| 1 | /* |
| 2 | * Copyright (c) 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 | /* |
| 29 | * Copyright (c) 1988 University of Utah. |
| 30 | * Copyright (c) 1991, 1993 |
| 31 | * The Regents of the University of California. All rights reserved. |
| 32 | * |
| 33 | * This code is derived from software contributed to Berkeley by |
| 34 | * the Systems Programming Group of the University of Utah Computer |
| 35 | * Science Department. |
| 36 | * |
| 37 | * Redistribution and use in source and binary forms, with or without |
| 38 | * modification, are permitted provided that the following conditions |
| 39 | * are met: |
| 40 | * 1. Redistributions of source code must retain the above copyright |
| 41 | * notice, this list of conditions and the following disclaimer. |
| 42 | * 2. Redistributions in binary form must reproduce the above copyright |
| 43 | * notice, this list of conditions and the following disclaimer in the |
| 44 | * documentation and/or other materials provided with the distribution. |
| 45 | * 3. All advertising materials mentioning features or use of this software |
| 46 | * must display the following acknowledgement: |
| 47 | * This product includes software developed by the University of |
| 48 | * California, Berkeley and its contributors. |
| 49 | * 4. Neither the name of the University nor the names of its contributors |
| 50 | * may be used to endorse or promote products derived from this software |
| 51 | * without specific prior written permission. |
| 52 | * |
| 53 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
| 54 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 55 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 56 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
| 57 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 58 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| 59 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 60 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 61 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 62 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 63 | * SUCH DAMAGE. |
| 64 | * |
| 65 | * from: Utah $Hdr: vm_mmap.c 1.6 91/10/21$ |
| 66 | * |
| 67 | * @(#)vm_mmap.c 8.10 (Berkeley) 2/19/95 |
| 68 | */ |
| 69 | /* |
| 70 | * NOTICE: This file was modified by SPARTA, Inc. in 2005 to introduce |
| 71 | * support for mandatory and extensible security protections. This notice |
| 72 | * is included in support of clause 2.2 (b) of the Apple Public License, |
| 73 | * Version 2.0. |
| 74 | */ |
| 75 | |
| 76 | /* |
| 77 | * Mapped file (mmap) interface to VM |
| 78 | */ |
| 79 | |
| 80 | #include <sys/param.h> |
| 81 | #include <sys/systm.h> |
| 82 | #include <sys/filedesc.h> |
| 83 | #include <sys/proc_internal.h> |
| 84 | #include <sys/kauth.h> |
| 85 | #include <sys/resourcevar.h> |
| 86 | #include <sys/vnode_internal.h> |
| 87 | #include <sys/acct.h> |
| 88 | #include <sys/wait.h> |
| 89 | #include <sys/file_internal.h> |
| 90 | #include <sys/vadvise.h> |
| 91 | #include <sys/trace.h> |
| 92 | #include <sys/mman.h> |
| 93 | #include <sys/conf.h> |
| 94 | #include <sys/stat.h> |
| 95 | #include <sys/ubc.h> |
| 96 | #include <sys/ubc_internal.h> |
| 97 | #include <sys/sysproto.h> |
| 98 | |
| 99 | #include <sys/syscall.h> |
| 100 | #include <sys/kdebug.h> |
| 101 | #include <sys/bsdtask_info.h> |
| 102 | |
| 103 | #include <security/audit/audit.h> |
| 104 | #include <bsm/audit_kevents.h> |
| 105 | |
| 106 | #include <mach/mach_types.h> |
| 107 | #include <mach/mach_traps.h> |
| 108 | #include <mach/vm_sync.h> |
| 109 | #include <mach/vm_behavior.h> |
| 110 | #include <mach/vm_inherit.h> |
| 111 | #include <mach/vm_statistics.h> |
| 112 | #include <mach/mach_vm.h> |
| 113 | #include <mach/vm_map.h> |
| 114 | #include <mach/host_priv.h> |
| 115 | #include <mach/sdt.h> |
| 116 | |
| 117 | #include <machine/machine_routines.h> |
| 118 | |
| 119 | #include <kern/cpu_number.h> |
| 120 | #include <kern/host.h> |
| 121 | #include <kern/task.h> |
| 122 | #include <kern/page_decrypt.h> |
| 123 | |
| 124 | #include <IOKit/IOReturn.h> |
| 125 | |
| 126 | #include <vm/vm_map.h> |
| 127 | #include <vm/vm_kern.h> |
| 128 | #include <vm/vm_pager.h> |
| 129 | #include <vm/vm_protos.h> |
| 130 | |
| 131 | #if CONFIG_MACF |
| 132 | #include <security/mac_framework.h> |
| 133 | #endif |
| 134 | |
| 135 | /* |
| 136 | * XXX Internally, we use VM_PROT_* somewhat interchangeably, but the correct |
| 137 | * XXX usage is PROT_* from an interface perspective. Thus the values of |
| 138 | * XXX VM_PROT_* and PROT_* need to correspond. |
| 139 | */ |
| 140 | int |
| 141 | mmap(proc_t p, struct mmap_args *uap, user_addr_t *retval) |
| 142 | { |
| 143 | /* |
| 144 | * Map in special device (must be SHARED) or file |
| 145 | */ |
| 146 | struct fileproc *fp; |
| 147 | struct vnode *vp; |
| 148 | int flags; |
| 149 | int prot; |
| 150 | int err=0; |
| 151 | vm_map_t user_map; |
| 152 | kern_return_t result; |
| 153 | vm_map_offset_t user_addr; |
| 154 | vm_map_size_t user_size; |
| 155 | vm_object_offset_t pageoff; |
| 156 | vm_object_offset_t file_pos; |
| 157 | int alloc_flags = 0; |
| 158 | vm_tag_t tag = VM_KERN_MEMORY_NONE; |
| 159 | vm_map_kernel_flags_t vmk_flags = VM_MAP_KERNEL_FLAGS_NONE; |
| 160 | boolean_t docow; |
| 161 | vm_prot_t maxprot; |
| 162 | void *handle; |
| 163 | memory_object_t = MEMORY_OBJECT_NULL; |
| 164 | memory_object_control_t control; |
| 165 | int mapanon=0; |
| 166 | int fpref=0; |
| 167 | int error =0; |
| 168 | int fd = uap->fd; |
| 169 | int num_retries = 0; |
| 170 | |
| 171 | /* |
| 172 | * Note that for UNIX03 conformance, there is additional parameter checking for |
