1 /*- 2 * Copyright (c) 2002 Marcel Moolenaar 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 16 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 17 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 18 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 19 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 20 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 21 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 22 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 23 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 24 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 25 */ 26 27 #include <sys/cdefs.h> 28 __FBSDID("$FreeBSD$"); 29 30 #include "opt_watchdog.h" 31 32 #include <sys/param.h> 33 #include <sys/systm.h> 34 #include <sys/conf.h> 35 #include <sys/cons.h> 36 #include <sys/sysctl.h> 37 #include <sys/kernel.h> 38 #include <sys/kerneldump.h> 39 #ifdef SW_WATCHDOG 40 #include <sys/watchdog.h> 41 #endif 42 #include <vm/vm.h> 43 #include <vm/pmap.h> 44 #include <machine/elf.h> 45 #include <machine/md_var.h> 46 47 #ifdef __amd64__ 48 #define KERNELDUMP_VERSION KERNELDUMP_AMD64_VERSION 49 #define EM_VALUE EM_X86_64 50 #else 51 #define KERNELDUMP_VERSION KERNELDUMP_I386_VERSION 52 #define EM_VALUE EM_386 53 #endif 54 55 CTASSERT(sizeof(struct kerneldumpheader) == 512); 56 57 int do_minidump = 1; 58 TUNABLE_INT("debug.minidump", &do_minidump); 59 SYSCTL_INT(_debug, OID_AUTO, minidump, CTLFLAG_RW, &do_minidump, 0, 60 "Enable mini crash dumps"); 61 62 /* 63 * Don't touch the first SIZEOF_METADATA bytes on the dump device. This 64 * is to protect us from metadata and to protect metadata from us. 65 */ 66 #define SIZEOF_METADATA (64*1024) 67 68 #define MD_ALIGN(x) (((off_t)(x) + PAGE_MASK) & ~PAGE_MASK) 69 #define DEV_ALIGN(x) (((off_t)(x) + (DEV_BSIZE-1)) & ~(DEV_BSIZE-1)) 70 71 struct md_pa { 72 vm_paddr_t md_start; 73 vm_paddr_t md_size; 74 }; 75 76 typedef int callback_t(struct md_pa *, int, void *); 77 78 static struct kerneldumpheader kdh; 79 static off_t dumplo, fileofs; 80 81 /* Handle buffered writes. */ 82 static char buffer[DEV_BSIZE]; 83 static size_t fragsz; 84 85 /* 20 phys_avail entry pairs correspond to 10 md_pa's */ 86 static struct md_pa dump_map[10]; 87 88 static void 89 md_pa_init(void) 90 { 91 int n, idx; 92 93 bzero(dump_map, sizeof(dump_map)); 94 for (n = 0; n < sizeof(dump_map) / sizeof(dump_map[0]); n++) { 95 idx = n * 2; 96 if (dump_avail[idx] == 0 && dump_avail[idx + 1] == 0) 97 break; 98 dump_map[n].md_start = dump_avail[idx]; 99 dump_map[n].md_size = dump_avail[idx + 1] - dump_avail[idx]; 100 } 101 } 102 103 static struct md_pa * 104 md_pa_first(void) 105 { 106 107 return (&dump_map[0]); 108 } 109 110 static struct md_pa * 111 md_pa_next(struct md_pa *mdp) 112 { 113 114 mdp++; 115 if (mdp->md_size == 0) 116 mdp = NULL; 117 return (mdp); 118 } 119 120 static int 121 buf_write(struct dumperinfo *di, char *ptr, size_t sz) 122 { 123 size_t len; 124 int error; 125 126 while (sz) { 127 len = DEV_BSIZE - fragsz; 128 if (len > sz) 129 len = sz; 130 bcopy(ptr, buffer + fragsz, len); 