | { |
| 876 | int segnum, mode, error; |
| 877 | int shmget_ret = 0; |
| 878 | |
| 879 | /* Auditing is actually done in shmget_allocate_segment() */ |
| 880 | |
| 881 | SYSV_SHM_SUBSYS_LOCK(); |
| 882 | |
| 883 | if ((shmget_ret = shminit())) { |
| 884 | goto shmget_out; |
| 885 | } |
| 886 | |
| 887 | mode = uap->shmflg & ACCESSPERMS; |
| 888 | if (uap->key != IPC_PRIVATE) { |
| 889 | again: |
| 890 | segnum = shm_find_segment_by_key(uap->key); |
| 891 | if (segnum >= 0) { |
| 892 | error = shmget_existing(uap, mode, segnum, retval); |
| 893 | if (error == EAGAIN) |
| 894 | goto again; |
| 895 | shmget_ret = error; |
| 896 | goto shmget_out; |
| 897 | } |
| 898 | if ((uap->shmflg & IPC_CREAT) == 0) { |
| 899 | shmget_ret = ENOENT; |
| 900 | goto shmget_out; |
| 901 | } |
| 902 | } |
| 903 | shmget_ret = shmget_allocate_segment(p, uap, mode, retval); |
| 904 | shmget_out: |
| 905 | SYSV_SHM_SUBSYS_UNLOCK(); |
| 906 | return shmget_ret; |
| 907 | } |
| 908 | |
| 909 | /* |
| 910 | * shmsys |
| 911 | * |
| 912 | * Entry point for all SHM calls: shmat, oshmctl, shmdt, shmget, shmctl |
| 913 | * |
| 914 | * Parameters: p Process requesting the call |
| 915 | * uap User argument descriptor (see below) |
| 916 | * retval Return value of the selected shm call |
| 917 | * |
| 918 | * Indirect parameters: uap->which msg call to invoke (index in array of shm calls) |
| 919 | * uap->a2 User argument descriptor |
| 920 | * |
| 921 | * Returns: 0 Success |
| 922 | * !0 Not success |
| 923 | * |
| 924 | * Implicit returns: retval Return value of the selected shm call |
| 925 | * |
| 926 | * DEPRECATED: This interface should not be used to call the other SHM |
| 927 | * functions (shmat, oshmctl, shmdt, shmget, shmctl). The correct |
| 928 | * usage is to call the other SHM functions directly. |
| 929 | */ |
| 930 | int |
| 931 | shmsys(struct proc *p, struct shmsys_args *uap, int32_t *retval) |
| 932 | { |
| 933 | |
| 934 | /* The routine that we are dispatching already does this */ |
| 935 | |
| 936 | if (uap->which >= sizeof(shmcalls)/sizeof(shmcalls[0])) |
| 937 | return EINVAL; |
| 938 | return ((*shmcalls[uap->which])(p, &uap->a2, retval)); |
| 939 | } |
| 940 | |
| 941 | /* |
| 942 | * Return 0 on success, 1 on failure. |
| 943 | */ |
| 944 | int |
| 945 | shmfork(struct proc *p1, struct proc *p2) |
| 946 | { |
| 947 | struct shmmap_state *shmmap_s; |
| 948 | size_t size; |
| 949 | int nsegs = 0; |
| 950 | int ret = 0; |
| 951 | |
| 952 | SYSV_SHM_SUBSYS_LOCK(); |
| 953 | |
| 954 | if (shminit()) { |
| 955 | ret = 1; |
| 956 | goto shmfork_out; |
| 957 | } |
| 958 | |
| 959 | struct shmmap_state *src = (struct shmmap_state *)p1->vm_shm; |
| 960 | assert(src); |
| 961 | |
| 962 | /* count number of shmid entries in src */ |
| 963 | for (struct shmmap_state *s = src; s->shmid != SHMID_SENTINEL; s++) { |
| 964 | nsegs++; |
| 965 | } |
| 966 | |
| 967 | if (os_add_and_mul_overflow(nsegs, 1, sizeof(struct shmmap_state), &size)) { |
| 968 | ret = 1; |
| 969 | goto shmfork_out; |