| 173 | * mmap() system call in libsyscall prior to entering the kernel. The sanity |
| 174 | * checks and argument validation done in this function are not the only places |
| 175 | * one can get returned errnos. |
| 176 | */ |
| 177 | |
| 178 | user_map = current_map(); |
| 179 | user_addr = (vm_map_offset_t)uap->addr; |
| 180 | user_size = (vm_map_size_t) uap->len; |
| 181 | |
| 182 | AUDIT_ARG(addr, user_addr); |
| 183 | AUDIT_ARG(len, user_size); |
| 184 | AUDIT_ARG(fd, uap->fd); |
| 185 | |
| 186 | prot = (uap->prot & VM_PROT_ALL); |
| 187 | #if 3777787 |
| 188 | /* |
| 189 | * Since the hardware currently does not support writing without |
| 190 | * read-before-write, or execution-without-read, if the request is |
| 191 | * for write or execute access, we must imply read access as well; |
| 192 | * otherwise programs expecting this to work will fail to operate. |
| 193 | */ |
| 194 | if (prot & (VM_PROT_EXECUTE | VM_PROT_WRITE)) |
| 195 | prot |= VM_PROT_READ; |
| 196 | #endif /* radar 3777787 */ |
| 197 | |
| 198 | flags = uap->flags; |
| 199 | vp = NULLVP; |
| 200 | |
| 201 | /* |
| 202 | * The vm code does not have prototypes & compiler doesn't do the' |
| 203 | * the right thing when you cast 64bit value and pass it in function |
| 204 | * call. So here it is. |
| 205 | */ |
| 206 | file_pos = (vm_object_offset_t)uap->pos; |
| 207 | |
| 208 | |
| 209 | /* make sure mapping fits into numeric range etc */ |
| 210 | if (file_pos + user_size > (vm_object_offset_t)-PAGE_SIZE_64) |
| 211 | return (EINVAL); |
| 212 | |
| 213 | /* |
| 214 | * Align the file position to a page boundary, |
| 215 | * and save its page offset component. |
| 216 | */ |
| 217 | pageoff = (file_pos & vm_map_page_mask(user_map)); |
| 218 | file_pos -= (vm_object_offset_t)pageoff; |
| 219 | |
| 220 | |
| 221 | /* Adjust size for rounding (on both ends). */ |
| 222 | user_size += pageoff; /* low end... */ |
| 223 | user_size = vm_map_round_page(user_size, |
| 224 | vm_map_page_mask(user_map)); /* hi end */ |
| 225 | |
| 226 | if (flags & MAP_JIT) { |
| 227 | if ((flags & MAP_FIXED) || |
| 228 | (flags & MAP_SHARED) || |
| 229 | !(flags & MAP_ANON) || |
| 230 | (flags & MAP_RESILIENT_CODESIGN) || |
| 231 | (flags & MAP_RESILIENT_MEDIA)) { |
| 232 | return EINVAL; |
| 233 | } |
| 234 | } |
| 235 | |
| 236 | if ((flags & MAP_RESILIENT_CODESIGN) || |
| 237 | (flags & MAP_RESILIENT_MEDIA)) { |
| 238 | if ((flags & MAP_ANON) || |
| 239 | (flags & MAP_JIT)) { |
| 240 | return EINVAL; |
| 241 | } |
| 242 | if (prot & (VM_PROT_WRITE | VM_PROT_EXECUTE)) { |
| 243 | return EPERM; |
| 244 | } |
| 245 | } |
| 246 | |
| 247 | /* |
| 248 | * Check for illegal addresses. Watch out for address wrap... Note |
| 249 | * that VM_*_ADDRESS are not constants due to casts (argh). |
| 250 | */ |
| 251 | if (flags & MAP_FIXED) { |
| 252 | /* |
| 253 | * The specified address must have the same remainder |
| 254 | * as the file offset taken modulo PAGE_SIZE, so it |
| 255 | * should be aligned after adjustment by pageoff. |
| 256 | */ |
| 257 | user_addr -= pageoff; |
| 258 | if (user_addr & vm_map_page_mask(user_map)) |
| 259 | return (EINVAL); |
| 260 | } |
| 261 | #ifdef notyet |
| 262 | /* DO not have apis to get this info, need to wait till then*/ |
| 263 | /* |
| 264 | * XXX for non-fixed mappings where no hint is provided or |
| 265 | * the hint would fall in the potential heap space, |
| 266 | * place it after the end of the largest possible heap. |
| 267 | * |
| 268 | * There should really be a pmap call to determine a reasonable |
| 269 | * location. |
| 270 | */ |
| 271 | else if (addr < vm_map_round_page(p->p_vmspace->vm_daddr + MAXDSIZ, |
| 272 | vm_map_page_mask(user_map))) |
| 273 | addr = vm_map_round_page(p->p_vmspace->vm_daddr + MAXDSIZ, |
| 274 | vm_map_page_mask(user_map)); |
| 275 | |
| 276 | #endif |
| 277 | |
| 278 | alloc_flags = 0; |
| 279 | |
| 280 | if (flags & MAP_ANON) { |
| 281 | |
| 282 | maxprot = VM_PROT_ALL; |
| 283 | #if CONFIG_MACF |
| 284 | /* |
| 285 | * Entitlement check. |
| 286 | */ |
| 287 | error = mac_proc_check_map_anon(p, user_addr, user_size, prot, flags, &maxprot); |
| 288 | if (error) { |
| 289 | return EINVAL; |
| 290 | } |
| 291 | #endif /* MAC */ |
| 292 | |
| 293 | /* |
| 294 | * Mapping blank space is trivial. Use positive fds as the alias |
| 295 | * value for memory tracking. |
| 296 | */ |
| 297 | if (fd != -1) { |
| 298 | /* |
| 299 | * Use "fd" to pass (some) Mach VM allocation flags, |
| 300 | * (see the VM_FLAGS_* definitions). |
| 301 | */ |
| 302 | alloc_flags = fd & (VM_FLAGS_ALIAS_MASK | |
| 303 | VM_FLAGS_SUPERPAGE_MASK | |
| 304 | VM_FLAGS_PURGABLE | |
| 305 | VM_FLAGS_4GB_CHUNK); |
| 306 | if (alloc_flags != fd) { |
| 307 | /* reject if there are any extra flags */ |
| 308 | return EINVAL; |
| 309 | } |
| 310 | VM_GET_FLAGS_ALIAS(alloc_flags, tag); |
| 311 | alloc_flags &= ~VM_FLAGS_ALIAS_MASK; |
| 312 | } |
| 313 | |
| 314 | handle = NULL; |
| 315 | file_pos = 0; |
| 316 | mapanon = 1; |
| 317 | } else { |
| 318 | struct vnode_attr va; |
| 319 | vfs_context_t ctx = vfs_context_current(); |
| 320 | |
| 321 | if (flags & MAP_JIT) |
| 322 | return EINVAL; |
| 323 | |
| 324 | /* |
| 325 | * Mapping file, get fp for validation. Obtain vnode and make |
| 326 | * sure it is of appropriate type. |
| 327 | */ |
| 328 | err = fp_lookup(p, fd, &fp, 0); |
| 329 | if (err) |
| 330 | return(err); |
| 331 | fpref = 1; |
| 332 | switch (FILEGLOB_DTYPE(fp->f_fglob)) { |
| 333 | case DTYPE_PSXSHM: |
| 334 | uap->addr = (user_addr_t)user_addr; |
| 335 | uap->len = (user_size_t)user_size; |
| 336 | uap->prot = prot; |
| 337 | uap->flags = flags; |
| 338 | uap->pos = file_pos; |
