1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1989, 1992, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software developed by the Computer Systems 8 * Engineering group at Lawrence Berkeley Laboratory under DARPA contract 9 * BG 91-66 and contributed to Berkeley. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 */ 35 36 #include <sys/cdefs.h> 37 __SCCSID("@(#)kvm.c 8.2 (Berkeley) 2/13/94"); 38 39 #include <sys/param.h> 40 #include <sys/fnv_hash.h> 41 42 #define _WANT_VNET 43 44 #include <sys/user.h> 45 #include <sys/linker.h> 46 #include <sys/pcpu.h> 47 #include <sys/stat.h> 48 #include <sys/sysctl.h> 49 #include <sys/mman.h> 50 51 #include <stdbool.h> 52 #include <net/vnet.h> 53 54 #include <fcntl.h> 55 #include <kvm.h> 56 #include <limits.h> 57 #include <paths.h> 58 #include <stdint.h> 59 #include <stdio.h> 60 #include <stdlib.h> 61 #include <string.h> 62 #include <unistd.h> 63 64 #include "kvm_private.h" 65 66 SET_DECLARE(kvm_arch, struct kvm_arch); 67 68 static char _kd_is_null[] = ""; 69 70 char * 71 kvm_geterr(kvm_t *kd) 72 { 73 74 if (kd == NULL) 75 return (_kd_is_null); 76 return (kd->errbuf); 77 } 78 79 static int 80 _kvm_read_kernel_ehdr(kvm_t *kd) 81 { 82 Elf *elf; 83 84 if (elf_version(EV_CURRENT) == EV_NONE) { 85 _kvm_err(kd, kd->program, "Unsupported libelf"); 86 return (-1); 87 } 88 elf = elf_begin(kd->nlfd, ELF_C_READ, NULL); 89 if (elf == NULL) { 90 _kvm_err(kd, kd->program, "%s", elf_errmsg(0)); 91 return (-1); 92 } 93 if (elf_kind(elf) != ELF_K_ELF) { 94 _kvm_err(kd, kd->program, "kernel is not an ELF file"); 95 return (-1); 96 } 97 if (gelf_getehdr(elf, &kd->nlehdr) == NULL) { 98 _kvm_err(kd, kd->program, "%s", elf_errmsg(0)); 99 elf_end(elf); 100 return (-1); 101 } 102 elf_end(elf); 103 104 switch (kd->nlehdr.e_ident[EI_DATA]) { 105 case ELFDATA2LSB: 106 case ELFDATA2MSB: 107 return (0); 108 default: 109 _kvm_err(kd, kd->program, 110 "unsupported ELF data encoding for kernel"); 111 return (-1); 112 } 113 } 114 115 static kvm_t * 116 _kvm_open(kvm_t *kd, const char *uf, const char *mf, int flag, char *errout) 117 { 118 struct kvm_arch **parch; 119 struct stat st; 120 121 kd->vmfd = -1; 122 kd->pmfd = -1; 123 kd->nlfd = -1; 124 kd->vmst = NULL; 125 kd->procbase = NULL; 126 kd->argspc = NULL; 127 kd->argv = NULL; 128 129 if (uf == NULL) 130 uf = getbootfile(); 131 else if (strlen(uf) >= MAXPATHLEN) { 132 _kvm_err(kd, kd->program, "exec file name too long"); 133 goto failed; 134 } 135 if (flag & ~O_RDWR) { 136 _kvm_err(kd, kd->program, "bad flags arg"); 137 goto failed; 138 } 139 if (mf == NULL) 140 mf = _PATH_MEM; 141 142 if ((kd->pmfd = open(mf, flag | O_CLOEXEC, 0)) < 0) { 143 _kvm_syserr(kd, kd->program, "%s", mf); 144 goto failed; 145 } 146 if (fstat(kd->pmfd, &st) < 0) { 147 _kvm_syserr(kd, kd->program, "%s", mf); 148 goto failed; 149 } 150 if (S_ISREG(st.st_mode) && st.st_size <= 0) { 151 errno = EINVAL; 152 _kvm_syserr(kd, kd->program, "empty file"); 153 goto failed; 154 } 155 if (S_ISCHR(st.st_mode)) { 156 /* 157 * If this is a character special device, then check that 158 * it's /dev/mem. If so, open kmem too. (Maybe we should 159 * make it work for either /dev/mem or /dev/kmem -- in either 160 * case you're working with a live kernel.) 