| 970 | } |
| 971 | MALLOC(shmmap_s, struct shmmap_state *, size, M_SHM, M_WAITOK); |
| 972 | if (shmmap_s == NULL) { |
| 973 | ret = 1; |
| 974 | goto shmfork_out; |
| 975 | } |
| 976 | |
| 977 | bcopy(src, (caddr_t)shmmap_s, size); |
| 978 | p2->vm_shm = (caddr_t)shmmap_s; |
| 979 | for (; shmmap_s->shmid != SHMID_SENTINEL; shmmap_s++) { |
| 980 | if (SHMID_IS_VALID(shmmap_s->shmid)) { |
| 981 | shmsegs[IPCID_TO_IX(shmmap_s->shmid)].u.shm_nattch++; |
| 982 | } |
| 983 | } |
| 984 | |
| 985 | shmfork_out: |
| 986 | SYSV_SHM_SUBSYS_UNLOCK(); |
| 987 | return ret; |
| 988 | } |
| 989 | |
| 990 | static void |
| 991 | shmcleanup(struct proc *p, int deallocate) |
| 992 | { |
| 993 | struct shmmap_state *shmmap_s; |
| 994 | |
| 995 | SYSV_SHM_SUBSYS_LOCK(); |
| 996 | |
| 997 | shmmap_s = (struct shmmap_state *)p->vm_shm; |
| 998 | for (; shmmap_s->shmid != SHMID_SENTINEL; shmmap_s++) { |
| 999 | if (SHMID_IS_VALID(shmmap_s->shmid)) { |
| 1000 | /* |
| 1001 | * XXX: Should the MAC framework enforce |
| 1002 | * check here as well. |
| 1003 | */ |
| 1004 | shm_delete_mapping(p, shmmap_s, deallocate); |
| 1005 | } |
| 1006 | } |
| 1007 | |
| 1008 | FREE((caddr_t)p->vm_shm, M_SHM); |
| 1009 | p->vm_shm = NULL; |
| 1010 | SYSV_SHM_SUBSYS_UNLOCK(); |
| 1011 | } |
| 1012 | |
| 1013 | void |
| 1014 | shmexit(struct proc *p) |
| 1015 | { |
| 1016 | shmcleanup(p, 1); |
| 1017 | } |
| 1018 | |
| 1019 | /* |
| 1020 | * shmexec() is like shmexit(), only it doesn't delete the mappings, |
| 1021 | * since the old address space has already been destroyed and the new |
| 1022 | * one instantiated. Instead, it just does the housekeeping work we |
| 1023 | * need to do to keep the System V shared memory subsystem sane. |
| 1024 | */ |
| 1025 | __private_extern__ void |
| 1026 | shmexec(struct proc *p) |
| 1027 | { |
| 1028 | shmcleanup(p, 0); |
| 1029 | } |
| 1030 | |
| 1031 | int |
| 1032 | shminit(void) |
| 1033 | { |
| 1034 | size_t sz; |
| 1035 | int i; |
| 1036 | |
| 1037 | if (!shm_inited) { |
| 1038 | /* |
| 1039 | * we store internally 64 bit, since if we didn't, we would |
| 1040 | * be unable to represent a segment size in excess of 32 bits |
| 1041 | * with the (struct shmid_ds)->shm_segsz field; also, POSIX |
| 1042 | * dictates this filed be a size_t, which is 64 bits when |
| 1043 | * running 64 bit binaries. |
| 1044 | */ |
| 1045 | if (os_mul_overflow(shminfo.shmmni, sizeof(struct shmid_kernel), &sz)) { |
| 1046 | return ENOMEM; |
| 1047 | } |
| 1048 | |
| 1049 | MALLOC(shmsegs, struct shmid_kernel *, sz, M_SHM, M_WAITOK); |
| 1050 | if (shmsegs == NULL) { |
| 1051 | return ENOMEM; |
| 1052 | } |
| 1053 | for (i = 0; i < shminfo.shmmni; i++) { |
| 1054 | shmsegs[i].u.shm_perm.mode = SHMSEG_FREE; |
| 1055 | shmsegs[i].u.shm_perm._seq = 0; |
| 1056 | #if CONFIG_MACF |
| 1057 | mac_sysvshm_label_init(&shmsegs[i]); |
| 1058 | #endif |
| 1059 | } |
| 1060 | shm_last_free = 0; |
| 1061 | shm_nused = 0; |
| 1062 | shm_committed = 0; |
| 1063 | shm_inited = 1; |
| 1064 | } |
| 1065 | |