131 fragsz += len; 132 ptr += len; 133 sz -= len; 134 if (fragsz == DEV_BSIZE) { 135 error = dump_write(di, buffer, 0, dumplo, 136 DEV_BSIZE); 137 if (error) 138 return error; 139 dumplo += DEV_BSIZE; 140 fragsz = 0; 141 } 142 } 143 144 return (0); 145 } 146 147 static int 148 buf_flush(struct dumperinfo *di) 149 { 150 int error; 151 152 if (fragsz == 0) 153 return (0); 154 155 error = dump_write(di, buffer, 0, dumplo, DEV_BSIZE); 156 dumplo += DEV_BSIZE; 157 fragsz = 0; 158 return (error); 159 } 160 161 #define PG2MB(pgs) ((pgs + (1 << 8) - 1) >> 8) 162 163 static int 164 cb_dumpdata(struct md_pa *mdp, int seqnr, void *arg) 165 { 166 struct dumperinfo *di = (struct dumperinfo*)arg; 167 vm_paddr_t a, pa; 168 void *va; 169 uint64_t pgs; 170 size_t counter, sz, chunk; 171 int i, c, error, twiddle; 172 u_int maxdumppgs; 173 174 error = 0; /* catch case in which chunk size is 0 */ 175 counter = 0; /* Update twiddle every 16MB */ 176 twiddle = 0; 177 va = 0; 178 pgs = mdp->md_size / PAGE_SIZE; 179 pa = mdp->md_start; 180 maxdumppgs = min(di->maxiosize / PAGE_SIZE, MAXDUMPPGS); 181 if (maxdumppgs == 0) /* seatbelt */ 182 maxdumppgs = 1; 183 184 printf(" chunk %d: %juMB (%ju pages)", seqnr, (uintmax_t)PG2MB(pgs), 185 (uintmax_t)pgs); 186 187 while (pgs) { 188 chunk = pgs; 189 if (chunk > maxdumppgs) 190 chunk = maxdumppgs; 191 sz = chunk << PAGE_SHIFT; 192 counter += sz; 193 if (counter >> 24) { 194 printf(" %ju", (uintmax_t)PG2MB(pgs)); 195 counter &= (1<<24) - 1; 196 } 197 for (i = 0; i < chunk; i++) { 198 a = pa + i * PAGE_SIZE; 199 va = pmap_kenter_temporary(trunc_page(a), i); 200 } 201 #ifdef SW_WATCHDOG 202 wdog_kern_pat(WD_LASTVAL); 203 #endif 204 error = dump_write(di, va, 0, dumplo, sz); 205 if (error) 206 break; 207 dumplo += sz; 208 pgs -= chunk; 209 pa += sz; 210 211 /* Check for user abort. */ 212 c = cncheckc(); 213 if (c == 0x03) 214 return (ECANCELED); 215 if (c != -1) 216 printf(" (CTRL-C to abort) "); 217 } 218 printf(" ... %s\n", (error) ? "fail" : "ok"); 219 return (error); 220 } 221 222 static int 223 cb_dumphdr(struct md_pa *mdp, int seqnr, void *arg) 224 { 225 struct dumperinfo *di = (struct dumperinfo*)arg; 226 Elf_Phdr phdr; 227 uint64_t size; 228 int error; 229 230 size = mdp->md_size; 231 bzero(&phdr, sizeof(phdr)); 232 phdr.p_type = PT_LOAD; 233 phdr.p_flags = PF_R; /* XXX */ 234 phdr.p_offset = fileofs; 235 phdr.p_vaddr = mdp->md_start; 236 phdr.p_paddr = mdp->md_start; 237 phdr.p_filesz = size; 238 phdr.p_memsz = size; 239 phdr.p_align = PAGE_SIZE; 240 241 error = buf_write(di, (char*)&phdr, sizeof(phdr)); 242 fileofs += phdr.p_filesz; 243 return (error); 244 } 245 246 static int 247 cb_size(struct md_pa *mdp, int seqnr, void *arg) 248 { 249 uint64_t *sz = (uint64_t*)arg; 250 251 *sz += (uint64_t)mdp->md_size; 252 return (0); 253 } 254 255 static int 256 foreach_chunk(callback_t cb, void *arg) 257 { 258 struct md_pa *mdp; 259 int error, seqnr; 260 261 seqnr = 0; 262 mdp = md_pa_first(); 263 while (mdp != NULL) { 264 error = (*cb)(mdp, seqnr++, arg); 265 if (error) 266 return (-error); 267 mdp = md_pa_next(mdp); 