161 */ 162 if (strcmp(mf, _PATH_DEVNULL) == 0) { 163 kd->vmfd = open(_PATH_DEVNULL, O_RDONLY | O_CLOEXEC); 164 return (kd); 165 } else if (strcmp(mf, _PATH_MEM) == 0) { 166 if ((kd->vmfd = open(_PATH_KMEM, flag | O_CLOEXEC)) < 167 0) { 168 _kvm_syserr(kd, kd->program, "%s", _PATH_KMEM); 169 goto failed; 170 } 171 return (kd); 172 } 173 } 174 175 /* 176 * This is either a crash dump or a remote live system with its physical 177 * memory fully accessible via a special device. 178 * Open the namelist fd and determine the architecture. 179 */ 180 if ((kd->nlfd = open(uf, O_RDONLY | O_CLOEXEC, 0)) < 0) { 181 _kvm_syserr(kd, kd->program, "%s", uf); 182 goto failed; 183 } 184 if (_kvm_read_kernel_ehdr(kd) < 0) 185 goto failed; 186 if (strncmp(mf, _PATH_FWMEM, strlen(_PATH_FWMEM)) == 0 || 187 strncmp(mf, _PATH_DEVVMM, strlen(_PATH_DEVVMM)) == 0) { 188 kd->rawdump = 1; 189 kd->writable = 1; 190 } 191 SET_FOREACH(parch, kvm_arch) { 192 if ((*parch)->ka_probe(kd)) { 193 kd->arch = *parch; 194 break; 195 } 196 } 197 if (kd->arch == NULL) { 198 _kvm_err(kd, kd->program, "unsupported architecture"); 199 goto failed; 200 } 201 202 /* 203 * Non-native kernels require a symbol resolver. 204 */ 205 if (!kd->arch->ka_native(kd) && kd->resolve_symbol == NULL) { 206 _kvm_err(kd, kd->program, 207 "non-native kernel requires a symbol resolver"); 208 goto failed; 209 } 210 211 /* 212 * Initialize the virtual address translation machinery. 213 */ 214 if (kd->arch->ka_initvtop(kd) < 0) 215 goto failed; 216 return (kd); 217 failed: 218 /* 219 * Copy out the error if doing sane error semantics. 220 */ 221 if (errout != NULL) 222 strlcpy(errout, kd->errbuf, _POSIX2_LINE_MAX); 223 (void)kvm_close(kd); 224 return (NULL); 225 } 226 227 kvm_t * 228 kvm_openfiles(const char *uf, const char *mf, const char *sf __unused, int flag, 229 char *errout) 230 { 231 kvm_t *kd; 232 233 if ((kd = calloc(1, sizeof(*kd))) == NULL) { 234 if (errout != NULL) 235 (void)strlcpy(errout, strerror(errno), 236 _POSIX2_LINE_MAX); 237 return (NULL); 238 } 239 return (_kvm_open(kd, uf, mf, flag, errout)); 240 } 241 242 kvm_t * 243 kvm_open(const char *uf, const char *mf, const char *sf __unused, int flag, 244 const char *errstr) 245 { 246 kvm_t *kd; 247 248 if ((kd = calloc(1, sizeof(*kd))) == NULL) { 249 if (errstr != NULL) 250 (void)fprintf(stderr, "%s: %s\n", 251 errstr, strerror(errno)); 252 return (NULL); 253 } 254 kd->program = errstr; 255 return (_kvm_open(kd, uf, mf, flag, NULL)); 256 } 257 258 kvm_t * 259 kvm_open2(const char *uf, const char *mf, int flag, char *errout, 260 int (*resolver)(const char *, kvaddr_t *)) 261 { 262 kvm_t *kd; 263 264 if ((kd = calloc(1, sizeof(*kd))) == NULL) { 265 if (errout != NULL) 266 (void)strlcpy(errout, strerror(errno), 267 _POSIX2_LINE_MAX); 268 return (NULL); 269 } 270 kd->resolve_symbol = resolver; 271 return (_kvm_open(kd, uf, mf, flag, errout)); 272 } 273 274 int 275 kvm_close(kvm_t *kd) 276 { 277 int error = 0; 278 279 if (kd == NULL) { 280 errno = EINVAL; 281 return (-1); 282 } 283 if (kd->vmst != NULL) 284 kd->arch->ka_freevtop(kd); 285 if (kd->pmfd >= 0) 286 error |= close(kd->pmfd); 287 if (kd->vmfd >= 0) 288 error |= close(kd->vmfd); 289 if (kd->nlfd >= 0) 290 error |= close(kd->nlfd); 291 if (kd->procbase != 0) 292 free((void *)kd->procbase); 293 if (kd->argbuf != 0) 294 free((void *) kd->argbuf); 295 if (kd->argspc != 0) 296 free((void *) kd->argspc); 297 if (kd->argv != 0) 298 free((void *)kd->argv); 299 if (kd->dpcpu_initialized != 0) 300 free(kd->dpcpu_off); 301 if (kd->pt_map != NULL) 302 free(kd->pt_map); 303 if (kd->page_map != NULL) 304 free(kd->page_map); 305 if (kd->sparse_map != MAP_FAILED && kd->sparse_map != NULL) 306 munmap(kd->sparse_map, kd->pt_sparse_size); 307 free((void *)kd); 308 309 return (error); 310 } 311 312 int 313 kvm_nlist2(kvm_t *kd, struct kvm_nlist *nl) 314 { 315 316 /* 317 * If called via the public interface, permit initialization of 318 * further virtualized modules on demand. 319 */ 320 return (_kvm_nlist(kd, nl, 1)); 321 } 322 323 int 324 kvm_nlist(kvm_t *kd, struct nlist *nl) 325 { 326 struct kvm_nlist *kl; 327 int count, i, nfail; 328 329 /* 330 * Avoid reporting truncated addresses by failing for non-native 331 * cores. 332 */ 333 if (!kvm_native(kd)) { 334 _kvm_err(kd, kd->program, "kvm_nlist of non-native vmcore"); 335 return (-1); 336 } 337 338 for (count = 0; nl[count].n_name != NULL && nl[count].n_name[0] != '\0'; 339 count++) 340 ; 341 if (count == 0) 342 return (0); 343 kl = calloc(count + 1, sizeof(*kl)); 344 if (kl == NULL) { 345 _kvm_err(kd, kd->program, "cannot allocate memory"); 346 return (-1); 347 } 348 for (i = 0; i < count; i++) 349 kl[i].n_name = nl[i].n_name; 350 nfail = kvm_nlist2(kd, kl); 351 for (i = 0; i < count; i++) { 352 nl[i].n_type = kl[i].n_type; 353 nl[i].n_other = 0; 354 nl[i].n_desc = 0; 355 nl[i].n_value = kl[i].n_value; 356 } 357 free(kl); 358 return (nfail); 359 } 360 361 ssize_t 362 kvm_read(kvm_t *kd, u_long kva, void *buf, size_t len) 363 { 364 365 return (kvm_read2(kd, kva, buf, len)); 366 } 367 368 ssize_t 369 kvm_read2(kvm_t *kd, kvaddr_t kva, void *buf, size_t len) 370 { 371 int cc; 372 ssize_t cr; 373 off_t pa; 374 char *cp; 375 376 if (ISALIVE(kd)) { 377 /* 378 * We're using /dev/kmem. Just read straight from the 379 * device and let the active kernel do the address translation. 380 */ 381 errno = 0; 382 if (lseek(kd->vmfd, (off_t)kva, 0) == -1 && errno != 0) { 383 _kvm_err(kd, 0, "invalid address (0x%jx)", 384 (uintmax_t)kva); 385 return (-1); 386 } 387 cr = read(kd->vmfd, buf, len); 388 if (cr < 0) { 389 _kvm_syserr(kd, 0, "kvm_read"); 390 return (-1); 391 } else if (cr < (ssize_t)len) 392 _kvm_err(kd, kd->program, "short read"); 393 return (cr); 394 } 395 396 cp = buf; 397 while (len > 0) { 398 cc = kd->arch->ka_kvatop(kd, kva, &pa); 399 if (cc == 0) 400 return (-1); 401 if (cc > (ssize_t)len) 402 cc = len; 403 errno = 0; 404 if (lseek(kd->pmfd, pa, 0) == -1 && errno != 0) { 405 _kvm_syserr(kd, 0, _PATH_MEM); 406 break; 407 } 408 cr = read(kd->pmfd, cp, cc); 409 if (cr < 0) { 410 _kvm_syserr(kd, kd->program, "kvm_read"); 411 break; 412 } 413 /* 414 * If ka_kvatop returns a bogus value or our core file is 415 * truncated, we might wind up seeking beyond the end of the 416 * core file in which case the read will return 0 (EOF). 