| 1066 | return 0; |
| 1067 | } |
| 1068 | |
| 1069 | /* Initialize the mutex governing access to the SysV shm subsystem */ |
| 1070 | __private_extern__ void |
| 1071 | sysv_shm_lock_init( void ) |
| 1072 | { |
| 1073 | |
| 1074 | sysv_shm_subsys_lck_grp_attr = lck_grp_attr_alloc_init(); |
| 1075 | |
| 1076 | sysv_shm_subsys_lck_grp = lck_grp_alloc_init("sysv_shm_subsys_lock" , sysv_shm_subsys_lck_grp_attr); |
| 1077 | |
| 1078 | sysv_shm_subsys_lck_attr = lck_attr_alloc_init(); |
| 1079 | lck_mtx_init(&sysv_shm_subsys_mutex, sysv_shm_subsys_lck_grp, sysv_shm_subsys_lck_attr); |
| 1080 | } |
| 1081 | |
| 1082 | /* (struct sysctl_oid *oidp, void *arg1, int arg2, \ |
| 1083 | struct sysctl_req *req) */ |
| 1084 | static int |
| 1085 | sysctl_shminfo(__unused struct sysctl_oid *oidp, void *arg1, |
| 1086 | __unused int arg2, struct sysctl_req *req) |
| 1087 | { |
| 1088 | int error = 0; |
| 1089 | int sysctl_shminfo_ret = 0; |
| 1090 | int64_t saved_shmmax; |
| 1091 | int64_t saved_shmmin; |
| 1092 | int64_t saved_shmseg; |
| 1093 | int64_t saved_shmmni; |
| 1094 | int64_t saved_shmall; |
| 1095 | |
| 1096 | error = SYSCTL_OUT(req, arg1, sizeof(int64_t)); |
| 1097 | if (error || req->newptr == USER_ADDR_NULL) |
| 1098 | return(error); |
| 1099 | |
| 1100 | SYSV_SHM_SUBSYS_LOCK(); |
| 1101 | |
| 1102 | /* shmmni can not be changed after SysV SHM has been initialized */ |
| 1103 | if (shm_inited && arg1 == &shminfo.shmmni) { |
| 1104 | sysctl_shminfo_ret = EPERM; |
| 1105 | goto sysctl_shminfo_out; |
| 1106 | } |
| 1107 | saved_shmmax = shminfo.shmmax; |
| 1108 | saved_shmmin = shminfo.shmmin; |
| 1109 | saved_shmseg = shminfo.shmseg; |
| 1110 | saved_shmmni = shminfo.shmmni; |
| 1111 | saved_shmall = shminfo.shmall; |
| 1112 | |
| 1113 | if ((error = SYSCTL_IN(req, arg1, sizeof(int64_t))) != 0) { |
| 1114 | sysctl_shminfo_ret = error; |
| 1115 | goto sysctl_shminfo_out; |
| 1116 | } |
| 1117 | |
| 1118 | if (arg1 == &shminfo.shmmax) { |
| 1119 | /* shmmax needs to be page-aligned */ |
| 1120 | if (shminfo.shmmax & PAGE_MASK_64 || shminfo.shmmax < 0) { |
| 1121 | shminfo.shmmax = saved_shmmax; |
| 1122 | sysctl_shminfo_ret = EINVAL; |
| 1123 | goto sysctl_shminfo_out; |
| 1124 | } |
| 1125 | } |
| 1126 | else if (arg1 == &shminfo.shmmin) { |
| 1127 | if (shminfo.shmmin < 0) { |
| 1128 | shminfo.shmmin = saved_shmmin; |
| 1129 | sysctl_shminfo_ret = EINVAL; |
| 1130 | goto sysctl_shminfo_out; |
| 1131 | } |
| 1132 | } |
| 1133 | else if (arg1 == &shminfo.shmseg) { |
| 1134 | /* add a sanity check - 20847256 */ |
| 1135 | if (shminfo.shmseg > INT32_MAX || shminfo.shmseg < 0) { |
| 1136 | shminfo.shmseg = saved_shmseg; |
| 1137 | sysctl_shminfo_ret = EINVAL; |
| 1138 | goto sysctl_shminfo_out; |
| 1139 | } |
| 1140 | } |
| 1141 | else if (arg1 == &shminfo.shmmni) { |
| 1142 | /* add a sanity check - 20847256 */ |
| 1143 | if (shminfo.shmmni > INT32_MAX || shminfo.shmmni < 0) { |
| 1144 | shminfo.shmmni = saved_shmmni; |
| 1145 | sysctl_shminfo_ret = EINVAL; |