268 } 269 return (seqnr); 270 } 271 272 void 273 dumpsys(struct dumperinfo *di) 274 { 275 Elf_Ehdr ehdr; 276 uint64_t dumpsize; 277 off_t hdrgap; 278 size_t hdrsz; 279 int error; 280 281 if (do_minidump) { 282 minidumpsys(di); 283 return; 284 } 285 bzero(&ehdr, sizeof(ehdr)); 286 ehdr.e_ident[EI_MAG0] = ELFMAG0; 287 ehdr.e_ident[EI_MAG1] = ELFMAG1; 288 ehdr.e_ident[EI_MAG2] = ELFMAG2; 289 ehdr.e_ident[EI_MAG3] = ELFMAG3; 290 ehdr.e_ident[EI_CLASS] = ELF_CLASS; 291 #if BYTE_ORDER == LITTLE_ENDIAN 292 ehdr.e_ident[EI_DATA] = ELFDATA2LSB; 293 #else 294 ehdr.e_ident[EI_DATA] = ELFDATA2MSB; 295 #endif 296 ehdr.e_ident[EI_VERSION] = EV_CURRENT; 297 ehdr.e_ident[EI_OSABI] = ELFOSABI_STANDALONE; /* XXX big picture? */ 298 ehdr.e_type = ET_CORE; 299 ehdr.e_machine = EM_VALUE; 300 ehdr.e_phoff = sizeof(ehdr); 301 ehdr.e_flags = 0; 302 ehdr.e_ehsize = sizeof(ehdr); 303 ehdr.e_phentsize = sizeof(Elf_Phdr); 304 ehdr.e_shentsize = sizeof(Elf_Shdr); 305 306 md_pa_init(); 307 308 /* Calculate dump size. */ 309 dumpsize = 0L; 310 ehdr.e_phnum = foreach_chunk(cb_size, &dumpsize); 311 hdrsz = ehdr.e_phoff + ehdr.e_phnum * ehdr.e_phentsize; 312 fileofs = MD_ALIGN(hdrsz); 313 dumpsize += fileofs; 314 hdrgap = fileofs - DEV_ALIGN(hdrsz); 315 316 /* Determine dump offset on device. */ 317 if (di->mediasize < SIZEOF_METADATA + dumpsize + sizeof(kdh) * 2) { 318 error = ENOSPC; 319 goto fail; 320 } 321 dumplo = di->mediaoffset + di->mediasize - dumpsize; 322 dumplo -= sizeof(kdh) * 2; 323 324 mkdumpheader(&kdh, KERNELDUMPMAGIC, KERNELDUMP_VERSION, dumpsize, 325 di->blocksize); 326 327 printf("Dumping %llu MB (%d chunks)\n", (long long)dumpsize >> 20, 328 ehdr.e_phnum); 329 330 /* Dump leader */ 331 error = dump_write(di, &kdh, 0, dumplo, sizeof(kdh)); 332 if (error) 333 goto fail; 334 dumplo += sizeof(kdh); 335 336 /* Dump ELF header */ 337 error = buf_write(di, (char*)&ehdr, sizeof(ehdr)); 338 if (error) 339 goto fail; 340 341 /* Dump program headers */ 342 error = foreach_chunk(cb_dumphdr, di); 343 if (error < 0) 344 goto fail; 345 buf_flush(di); 346 347 /* 348 * All headers are written using blocked I/O, so we know the 349 * current offset is (still) block aligned. Skip the alignement 350 * in the file to have the segment contents aligned at page 351 * boundary. We cannot use MD_ALIGN on dumplo, because we don't 352 * care and may very well be unaligned within the dump device. 353 */ 354 dumplo += hdrgap; 355 356 /* Dump memory chunks (updates dumplo) */ 357 error = foreach_chunk(cb_dumpdata, di); 358 if (error < 0) 359 goto fail; 360 361 /* Dump trailer */ 362 error = dump_write(di, &kdh, 0, dumplo, sizeof(kdh)); 363 if (error) 364 goto fail; 365 366 /* Signal completion, signoff and exit stage left. */ 367 dump_write(di, NULL, 0, 0, 0); 368 printf("\nDump complete\n"); 369 return; 370 371 fail: 372 if (error < 0) 373 error = -error; 374 375 if (error == ECANCELED) 376 printf("\nDump aborted\n"); 377 else if (error == ENOSPC) 378 printf("\nDump failed. Partition too small.\n"); 379 else 380 printf("\n** DUMP FAILED (ERROR %d) **\n", error); 381 } 382