| 339 | error = pshm_mmap(p, uap, retval, fp, (off_t)pageoff); |
| 340 | goto bad; |
| 341 | case DTYPE_VNODE: |
| 342 | break; |
| 343 | default: |
| 344 | error = EINVAL; |
| 345 | goto bad; |
| 346 | } |
| 347 | vp = (struct vnode *)fp->f_fglob->fg_data; |
| 348 | error = vnode_getwithref(vp); |
| 349 | if(error != 0) |
| 350 | goto bad; |
| 351 | |
| 352 | if (vp->v_type != VREG && vp->v_type != VCHR) { |
| 353 | (void)vnode_put(vp); |
| 354 | error = EINVAL; |
| 355 | goto bad; |
| 356 | } |
| 357 | |
| 358 | AUDIT_ARG(vnpath, vp, ARG_VNODE1); |
| 359 | |
| 360 | /* |
| 361 | * POSIX: mmap needs to update access time for mapped files |
| 362 | */ |
| 363 | if ((vnode_vfsvisflags(vp) & MNT_NOATIME) == 0) { |
| 364 | VATTR_INIT(&va); |
| 365 | nanotime(&va.va_access_time); |
| 366 | VATTR_SET_ACTIVE(&va, va_access_time); |
| 367 | vnode_setattr(vp, &va, ctx); |
| 368 | } |
| 369 | |
| 370 | /* |
| 371 | * XXX hack to handle use of /dev/zero to map anon memory (ala |
| 372 | * SunOS). |
| 373 | */ |
| 374 | if (vp->v_type == VCHR || vp->v_type == VSTR) { |
| 375 | (void)vnode_put(vp); |
| 376 | error = ENODEV; |
| 377 | goto bad; |
| 378 | } else { |
| 379 | /* |
| 380 | * Ensure that file and memory protections are |
| 381 | * compatible. Note that we only worry about |
| 382 | * writability if mapping is shared; in this case, |
| 383 | * current and max prot are dictated by the open file. |
| 384 | * XXX use the vnode instead? Problem is: what |
| 385 | * credentials do we use for determination? What if |
| 386 | * proc does a setuid? |
| 387 | */ |
| 388 | maxprot = VM_PROT_EXECUTE; /* ??? */ |
| 389 | if (fp->f_fglob->fg_flag & FREAD) |
| 390 | maxprot |= VM_PROT_READ; |
| 391 | else if (prot & PROT_READ) { |
| 392 | (void)vnode_put(vp); |
| 393 | error = EACCES; |
| 394 | goto bad; |
| 395 | } |
| 396 | /* |
| 397 | * If we are sharing potential changes (either via |
| 398 | * MAP_SHARED or via the implicit sharing of character |
| 399 | * device mappings), and we are trying to get write |
| 400 | * permission although we opened it without asking |
| 401 | * for it, bail out. |
| 402 | */ |
| 403 | |
| 404 | if ((flags & MAP_SHARED) != 0) { |
| 405 | if ((fp->f_fglob->fg_flag & FWRITE) != 0 && |
| 406 | /* |
| 407 | * Do not allow writable mappings of |
| 408 | * swap files (see vm_swapfile_pager.c). |
| 409 | */ |
| 410 | !vnode_isswap(vp)) { |
| 411 | /* |
| 412 | * check for write access |
| 413 | * |
| 414 | * Note that we already made this check when granting FWRITE |
| 415 | * against the file, so it seems redundant here. |
| 416 | */ |
| 417 | error = vnode_authorize(vp, NULL, KAUTH_VNODE_CHECKIMMUTABLE, ctx); |
| 418 | |
| 419 | /* if not granted for any reason, but we wanted it, bad */ |
| 420 | if ((prot & PROT_WRITE) && (error != 0)) { |
| 421 | vnode_put(vp); |
| 422 | goto bad; |
| 423 | } |
| 424 | |
| 425 | /* if writable, remember */ |
| 426 | if (error == 0) |
| 427 | maxprot |= VM_PROT_WRITE; |
| 428 | |
| 429 | } else if ((prot & PROT_WRITE) != 0) { |
| 430 | (void)vnode_put(vp); |
| 431 | error = EACCES; |
| 432 | goto bad; |
| 433 | } |
| 434 | } else |
| 435 | maxprot |= VM_PROT_WRITE; |
| 436 | |
| 437 | handle = (void *)vp; |
| 438 | #if CONFIG_MACF |
| 439 | error = mac_file_check_mmap(vfs_context_ucred(ctx), |
| 440 | fp->f_fglob, prot, flags, file_pos, &maxprot); |
| 441 | if (error) { |
| 442 | (void)vnode_put(vp); |
| 443 | goto bad; |
| 444 | } |
| 445 | #endif /* MAC */ |
| 446 | } |
| 447 | } |
| 448 | |
| 449 | if (user_size == 0) { |
| 450 | if (!mapanon) |
| 451 | (void)vnode_put(vp); |
| 452 | error = 0; |
| 453 | goto bad; |
| 454 | } |
| 455 | |
| 456 | /* |
| 457 | * We bend a little - round the start and end addresses |
| 458 | * to the nearest page boundary. |
| 459 | */ |
| 460 | user_size = vm_map_round_page(user_size, |
| 461 | vm_map_page_mask(user_map)); |
| 462 | |
| 463 | if (file_pos & vm_map_page_mask(user_map)) { |
| 464 | if (!mapanon) |
| 465 | (void)vnode_put(vp); |
| 466 | error = EINVAL; |
| 467 | goto bad; |
| 468 | } |
| 469 | |
| 470 | if ((flags & MAP_FIXED) == 0) { |
| 471 | alloc_flags |= VM_FLAGS_ANYWHERE; |
| 472 | user_addr = vm_map_round_page(user_addr, |
| 473 | vm_map_page_mask(user_map)); |
| 474 | } else { |
| 475 | if (user_addr != vm_map_trunc_page(user_addr, |
| 476 | vm_map_page_mask(user_map))) { |
| 477 | if (!mapanon) |
| 478 | (void)vnode_put(vp); |
| 479 | error = EINVAL; |
| 480 | goto bad; |
| 481 | } |
| 482 | /* |
| 483 | * mmap(MAP_FIXED) will replace any existing mappings in the |
| 484 | * specified range, if the new mapping is successful. |
| 485 | * If we just deallocate the specified address range here, |
| 486 | * another thread might jump in and allocate memory in that |
| 487 | * range before we get a chance to establish the new mapping, |
| 488 | * and we won't have a chance to restore the old mappings. |
| 489 | * So we use VM_FLAGS_OVERWRITE to let Mach VM know that it |
| 490 | * has to deallocate the existing mappings and establish the |
| 491 | * new ones atomically. |
| 492 | */ |
| 493 | alloc_flags |= VM_FLAGS_FIXED | VM_FLAGS_OVERWRITE; |
| 494 | } |
| 495 | |
| 496 | if (flags & MAP_NOCACHE) |
| 497 | alloc_flags |= VM_FLAGS_NO_CACHE; |
| 498 | |
| 499 | if (flags & MAP_JIT) { |
| 500 | vmk_flags.vmkf_map_jit = TRUE; |
| 501 | } |
| 502 | |
| 503 | if (flags & MAP_RESILIENT_CODESIGN) { |
| 504 | alloc_flags |= VM_FLAGS_RESILIENT_CODESIGN; |
| 505 | } |
| 506 | |
| 507 | /* |
| 508 | * Lookup/allocate object. |
| 509 | */ |
| 510 | if (handle == NULL) { |
| 511 | control = NULL; |
| 512 | #ifdef notyet |
| 513 | /* Hmm .. */ |
| 514 | #if defined(VM_PROT_READ_IS_EXEC) |