| 1146 | goto sysctl_shminfo_out; |
| 1147 | } |
| 1148 | } |
| 1149 | else if (arg1 == &shminfo.shmall) { |
| 1150 | /* add a sanity check - 20847256 */ |
| 1151 | if (shminfo.shmall > INT32_MAX || shminfo.shmall < 0) { |
| 1152 | shminfo.shmall = saved_shmall; |
| 1153 | sysctl_shminfo_ret = EINVAL; |
| 1154 | goto sysctl_shminfo_out; |
| 1155 | } |
| 1156 | } |
| 1157 | sysctl_shminfo_ret = 0; |
| 1158 | sysctl_shminfo_out: |
| 1159 | SYSV_SHM_SUBSYS_UNLOCK(); |
| 1160 | return sysctl_shminfo_ret; |
| 1161 | } |
| 1162 | |
| 1163 | static int |
| 1164 | IPCS_shm_sysctl(__unused struct sysctl_oid *oidp, __unused void *arg1, |
| 1165 | __unused int arg2, struct sysctl_req *req) |
| 1166 | { |
| 1167 | int error; |
| 1168 | int cursor; |
| 1169 | union { |
| 1170 | struct user32_IPCS_command u32; |
| 1171 | struct user_IPCS_command u64; |
| 1172 | } ipcs; |
| 1173 | struct user32_shmid_ds shmid_ds32 = {}; /* post conversion, 32 bit version */ |
| 1174 | struct user_shmid_ds shmid_ds; /* 64 bit version */ |
| 1175 | void *shmid_dsp; |
| 1176 | size_t ipcs_sz = sizeof(struct user_IPCS_command); |
| 1177 | size_t shmid_ds_sz = sizeof(struct user_shmid_ds); |
| 1178 | struct proc *p = current_proc(); |
| 1179 | |
| 1180 | SYSV_SHM_SUBSYS_LOCK(); |
| 1181 | |
| 1182 | if ((error = shminit())) { |
| 1183 | goto ipcs_shm_sysctl_out; |
| 1184 | } |
| 1185 | |
| 1186 | if (!IS_64BIT_PROCESS(p)) { |
| 1187 | ipcs_sz = sizeof(struct user32_IPCS_command); |
| 1188 | shmid_ds_sz = sizeof(struct user32_shmid_ds); |
| 1189 | } |
| 1190 | |
| 1191 | /* Copy in the command structure */ |
| 1192 | if ((error = SYSCTL_IN(req, &ipcs, ipcs_sz)) != 0) { |
| 1193 | goto ipcs_shm_sysctl_out; |
| 1194 | } |
| 1195 | |
| 1196 | if (!IS_64BIT_PROCESS(p)) /* convert in place */ |
| 1197 | ipcs.u64.ipcs_data = CAST_USER_ADDR_T(ipcs.u32.ipcs_data); |
| 1198 | |
| 1199 | /* Let us version this interface... */ |
| 1200 | if (ipcs.u64.ipcs_magic != IPCS_MAGIC) { |
| 1201 | error = EINVAL; |
| 1202 | goto ipcs_shm_sysctl_out; |
| 1203 | } |
| 1204 | |
| 1205 | switch(ipcs.u64.ipcs_op) { |
| 1206 | case IPCS_SHM_CONF: /* Obtain global configuration data */ |
| 1207 | if (ipcs.u64.ipcs_datalen != sizeof(struct shminfo)) { |
| 1208 | if (ipcs.u64.ipcs_cursor != 0) { /* fwd. compat. */ |
| 1209 | error = ENOMEM; |
| 1210 | break; |
| 1211 | } |
| 1212 | error = ERANGE; |
| 1213 | break; |
| 1214 | } |
| 1215 | error = copyout(&shminfo, ipcs.u64.ipcs_data, ipcs.u64.ipcs_datalen); |
| 1216 | break; |
| 1217 | |
| 1218 | case IPCS_SHM_ITER: /* Iterate over existing segments */ |
| 1219 | cursor = ipcs.u64.ipcs_cursor; |
| 1220 | if (cursor < 0 || cursor >= shminfo.shmmni) { |
| 1221 | error = ERANGE; |
| 1222 | break; |
| 1223 | } |
| 1224 | if (ipcs.u64.ipcs_datalen != (int)shmid_ds_sz) { |
| 1225 | error = EINVAL; |
| 1226 | break; |
| 1227 | } |
| 1228 | for( ; cursor < shminfo.shmmni; cursor++) { |
| 1229 | if (shmsegs[cursor].u.shm_perm.mode & SHMSEG_ALLOCATED) |
| 1230 | break; |
| 1231 | continue; |
| 1232 | } |