417 */ 418 if (cr == 0) 419 break; 420 cp += cr; 421 kva += cr; 422 len -= cr; 423 } 424 425 return (cp - (char *)buf); 426 } 427 428 ssize_t 429 kvm_write(kvm_t *kd, u_long kva, const void *buf, size_t len) 430 { 431 int cc; 432 ssize_t cw; 433 off_t pa; 434 const char *cp; 435 436 if (!ISALIVE(kd) && !kd->writable) { 437 _kvm_err(kd, kd->program, 438 "kvm_write not implemented for dead kernels"); 439 return (-1); 440 } 441 442 if (ISALIVE(kd)) { 443 /* 444 * Just like kvm_read, only we write. 445 */ 446 errno = 0; 447 if (lseek(kd->vmfd, (off_t)kva, 0) == -1 && errno != 0) { 448 _kvm_err(kd, 0, "invalid address (%lx)", kva); 449 return (-1); 450 } 451 cc = write(kd->vmfd, buf, len); 452 if (cc < 0) { 453 _kvm_syserr(kd, 0, "kvm_write"); 454 return (-1); 455 } else if ((size_t)cc < len) 456 _kvm_err(kd, kd->program, "short write"); 457 return (cc); 458 } 459 460 cp = buf; 461 while (len > 0) { 462 cc = kd->arch->ka_kvatop(kd, kva, &pa); 463 if (cc == 0) 464 return (-1); 465 if (cc > (ssize_t)len) 466 cc = len; 467 errno = 0; 468 if (lseek(kd->pmfd, pa, 0) == -1 && errno != 0) { 469 _kvm_syserr(kd, 0, _PATH_MEM); 470 break; 471 } 472 cw = write(kd->pmfd, cp, cc); 473 if (cw < 0) { 474 _kvm_syserr(kd, kd->program, "kvm_write"); 475 break; 476 } 477 /* 478 * If ka_kvatop returns a bogus value or our core file is 479 * truncated, we might wind up seeking beyond the end of the 480 * core file in which case the read will return 0 (EOF). 481 */ 482 if (cw == 0) 483 break; 484 cp += cw; 485 kva += cw; 486 len -= cw; 487 } 488 489 return (cp - (const char *)buf); 490 } 491 492 int 493 kvm_native(kvm_t *kd) 494 { 495 496 if (ISALIVE(kd)) 497 return (1); 498 return (kd->arch->ka_native(kd)); 499 } 500 501 int 502 kvm_walk_pages(kvm_t *kd, kvm_walk_pages_cb_t *cb, void *closure) 503 { 504 505 if (kd->arch->ka_walk_pages == NULL) 506 return (0); 507 508 return (kd->arch->ka_walk_pages(kd, cb, closure)); 509 } 510 511 kssize_t 512 kvm_kerndisp(kvm_t *kd) 513 { 514 unsigned long kernbase, rel_kernbase; 515 size_t kernbase_len = sizeof(kernbase); 516 size_t rel_kernbase_len = sizeof(rel_kernbase); 517 518 if (ISALIVE(kd)) { 519 if (sysctlbyname("kern.base_address", &kernbase, 520 &kernbase_len, NULL, 0) == -1) { 521 _kvm_syserr(kd, kd->program, 522 "failed to get kernel base address"); 523 return (0); 524 } 525 if (sysctlbyname("kern.relbase_address", &rel_kernbase, 526 &rel_kernbase_len, NULL, 0) == -1) { 527 _kvm_syserr(kd, kd->program, 528 "failed to get relocated kernel base address"); 529 return (0); 530 } 531 return (rel_kernbase - kernbase); 532 } 533 534 if (kd->arch->ka_kerndisp == NULL) 535 return (0); 536 537 return (kd->arch->ka_kerndisp(kd)); 538 } 539