| 515 | if (prot & VM_PROT_READ) |
| 516 | prot |= VM_PROT_EXECUTE; |
| 517 | if (maxprot & VM_PROT_READ) |
| 518 | maxprot |= VM_PROT_EXECUTE; |
| 519 | #endif |
| 520 | #endif |
| 521 | |
| 522 | #if 3777787 |
| 523 | if (prot & (VM_PROT_EXECUTE | VM_PROT_WRITE)) |
| 524 | prot |= VM_PROT_READ; |
| 525 | if (maxprot & (VM_PROT_EXECUTE | VM_PROT_WRITE)) |
| 526 | maxprot |= VM_PROT_READ; |
| 527 | #endif /* radar 3777787 */ |
| 528 | map_anon_retry: |
| 529 | result = vm_map_enter_mem_object(user_map, |
| 530 | &user_addr, user_size, |
| 531 | 0, alloc_flags, vmk_flags, |
| 532 | tag, |
| 533 | IPC_PORT_NULL, 0, FALSE, |
| 534 | prot, maxprot, |
| 535 | (flags & MAP_SHARED) ? |
| 536 | VM_INHERIT_SHARE : |
| 537 | VM_INHERIT_DEFAULT); |
| 538 | |
| 539 | /* If a non-binding address was specified for this anonymous |
| 540 | * mapping, retry the mapping with a zero base |
| 541 | * in the event the mapping operation failed due to |
| 542 | * lack of space between the address and the map's maximum. |
| 543 | */ |
| 544 | if ((result == KERN_NO_SPACE) && ((flags & MAP_FIXED) == 0) && user_addr && (num_retries++ == 0)) { |
| 545 | user_addr = vm_map_page_size(user_map); |
| 546 | goto map_anon_retry; |
| 547 | } |
| 548 | } else { |
| 549 | if (vnode_isswap(vp)) { |
| 550 | /* |
| 551 | * Map swap files with a special pager |
| 552 | * that returns obfuscated contents. |
| 553 | */ |
| 554 | control = NULL; |
| 555 | pager = swapfile_pager_setup(vp); |
| 556 | if (pager != MEMORY_OBJECT_NULL) { |
| 557 | control = swapfile_pager_control(pager); |
| 558 | } |
| 559 | } else { |
| 560 | control = ubc_getobject(vp, UBC_FLAGS_NONE); |
| 561 | } |
| 562 | |
| 563 | if (control == NULL) { |
| 564 | (void)vnode_put(vp); |
| 565 | error = ENOMEM; |
| 566 | goto bad; |
| 567 | } |
| 568 | |
| 569 | /* |
| 570 | * Set credentials: |
| 571 | * FIXME: if we're writing the file we need a way to |
| 572 | * ensure that someone doesn't replace our R/W creds |
| 573 | * with ones that only work for read. |
| 574 | */ |
| 575 | |
| 576 | ubc_setthreadcred(vp, p, current_thread()); |
| 577 | docow = FALSE; |
| 578 | if ((flags & (MAP_ANON|MAP_SHARED)) == 0) { |
| 579 | docow = TRUE; |
| 580 | } |
| 581 | |
| 582 | #ifdef notyet |
| 583 | /* Hmm .. */ |
| 584 | #if defined(VM_PROT_READ_IS_EXEC) |
| 585 | if (prot & VM_PROT_READ) |
| 586 | prot |= VM_PROT_EXECUTE; |
| 587 | if (maxprot & VM_PROT_READ) |
| 588 | maxprot |= VM_PROT_EXECUTE; |
| 589 | #endif |
| 590 | #endif /* notyet */ |
| 591 | |
| 592 | #if 3777787 |
| 593 | if (prot & (VM_PROT_EXECUTE | VM_PROT_WRITE)) |
| 594 | prot |= VM_PROT_READ; |
| 595 | if (maxprot & (VM_PROT_EXECUTE | VM_PROT_WRITE)) |
| 596 | maxprot |= VM_PROT_READ; |
| 597 | #endif /* radar 3777787 */ |
| 598 | |
| 599 | map_file_retry: |
| 600 | if ((flags & MAP_RESILIENT_CODESIGN) || |
| 601 | (flags & MAP_RESILIENT_MEDIA)) { |
| 602 | if (prot & (VM_PROT_WRITE | VM_PROT_EXECUTE)) { |
| 603 | assert(!mapanon); |
| 604 | vnode_put(vp); |
| 605 | error = EPERM; |
| 606 | goto bad; |
| 607 | } |
| 608 | /* strictly limit access to "prot" */ |
| 609 | maxprot &= prot; |
| 610 | } |
| 611 | |
| 612 | vm_object_offset_t end_pos = 0; |
| 613 | if (os_add_overflow(user_size, file_pos, &end_pos)) { |
| 614 | vnode_put(vp); |
| 615 | error = EINVAL; |
| 616 | goto bad; |
| 617 | } |
| 618 | |
| 619 | result = vm_map_enter_mem_object_control(user_map, |
| 620 | &user_addr, user_size, |
| 621 | 0, alloc_flags, vmk_flags, |
| 622 | tag, |
| 623 | control, file_pos, |
| 624 | docow, prot, maxprot, |
| 625 | (flags & MAP_SHARED) ? |
| 626 | VM_INHERIT_SHARE : |
| 627 | VM_INHERIT_DEFAULT); |
| 628 | |
| 629 | /* If a non-binding address was specified for this file backed |
| 630 | * mapping, retry the mapping with a zero base |
| 631 | * in the event the mapping operation failed due to |
| 632 | * lack of space between the address and the map's maximum. |
| 633 | */ |
| 634 | if ((result == KERN_NO_SPACE) && ((flags & MAP_FIXED) == 0) && user_addr && (num_retries++ == 0)) { |
| 635 | user_addr = vm_map_page_size(user_map); |
| 636 | goto map_file_retry; |
| 637 | } |
| 638 | } |
| 639 | |
| 640 | if (!mapanon) { |
| 641 | (void)vnode_put(vp); |
| 642 | } |
| 643 | |
| 644 | switch (result) { |
| 645 | case KERN_SUCCESS: |
| 646 | *retval = user_addr + pageoff; |
| 647 | error = 0; |
| 648 | break; |
| 649 | case KERN_INVALID_ADDRESS: |
| 650 | case KERN_NO_SPACE: |
| 651 | error = ENOMEM; |
| 652 | break; |
| 653 | case KERN_PROTECTION_FAILURE: |
| 654 | error = EACCES; |
| 655 | break; |
| 656 | default: |
| 657 | error = EINVAL; |
| 658 | break; |
| 659 | } |
| 660 | bad: |
| 661 | if (pager != MEMORY_OBJECT_NULL) { |
| 662 | /* |
| 663 | * Release the reference on the pager. |
| 664 | * If the mapping was successful, it now holds |
| 665 | * an extra reference. |
| 666 | */ |
| 667 | memory_object_deallocate(pager); |
| 668 | } |
| 669 | if (fpref) |
| 670 | fp_drop(p, fd, fp, 0); |
| 671 | |
| 672 | KERNEL_DEBUG_CONSTANT((BSDDBG_CODE(DBG_BSD_SC_EXTENDED_INFO, SYS_mmap) | DBG_FUNC_NONE), fd, (uint32_t)(*retval), (uint32_t)user_size, error, 0); |
| 673 | #ifndef CONFIG_EMBEDDED |
| 674 | KERNEL_DEBUG_CONSTANT((BSDDBG_CODE(DBG_BSD_SC_EXTENDED_INFO2, SYS_mmap) | DBG_FUNC_NONE), (uint32_t)(*retval >> 32), (uint32_t)(user_size >> 32), |
| 675 | (uint32_t)(file_pos >> 32), (uint32_t)file_pos, 0); |
| 676 | #endif |
| 677 | return(error); |
| 678 | } |
| 679 | |
| 680 | int |
| 681 | msync(__unused proc_t p, struct msync_args *uap, int32_t *retval) |
| 682 | { |
| 683 | __pthread_testcancel(1); |
| 684 | return(msync_nocancel(p, (struct msync_nocancel_args *)uap, retval)); |