| 1233 | if (cursor == shminfo.shmmni) { |
| 1234 | error = ENOENT; |
| 1235 | break; |
| 1236 | } |
| 1237 | |
| 1238 | shmid_dsp = &shmsegs[cursor]; /* default: 64 bit */ |
| 1239 | |
| 1240 | /* |
| 1241 | * If necessary, convert the 64 bit kernel segment |
| 1242 | * descriptor to a 32 bit user one. |
| 1243 | */ |
| 1244 | if (!IS_64BIT_PROCESS(p)) { |
| 1245 | shmid_ds_64to32(shmid_dsp, &shmid_ds32); |
| 1246 | |
| 1247 | /* Clear kernel reserved pointer before copying to user space */ |
| 1248 | shmid_ds32.shm_internal = (user32_addr_t)0; |
| 1249 | |
| 1250 | shmid_dsp = &shmid_ds32; |
| 1251 | } else { |
| 1252 | memcpy(&shmid_ds, shmid_dsp, sizeof(shmid_ds)); |
| 1253 | |
| 1254 | /* Clear kernel reserved pointer before copying to user space */ |
| 1255 | shmid_ds.shm_internal = USER_ADDR_NULL; |
| 1256 | |
| 1257 | shmid_dsp = &shmid_ds; |
| 1258 | } |
| 1259 | error = copyout(shmid_dsp, ipcs.u64.ipcs_data, ipcs.u64.ipcs_datalen); |
| 1260 | if (!error) { |
| 1261 | /* update cursor */ |
| 1262 | ipcs.u64.ipcs_cursor = cursor + 1; |
| 1263 | |
| 1264 | if (!IS_64BIT_PROCESS(p)) /* convert in place */ |
| 1265 | ipcs.u32.ipcs_data = CAST_DOWN_EXPLICIT(user32_addr_t,ipcs.u64.ipcs_data); |
| 1266 | |
| 1267 | error = SYSCTL_OUT(req, &ipcs, ipcs_sz); |
| 1268 | } |
| 1269 | break; |
| 1270 | |
| 1271 | default: |
| 1272 | error = EINVAL; |
| 1273 | break; |
| 1274 | } |
| 1275 | ipcs_shm_sysctl_out: |
| 1276 | SYSV_SHM_SUBSYS_UNLOCK(); |
| 1277 | return(error); |
| 1278 | } |
| 1279 | |
| 1280 | SYSCTL_NODE(_kern, KERN_SYSV, sysv, CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_ANYBODY, 0, "SYSV" ); |
| 1281 | |
| 1282 | SYSCTL_PROC(_kern_sysv, OID_AUTO, shmmax, CTLTYPE_QUAD | CTLFLAG_RW | CTLFLAG_LOCKED, |
| 1283 | &shminfo.shmmax, 0, &sysctl_shminfo ,"Q" ,"shmmax" ); |
| 1284 | |
| 1285 | SYSCTL_PROC(_kern_sysv, OID_AUTO, shmmin, CTLTYPE_QUAD | CTLFLAG_RW | CTLFLAG_LOCKED, |
| 1286 | &shminfo.shmmin, 0, &sysctl_shminfo ,"Q" ,"shmmin" ); |
| 1287 | |
| 1288 | SYSCTL_PROC(_kern_sysv, OID_AUTO, shmmni, CTLTYPE_QUAD | CTLFLAG_RW | CTLFLAG_LOCKED, |
| 1289 | &shminfo.shmmni, 0, &sysctl_shminfo ,"Q" ,"shmmni" ); |
| 1290 | |
| 1291 | SYSCTL_PROC(_kern_sysv, OID_AUTO, shmseg, CTLTYPE_QUAD | CTLFLAG_RW | CTLFLAG_LOCKED, |
| 1292 | &shminfo.shmseg, 0, &sysctl_shminfo ,"Q" ,"shmseg" ); |
| 1293 | |
| 1294 | SYSCTL_PROC(_kern_sysv, OID_AUTO, shmall, CTLTYPE_QUAD | CTLFLAG_RW | CTLFLAG_LOCKED, |
| 1295 | &shminfo.shmall, 0, &sysctl_shminfo ,"Q" ,"shmall" ); |
| 1296 | |
| 1297 | SYSCTL_NODE(_kern_sysv, OID_AUTO, ipcs, CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_ANYBODY, 0, "SYSVIPCS" ); |
| 1298 | |
| 1299 | SYSCTL_PROC(_kern_sysv_ipcs, OID_AUTO, shm, CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_LOCKED, |
| 1300 | 0, 0, IPCS_shm_sysctl, |
| 1301 | "S,IPCS_shm_command" , |
| 1302 | "ipcs shm command interface" ); |
| 1303 | #endif /* SYSV_SHM */ |
| 1304 | |
| 1305 | /* DSEP Review Done pl-20051108-v02 @2743,@2908,@2913,@3009 */ |
| 1306 | |