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