| 685 | } |
| 686 | |
| 687 | int |
| 688 | msync_nocancel(__unused proc_t p, struct msync_nocancel_args *uap, __unused int32_t *retval) |
| 689 | { |
| 690 | mach_vm_offset_t addr; |
| 691 | mach_vm_size_t size; |
| 692 | int flags; |
| 693 | vm_map_t user_map; |
| 694 | int rv; |
| 695 | vm_sync_t sync_flags=0; |
| 696 | |
| 697 | user_map = current_map(); |
| 698 | addr = (mach_vm_offset_t) uap->addr; |
| 699 | size = (mach_vm_size_t)uap->len; |
| 700 | #ifndef CONFIG_EMBEDDED |
| 701 | KERNEL_DEBUG_CONSTANT((BSDDBG_CODE(DBG_BSD_SC_EXTENDED_INFO, SYS_msync) | DBG_FUNC_NONE), (uint32_t)(addr >> 32), (uint32_t)(size >> 32), 0, 0, 0); |
| 702 | #endif |
| 703 | if (addr & vm_map_page_mask(user_map)) { |
| 704 | /* UNIX SPEC: user address is not page-aligned, return EINVAL */ |
| 705 | return EINVAL; |
| 706 | } |
| 707 | if (size == 0) { |
| 708 | /* |
| 709 | * We cannot support this properly without maintaining |
| 710 | * list all mmaps done. Cannot use vm_map_entry as they could be |
| 711 | * split or coalesced by indepenedant actions. So instead of |
| 712 | * inaccurate results, lets just return error as invalid size |
| 713 | * specified |
| 714 | */ |
| 715 | return (EINVAL); /* XXX breaks posix apps */ |
| 716 | } |
| 717 | |
| 718 | flags = uap->flags; |
| 719 | /* disallow contradictory flags */ |
| 720 | if ((flags & (MS_SYNC|MS_ASYNC)) == (MS_SYNC|MS_ASYNC)) |
| 721 | return (EINVAL); |
| 722 | |
| 723 | if (flags & MS_KILLPAGES) |
| 724 | sync_flags |= VM_SYNC_KILLPAGES; |
| 725 | if (flags & MS_DEACTIVATE) |
| 726 | sync_flags |= VM_SYNC_DEACTIVATE; |
| 727 | if (flags & MS_INVALIDATE) |
| 728 | sync_flags |= VM_SYNC_INVALIDATE; |
| 729 | |
| 730 | if ( !(flags & (MS_KILLPAGES | MS_DEACTIVATE))) { |
| 731 | if (flags & MS_ASYNC) |
| 732 | sync_flags |= VM_SYNC_ASYNCHRONOUS; |
| 733 | else |
| 734 | sync_flags |= VM_SYNC_SYNCHRONOUS; |
| 735 | } |
| 736 | |
| 737 | sync_flags |= VM_SYNC_CONTIGUOUS; /* complain if holes */ |
| 738 | |
| 739 | rv = mach_vm_msync(user_map, addr, size, sync_flags); |
| 740 | |
| 741 | switch (rv) { |
| 742 | case KERN_SUCCESS: |
| 743 | break; |
| 744 | case KERN_INVALID_ADDRESS: /* hole in region being sync'ed */ |
| 745 | return (ENOMEM); |
| 746 | case KERN_FAILURE: |
| 747 | return (EIO); |
| 748 | default: |
| 749 | return (EINVAL); |
| 750 | } |
| 751 | return (0); |
| 752 | } |
| 753 | |
| 754 | |
| 755 | int |
| 756 | munmap(__unused proc_t p, struct munmap_args *uap, __unused int32_t *retval) |
| 757 | { |
| 758 | mach_vm_offset_t user_addr; |
| 759 | mach_vm_size_t user_size; |
| 760 | kern_return_t result; |
| 761 | vm_map_t user_map; |
| 762 | |
| 763 | user_map = current_map(); |
| 764 | user_addr = (mach_vm_offset_t) uap->addr; |
| 765 | user_size = (mach_vm_size_t) uap->len; |
| 766 | |
| 767 | AUDIT_ARG(addr, user_addr); |
| 768 | AUDIT_ARG(len, user_size); |
| 769 | |
| 770 | if (user_addr & vm_map_page_mask(user_map)) { |
| 771 | /* UNIX SPEC: user address is not page-aligned, return EINVAL */ |
| 772 | return EINVAL; |
| 773 | } |
| 774 | |
| 775 | if (user_addr + user_size < user_addr) |
| 776 | return(EINVAL); |
| 777 | |
| 778 | if (user_size == 0) { |
| 779 | /* UNIX SPEC: size is 0, return EINVAL */ |
| 780 | return EINVAL; |
| 781 | } |
| 782 | |
| 783 | result = mach_vm_deallocate(user_map, user_addr, user_size); |
| 784 | if (result != KERN_SUCCESS) { |
| 785 | return(EINVAL); |
| 786 | } |
| 787 | return(0); |
| 788 | } |
| 789 | |
| 790 | int |
| 791 | mprotect(__unused proc_t p, struct mprotect_args *uap, __unused int32_t *retval) |
| 792 | { |
| 793 | vm_prot_t prot; |
| 794 | mach_vm_offset_t user_addr; |
| 795 | mach_vm_size_t user_size; |
| 796 | kern_return_t result; |
| 797 | vm_map_t user_map; |
| 798 | #if CONFIG_MACF |
| 799 | int error; |
| 800 | #endif |
| 801 | |
| 802 | AUDIT_ARG(addr, uap->addr); |
| 803 | AUDIT_ARG(len, uap->len); |
| 804 | AUDIT_ARG(value32, uap->prot); |
| 805 | |
| 806 | user_map = current_map(); |
| 807 | user_addr = (mach_vm_offset_t) uap->addr; |
| 808 | user_size = (mach_vm_size_t) uap->len; |
| 809 | prot = (vm_prot_t)(uap->prot & (VM_PROT_ALL | VM_PROT_TRUSTED | VM_PROT_STRIP_READ)); |
| 810 | |
| 811 | if (user_addr & vm_map_page_mask(user_map)) { |
| 812 | /* UNIX SPEC: user address is not page-aligned, return EINVAL */ |
| 813 | return EINVAL; |
| 814 | } |
| 815 | |
| 816 | #ifdef notyet |
| 817 | /* Hmm .. */ |
| 818 | #if defined(VM_PROT_READ_IS_EXEC) |
| 819 | if (prot & VM_PROT_READ) |
| 820 | prot |= VM_PROT_EXECUTE; |
| 821 | #endif |
| 822 | #endif /* notyet */ |
| 823 | |
| 824 | #if 3936456 |
| 825 | if (prot & (VM_PROT_EXECUTE | VM_PROT_WRITE)) |
| 826 | prot |= VM_PROT_READ; |
| 827 | #endif /* 3936456 */ |
| 828 | |
| 829 | #if defined(__arm64__) |
| 830 | if (prot & VM_PROT_STRIP_READ) |
| 831 | prot &= ~(VM_PROT_READ | VM_PROT_STRIP_READ); |
| 832 | #endif |
| 833 | |
| 834 | #if CONFIG_MACF |
| 835 | /* |
| 836 | * The MAC check for mprotect is of limited use for 2 reasons: |
| 837 | * Without mmap revocation, the caller could have asked for the max |
| 838 | * protections initially instead of a reduced set, so a mprotect |
| 839 | * check would offer no new security. |
| 840 | * It is not possible to extract the vnode from the pager object(s) |
| 841 | * of the target memory range. |
| 842 | * However, the MAC check may be used to prevent a process from, |
| 843 | * e.g., making the stack executable. |
| 844 | */ |
| 845 | error = mac_proc_check_mprotect(p, user_addr, |
| 846 | user_size, prot); |
| 847 | if (error) |
| 848 | return (error); |
| 849 | #endif |
| 850 | |
| 851 | if(prot & VM_PROT_TRUSTED) { |
| 852 | #if CONFIG_DYNAMIC_CODE_SIGNING |
| 853 | /* CODE SIGNING ENFORCEMENT - JIT support */ |
| 854 | /* The special protection value VM_PROT_TRUSTED requests that we treat |
| 855 | * this page as if it had a valid code signature. |
| 856 | * If this is enabled, there MUST be a MAC policy implementing the |
| 857 | * mac_proc_check_mprotect() hook above. Otherwise, Codesigning will be |
| 858 | * compromised because the check would always succeed and thusly any |
| 859 | * process could sign dynamically. */ |
| 860 | result = vm_map_sign( |
| 861 | user_map, |
| 862 | vm_map_trunc_page(user_addr, |
| 863 | vm_map_page_mask(user_map)), |
| 864 | vm_map_round_page(user_addr+user_size, |
| 865 | vm_map_page_mask(user_map))); |
| 866 | switch (result) { |
| 867 | case KERN_SUCCESS: |
| 868 | break; |
| 869 | case KERN_INVALID_ADDRESS: |
| 870 | /* UNIX SPEC: for an invalid address range, return ENOMEM */ |
| 871 | return ENOMEM; |
| 872 | default: |
| 873 | return EINVAL; |
| 874 | } |
| 875 | #else |
| 876 | return ENOTSUP; |
| 877 | #endif |
| 878 | } |
| 879 | prot &= ~VM_PROT_TRUSTED; |
| 880 | |
| 881 | result = mach_vm_protect(user_map, user_addr, user_size, |
| 882 | FALSE, prot); |
| 883 | switch (result) { |
| 884 | case KERN_SUCCESS: |
| 885 | return (0); |
| 886 | case KERN_PROTECTION_FAILURE: |
| 887 | return (EACCES); |
| 888 | case KERN_INVALID_ADDRESS: |
| 889 | /* UNIX SPEC: for an invalid address range, return ENOMEM */ |
| 890 | return ENOMEM; |
| 891 | } |
| 892 | return (EINVAL); |
| 893 | } |
| 894 | |
| 895 | |
| 896 | int |
| 897 | minherit(__unused proc_t p, struct minherit_args *uap, __unused int32_t *retval) |
| 898 | { |
| 899 | mach_vm_offset_t addr; |
| 900 | mach_vm_size_t size; |
| 901 | vm_inherit_t inherit; |
| 902 | vm_map_t user_map; |
| 903 | kern_return_t result; |
| 904 | |
| 905 | AUDIT_ARG(addr, uap->addr); |
| 906 | AUDIT_ARG(len, uap->len); |
| 907 | AUDIT_ARG(value32, uap->inherit); |
| 908 | |
| 909 | addr = (mach_vm_offset_t)uap->addr; |
| 910 | size = (mach_vm_size_t)uap->len; |
| 911 | inherit = uap->inherit; |
| 912 | |
| 913 | user_map = current_map(); |
| 914 | result = mach_vm_inherit(user_map, addr, size, |
| 915 | inherit); |
| 916 | switch (result) { |
| 917 | case KERN_SUCCESS: |
| 918 | return (0); |
| 919 | case KERN_PROTECTION_FAILURE: |
| 920 | return (EACCES); |
| 921 | } |
| 922 | return (EINVAL); |
| 923 | } |
| 924 | |
| 925 | int |
| 926 | madvise(__unused proc_t p, struct madvise_args *uap, __unused int32_t *retval) |
| 927 | { |
| 928 | vm_map_t user_map; |
| 929 | mach_vm_offset_t start; |
| 930 | mach_vm_size_t size; |
| 931 | vm_behavior_t new_behavior; |
| 932 | kern_return_t result; |
| 933 | |
| 934 | /* |
| 935 | * Since this routine is only advisory, we default to conservative |
| 936 | * behavior. |
| 937 | */ |
| 938 | switch (uap->behav) { |
| 939 | case MADV_RANDOM: |
| 940 | new_behavior = VM_BEHAVIOR_RANDOM; |
| 941 | break; |
| 942 | case MADV_SEQUENTIAL: |
| 943 | new_behavior = VM_BEHAVIOR_SEQUENTIAL; |
| 944 | break; |
| 945 | case MADV_NORMAL: |
| 946 | new_behavior = VM_BEHAVIOR_DEFAULT; |
| 947 | break; |
| 948 | case MADV_WILLNEED: |
| 949 | new_behavior = VM_BEHAVIOR_WILLNEED; |
| 950 | break; |
| 951 | case MADV_DONTNEED: |
| 952 | new_behavior = VM_BEHAVIOR_DONTNEED; |
| 953 | break; |
| 954 | case MADV_FREE: |
| 955 | new_behavior = VM_BEHAVIOR_FREE; |
| 956 | break; |
| 957 | case MADV_ZERO_WIRED_PAGES: |
| 958 | new_behavior = VM_BEHAVIOR_ZERO_WIRED_PAGES; |
| 959 | break; |
| 960 | case MADV_FREE_REUSABLE: |
| 961 | new_behavior = VM_BEHAVIOR_REUSABLE; |
| 962 | break; |
| 963 | case MADV_FREE_REUSE: |
| 964 | new_behavior = VM_BEHAVIOR_REUSE; |
| 965 | break; |
| 966 | case MADV_CAN_REUSE: |
| 967 | new_behavior = VM_BEHAVIOR_CAN_REUSE; |
| 968 | break; |
| 969 | case MADV_PAGEOUT: |
| 970 | #if MACH_ASSERT |
| 971 | new_behavior = VM_BEHAVIOR_PAGEOUT; |
| 972 | break; |
| 973 | #else /* MACH_ASSERT */ |
| 974 | return ENOTSUP; |
| 975 | #endif /* MACH_ASSERT */ |
| 976 | default: |
| 977 | return(EINVAL); |
| 978 | } |
| 979 | |
| 980 | start = (mach_vm_offset_t) uap->addr; |
| 981 | size = (mach_vm_size_t) uap->len; |
| 982 | |
| 983 | #if __arm64__ |
| 984 | if (start == 0 && |
| 985 | size != 0 && |
| 986 | (uap->behav == MADV_FREE || |
| 987 | uap->behav == MADV_FREE_REUSABLE)) { |
| 988 | printf("** FOURK_COMPAT: %d[%s] " |
| 989 | "failing madvise(0x%llx,0x%llx,%s)\n" , |
| 990 | p->p_pid, p->p_comm, start, size, |
| 991 | ((uap->behav == MADV_FREE_REUSABLE) |
| 992 | ? "MADV_FREE_REUSABLE" |
| 993 | : "MADV_FREE" )); |
| 994 | DTRACE_VM3(fourk_compat_madvise, |
| 995 | uint64_t, start, |
| 996 | uint64_t, size, |
| 997 | int, uap->behav); |
| 998 | return EINVAL; |
| 999 | } |
| 1000 | #endif /* __arm64__ */ |
| 1001 | |
| 1002 | user_map = current_map(); |
| 1003 | |
| 1004 | result = mach_vm_behavior_set(user_map, start, size, new_behavior); |
| 1005 | switch (result) { |
| 1006 | case KERN_SUCCESS: |
| 1007 | return 0; |
| 1008 | case KERN_INVALID_ADDRESS: |
| 1009 | return EINVAL; |
| 1010 | case KERN_NO_SPACE: |
| 1011 | return ENOMEM; |
| 1012 | } |
| 1013 | |
| 1014 | return EINVAL; |
| 1015 | } |
| 1016 | |
| 1017 | int |
| 1018 | mincore(__unused proc_t p, struct mincore_args *uap, __unused int32_t *retval) |
| 1019 | { |
| 1020 | mach_vm_offset_t addr = 0, first_addr = 0, end = 0, cur_end = 0; |
| 1021 | vm_map_t map = VM_MAP_NULL; |
| 1022 | user_addr_t vec = 0; |
| 1023 | int error = 0; |
| 1024 | int lastvecindex = 0; |
| 1025 | int mincoreinfo=0; |
| 1026 | int pqueryinfo = 0; |
| 1027 | unsigned int pqueryinfo_vec_size = 0; |
| 1028 | vm_page_info_basic_t info = NULL; |
| 1029 | mach_msg_type_number_t count = 0; |
| 1030 | char *kernel_vec = NULL; |
| 1031 | unsigned int req_vec_size_pages = 0, cur_vec_size_pages = 0, vecindex = 0; |
| 1032 | kern_return_t kr = KERN_SUCCESS; |
| 1033 | |
| 1034 | map = current_map(); |
| 1035 | |
| 1036 | /* |
| 1037 | * Make sure that the addresses presented are valid for user |
| 1038 | * mode. |
| 1039 | */ |
| 1040 | first_addr = addr = vm_map_trunc_page(uap->addr, |
| 1041 | vm_map_page_mask(map)); |
| 1042 | end = vm_map_round_page(uap->addr + uap->len, |
| 1043 | vm_map_page_mask(map)); |
| 1044 | |
| 1045 | if (end < addr) |
| 1046 | return (EINVAL); |
| 1047 | |
| 1048 | if (end == addr) |
| 1049 | return (0); |
| 1050 | |
| 1051 | /* |
| 1052 | * We are going to loop through the whole 'req_vec_size' pages |
| 1053 | * range in chunks of 'cur_vec_size'. |
| 1054 | */ |
| 1055 | |
| 1056 | req_vec_size_pages = (end - addr) >> PAGE_SHIFT; |
| 1057 | cur_vec_size_pages = MIN(req_vec_size_pages, (MAX_PAGE_RANGE_QUERY >> PAGE_SHIFT)); |
| 1058 | |
| 1059 | kernel_vec = (void*) _MALLOC(cur_vec_size_pages * sizeof(char), M_TEMP, M_WAITOK | M_ZERO); |
| 1060 | |
| 1061 | if (kernel_vec == NULL) { |
| 1062 | return (ENOMEM); |
| 1063 | } |
| 1064 | |
| 1065 | /* |
| 1066 | * Address of byte vector |
| 1067 | */ |
| 1068 | vec = uap->vec; |
| 1069 | |
| 1070 | pqueryinfo_vec_size = cur_vec_size_pages * sizeof(struct vm_page_info_basic); |
| 1071 | info = (void*) _MALLOC(pqueryinfo_vec_size, M_TEMP, M_WAITOK); |
| 1072 | |
| 1073 | if (info == NULL) { |
| 1074 | FREE(kernel_vec, M_TEMP); |
| 1075 | return (ENOMEM); |
| 1076 | } |
| 1077 | |
| 1078 | while (addr < end) { |
| 1079 | |
| 1080 | cur_end = addr + (cur_vec_size_pages * PAGE_SIZE_64); |
| 1081 | |
| 1082 | count = VM_PAGE_INFO_BASIC_COUNT; |
| 1083 | kr = vm_map_page_range_info_internal(map, |
| 1084 | addr, |
| 1085 | cur_end, |
| 1086 | VM_PAGE_INFO_BASIC, |
| 1087 | (vm_page_info_t) info, |
| 1088 | &count); |
| 1089 | |
| 1090 | assert(kr == KERN_SUCCESS); |
| 1091 | |
| 1092 | /* |
| 1093 | * Do this on a map entry basis so that if the pages are not |
| 1094 | * in the current processes address space, we can easily look |
| 1095 | * up the pages elsewhere. |
| 1096 | */ |
| 1097 | lastvecindex = -1; |
| 1098 | for( ; addr < cur_end; addr += PAGE_SIZE ) { |
| 1099 | |
| 1100 | pqueryinfo = info[lastvecindex + 1].disposition; |
| 1101 | |
| 1102 | mincoreinfo = 0; |
| 1103 | |
| 1104 | if (pqueryinfo & VM_PAGE_QUERY_PAGE_PRESENT) |
| 1105 | mincoreinfo |= MINCORE_INCORE; |
| 1106 | if (pqueryinfo & VM_PAGE_QUERY_PAGE_REF) |
| 1107 | mincoreinfo |= MINCORE_REFERENCED; |
| 1108 | if (pqueryinfo & VM_PAGE_QUERY_PAGE_DIRTY) |
| 1109 | mincoreinfo |= MINCORE_MODIFIED; |
| 1110 | if (pqueryinfo & VM_PAGE_QUERY_PAGE_PAGED_OUT) |
| 1111 | mincoreinfo |= MINCORE_PAGED_OUT; |
| 1112 | if (pqueryinfo & VM_PAGE_QUERY_PAGE_COPIED) |
| 1113 | mincoreinfo |= MINCORE_COPIED; |
| 1114 | if ((pqueryinfo & VM_PAGE_QUERY_PAGE_EXTERNAL) == 0) |
| 1115 | mincoreinfo |= MINCORE_ANONYMOUS; |
| 1116 | /* |
| 1117 | * calculate index into user supplied byte vector |
| 1118 | */ |
| 1119 | vecindex = (addr - first_addr)>> PAGE_SHIFT; |
| 1120 | kernel_vec[vecindex] = (char)mincoreinfo; |
| 1121 | lastvecindex = vecindex; |
| 1122 | } |
| 1123 | |
| 1124 | |
| 1125 | assert(vecindex == (cur_vec_size_pages - 1)); |
| 1126 | |
| 1127 | error = copyout(kernel_vec, vec, cur_vec_size_pages * sizeof(char) /* a char per page */); |
| 1128 | |
| 1129 | if (error) { |
| 1130 | break; |
| 1131 | } |
| 1132 | |
| 1133 | /* |
| 1134 | * For the next chunk, we'll need: |
| 1135 | * - bump the location in the user buffer for our next disposition. |
| 1136 | * - new length |
| 1137 | * - starting address |
| 1138 | */ |
| 1139 | vec += cur_vec_size_pages * sizeof(char); |
| 1140 | req_vec_size_pages = (end - addr) >> PAGE_SHIFT; |
| 1141 | cur_vec_size_pages = MIN(req_vec_size_pages, (MAX_PAGE_RANGE_QUERY >> PAGE_SHIFT)); |
| 1142 | |
| 1143 | first_addr = addr; |
| 1144 | } |
| 1145 | |
| 1146 | FREE(kernel_vec, M_TEMP); |
| 1147 | FREE(info, M_TEMP); |
| 1148 | |
| 1149 | if (error) { |
| 1150 | return (EFAULT); |
| 1151 | } |
| 1152 | |
| 1153 | return (0); |
| 1154 | } |
| 1155 | |
| 1156 | int |
| 1157 | mlock(__unused proc_t p, struct mlock_args *uap, __unused int32_t *retvalval) |
| 1158 | { |
| 1159 | vm_map_t user_map; |
| 1160 | vm_map_offset_t addr; |
| 1161 | vm_map_size_t size, pageoff; |
| 1162 | kern_return_t result; |
| 1163 | |
| 1164 | AUDIT_ARG(addr, uap->addr); |
| 1165 | AUDIT_ARG(len, uap->len); |
| 1166 | |
| 1167 | addr = (vm_map_offset_t) uap->addr; |
| 1168 | size = (vm_map_size_t)uap->len; |
| 1169 | |
| 1170 | /* disable wrap around */ |
| 1171 | if (addr + size < addr) |
| 1172 | return (EINVAL); |
| 1173 | |
| 1174 | if (size == 0) |
| 1175 | return (0); |
| 1176 | |
| 1177 | user_map = current_map(); |
| 1178 | pageoff = (addr & vm_map_page_mask(user_map)); |
| 1179 | addr -= pageoff; |
| 1180 | size = vm_map_round_page(size+pageoff, vm_map_page_mask(user_map)); |
| 1181 | |
| 1182 | /* have to call vm_map_wire directly to pass "I don't know" protections */ |
| 1183 | result = vm_map_wire_kernel(user_map, addr, addr+size, VM_PROT_NONE, VM_KERN_MEMORY_MLOCK, TRUE); |
| 1184 | |
| 1185 | if (result == KERN_RESOURCE_SHORTAGE) |
| 1186 | return EAGAIN; |
| 1187 | else if (result == KERN_PROTECTION_FAILURE) |
| 1188 | return EACCES; |
| 1189 | else if (result != KERN_SUCCESS) |
| 1190 | return ENOMEM; |
| 1191 | |
| 1192 | return 0; /* KERN_SUCCESS */ |
| 1193 | } |
| 1194 | |
| 1195 | int |
| 1196 | munlock(__unused proc_t p, struct munlock_args *uap, __unused int32_t *retval) |
| 1197 | { |
| 1198 | mach_vm_offset_t addr; |
| 1199 | mach_vm_size_t size; |
| 1200 | vm_map_t user_map; |
| 1201 | kern_return_t result; |
| 1202 | |
| 1203 | AUDIT_ARG(addr, uap->addr); |
| 1204 | AUDIT_ARG(addr, uap->len); |
| 1205 | |
| 1206 | addr = (mach_vm_offset_t) uap->addr; |
| 1207 | size = (mach_vm_size_t)uap->len; |
| 1208 | user_map = current_map(); |
| 1209 | |
| 1210 | /* JMM - need to remove all wirings by spec - this just removes one */ |
| 1211 | result = mach_vm_wire_kernel(host_priv_self(), user_map, addr, size, VM_PROT_NONE, VM_KERN_MEMORY_MLOCK); |
| 1212 | return (result == KERN_SUCCESS ? 0 : ENOMEM); |
| 1213 | } |
| 1214 | |
| 1215 | |
| 1216 | int |
| 1217 | mlockall(__unused proc_t p, __unused struct mlockall_args *uap, __unused int32_t *retval) |
| 1218 | { |
| 1219 | return (ENOSYS); |
| 1220 | } |
| 1221 | |
| 1222 | int |
| 1223 | munlockall(__unused proc_t p, __unused struct munlockall_args *uap, __unused int32_t *retval) |
| 1224 | { |
| 1225 | return(ENOSYS); |
| 1226 | } |
| 1227 | |
| 1228 | #if CONFIG_CODE_DECRYPTION |
| 1229 | int |
| 1230 | mremap_encrypted(__unused struct proc *p, struct mremap_encrypted_args *uap, __unused int32_t *retval) |
| 1231 | { |
| 1232 | mach_vm_offset_t user_addr; |
| 1233 | mach_vm_size_t user_size; |
| 1234 | kern_return_t result; |
| 1235 | vm_map_t user_map; |
| 1236 | uint32_t cryptid; |
| 1237 | cpu_type_t cputype; |
| 1238 | cpu_subtype_t cpusubtype; |
| 1239 | pager_crypt_info_t crypt_info; |
| 1240 | const char * cryptname = 0; |
| 1241 | char *vpath; |
| 1242 | int len, ret; |
| 1243 | struct proc_regioninfo_internal pinfo; |
| 1244 | vnode_t vp; |
| 1245 | uintptr_t vnodeaddr; |
| 1246 | uint32_t vid; |
| 1247 | |
| 1248 | AUDIT_ARG(addr, uap->addr); |
| 1249 | AUDIT_ARG(len, uap->len); |
| 1250 | |
| 1251 | user_map = current_map(); |
| 1252 | user_addr = (mach_vm_offset_t) uap->addr; |
| 1253 | user_size = (mach_vm_size_t) uap->len; |
| 1254 | |
| 1255 | cryptid = uap->cryptid; |
| 1256 | cputype = uap->cputype; |
| 1257 | cpusubtype = uap->cpusubtype; |
| 1258 | |
| 1259 | if (user_addr & vm_map_page_mask(user_map)) { |
| 1260 | /* UNIX SPEC: user address is not page-aligned, return EINVAL */ |
| 1261 | return EINVAL; |
| 1262 | } |
| 1263 | |
| 1264 | switch(cryptid) { |
| 1265 | case 0: |
| 1266 | /* not encrypted, just an empty load command */ |
| 1267 | return 0; |
| 1268 | case 1: |
| 1269 | cryptname="com.apple.unfree" ; |
| 1270 | break; |
| 1271 | case 0x10: |
| 1272 | /* some random cryptid that you could manually put into |
| 1273 | * your binary if you want NULL */ |
| 1274 | cryptname="com.apple.null" ; |
| 1275 | break; |
| 1276 | default: |
| 1277 | return EINVAL; |
| 1278 | } |
| 1279 | |
| 1280 | if (NULL == text_crypter_create) return ENOTSUP; |
| 1281 | |
| 1282 | ret = fill_procregioninfo_onlymappedvnodes( proc_task(p), user_addr, &pinfo, &vnodeaddr, &vid); |
| 1283 | if (ret == 0 || !vnodeaddr) { |
| 1284 | /* No really, this returns 0 if the memory address is not backed by a file */ |
| 1285 | return (EINVAL); |
| 1286 | } |
| 1287 | |
| 1288 | vp = (vnode_t)vnodeaddr; |
| 1289 | if ((vnode_getwithvid(vp, vid)) == 0) { |
| 1290 | MALLOC_ZONE(vpath, char *, MAXPATHLEN, M_NAMEI, M_WAITOK); |
| 1291 | if(vpath == NULL) { |
| 1292 | vnode_put(vp); |
| 1293 | return (ENOMEM); |
| 1294 | } |
| 1295 | |
| 1296 | len = MAXPATHLEN; |
| 1297 | ret = vn_getpath(vp, vpath, &len); |
| 1298 | if(ret) { |
| 1299 | FREE_ZONE(vpath, MAXPATHLEN, M_NAMEI); |
| 1300 | vnode_put(vp); |
| 1301 | return (ret); |
| 1302 | } |
| 1303 | |
| 1304 | vnode_put(vp); |
| 1305 | } else { |
| 1306 | return (EINVAL); |
| 1307 | } |
| 1308 | |
| 1309 | #if 0 |
| 1310 | kprintf("%s vpath %s cryptid 0x%08x cputype 0x%08x cpusubtype 0x%08x range 0x%016llx size 0x%016llx\n" , |
| 1311 | __FUNCTION__, vpath, cryptid, cputype, cpusubtype, (uint64_t)user_addr, (uint64_t)user_size); |
| 1312 | #endif |
| 1313 | |
| 1314 | /* set up decrypter first */ |
| 1315 | crypt_file_data_t crypt_data = { |
| 1316 | .filename = vpath, |
| 1317 | .cputype = cputype, |
| 1318 | .cpusubtype = cpusubtype }; |
| 1319 | result = text_crypter_create(&crypt_info, cryptname, (void*)&crypt_data); |
| 1320 | #if VM_MAP_DEBUG_APPLE_PROTECT |
| 1321 | if (vm_map_debug_apple_protect) { |
| 1322 | printf("APPLE_PROTECT: %d[%s] map %p [0x%llx:0x%llx] %s(%s) -> 0x%x\n" , |
| 1323 | p->p_pid, p->p_comm, |
| 1324 | user_map, |
| 1325 | (uint64_t) user_addr, |
| 1326 | (uint64_t) (user_addr + user_size), |
| 1327 | __FUNCTION__, vpath, result); |
| 1328 | } |
| 1329 | #endif /* VM_MAP_DEBUG_APPLE_PROTECT */ |
| 1330 | FREE_ZONE(vpath, MAXPATHLEN, M_NAMEI); |
| 1331 | |
| 1332 | if(result) { |
| 1333 | printf("%s: unable to create decrypter %s, kr=%d\n" , |
| 1334 | __FUNCTION__, cryptname, result); |
| 1335 | if (result == kIOReturnNotPrivileged) { |
| 1336 | /* text encryption returned decryption failure */ |
| 1337 | return (EPERM); |
| 1338 | } else { |
| 1339 | return (ENOMEM); |
| 1340 | } |
| 1341 | } |
| 1342 | |
| 1343 | /* now remap using the decrypter */ |
| 1344 | vm_object_offset_t crypto_backing_offset; |
| 1345 | crypto_backing_offset = -1; /* i.e. use map entry's offset */ |
| 1346 | result = vm_map_apple_protected(user_map, |
| 1347 | user_addr, |
| 1348 | user_addr+user_size, |
| 1349 | crypto_backing_offset, |
| 1350 | &crypt_info); |
| 1351 | if (result) { |
| 1352 | printf("%s: mapping failed with %d\n" , __FUNCTION__, result); |
| 1353 | } |
| 1354 | |
| 1355 | if (result) { |
| 1356 | return (EPERM); |
| 1357 | } |
| 1358 | return 0; |
| 1359 | } |
| 1360 | #endif /* CONFIG_CODE_DECRYPTION */ |
| 1361 | |