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 #include <sys/watchdog.h> 40 #include <vm/vm.h> 41 #include <vm/pmap.h> 42 #include <machine/elf.h> 43 #include <machine/md_var.h> 44 45 #ifdef __amd64__ 46 #define KERNELDUMP_VERSION KERNELDUMP_AMD64_VERSION 47 #define EM_VALUE EM_X86_64 48 #else 49 #define KERNELDUMP_VERSION KERNELDUMP_I386_VERSION 50 #define EM_VALUE EM_386 51 #endif 52 53 CTASSERT(sizeof(struct kerneldumpheader) == 512); 54 55 int do_minidump = 1; 56 TUNABLE_INT("debug.minidump", &do_minidump); 57 SYSCTL_INT(_debug, OID_AUTO, minidump, CTLFLAG_RW, &do_minidump, 0, 58 "Enable mini crash dumps"); 59 60 /* 61 * Don't touch the first SIZEOF_METADATA bytes on the dump device. This 62 * is to protect us from metadata and to protect metadata from us. 63 */ 64 #define SIZEOF_METADATA (64*1024) 65 66 #define MD_ALIGN(x) (((off_t)(x) + PAGE_MASK) & ~PAGE_MASK) 67 #define DEV_ALIGN(x) (((off_t)(x) + (DEV_BSIZE-1)) & ~(DEV_BSIZE-1)) 68 69 struct md_pa { 70 vm_paddr_t md_start; 71 vm_paddr_t md_size; 72 }; 73 74 typedef int callback_t(struct md_pa *, int, void *); 75 76 static struct kerneldumpheader kdh; 77 static off_t dumplo, fileofs; 78 79 /* Handle buffered writes. */ 80 static char buffer[DEV_BSIZE]; 81 static size_t fragsz; 82 83 /* 20 phys_avail entry pairs correspond to 10 md_pa's */ 84 static struct md_pa dump_map[10]; 85 86 static void 87 md_pa_init(void) 88 { 89 int n, idx; 90 91 bzero(dump_map, sizeof(dump_map)); 92 for (n = 0; n < sizeof(dump_map) / sizeof(dump_map[0]); n++) { 93 idx = n * 2; 94 if (dump_avail[idx] == 0 && dump_avail[idx + 1] == 0) 95 break; 96 dump_map[n].md_start = dump_avail[idx]; 97 dump_map[n].md_size = dump_avail[idx + 1] - dump_avail[idx]; 98 } 99 } 100 101 static struct md_pa * 102 md_pa_first(void) 103 { 104 105 return (&dump_map[0]); 106 } 107 108 static struct md_pa * 109 md_pa_next(struct md_pa *mdp) 110 { 111 112 mdp++; 113 if (mdp->md_size == 0) 114 mdp = NULL; 115 return (mdp); 116 } 117 118 static int 119 buf_write(struct dumperinfo *di, char *ptr, size_t sz) 120 { 121 size_t len; 122 int error; 123 124 while (sz) { 125 len = DEV_BSIZE - fragsz; 126 if (len > sz) 127 len = sz; 128 bcopy(ptr, buffer + fragsz, len); 129 fragsz += len; 130 ptr += len; 131 sz -= len; 132 if (fragsz == DEV_BSIZE) { 133 error = dump_write(di, buffer, 0, dumplo, 134 DEV_BSIZE); 135 if (error) 136 return error; 137 dumplo += DEV_BSIZE; 138 fragsz = 0; 139 } 140 } 141 142 return (0); 143 } 144 145 static int 146 buf_flush(struct dumperinfo *di) 147 { 148 int error; 149 150 if (fragsz == 0) 151 return (0); 152 153 error = dump_write(di, buffer, 0, dumplo, DEV_BSIZE); 154 dumplo += DEV_BSIZE; 155 fragsz = 0; 156 return (error); 157 } 158 159 #define PG2MB(pgs) ((pgs + (1 << 8) - 1) >> 8) 160 161 static int 162 cb_dumpdata(struct md_pa *mdp, int seqnr, void *arg) 163 { 164 struct dumperinfo *di = (struct dumperinfo*)arg; 165 vm_paddr_t a, pa; 166 void *va; 167 uint64_t pgs; 168 size_t counter, sz, chunk; 169 int i, c, error, twiddle; 170 u_int maxdumppgs; 171 172 error = 0; /* catch case in which chunk size is 0 */ 173 counter = 0; /* Update twiddle every 16MB */ 174 twiddle = 0; 175 va = 0; 176 pgs = mdp->md_size / PAGE_SIZE; 177 pa = mdp->md_start; 178 maxdumppgs = min(di->maxiosize / PAGE_SIZE, MAXDUMPPGS); 179 if (maxdumppgs == 0) /* seatbelt */ 180 maxdumppgs = 1; 181 182 printf(" chunk %d: %juMB (%ju pages)", seqnr, (uintmax_t)PG2MB(pgs), 183 (uintmax_t)pgs); 184 185 while (pgs) { 186 chunk = pgs; 187 if (chunk > maxdumppgs) 188 chunk = maxdumppgs; 189 sz = chunk << PAGE_SHIFT; 190 counter += sz; 191 if (counter >> 24) { 192 printf(" %ju", (uintmax_t)PG2MB(pgs)); 193 counter &= (1<<24) - 1; 194 } 195 for (i = 0; i < chunk; i++) { 196 a = pa + i * PAGE_SIZE; 197 va = pmap_kenter_temporary(trunc_page(a), i); 198 } 199 200 wdog_kern_pat(WD_LASTVAL); 201 202 error = dump_write(di, va, 0, dumplo, sz); 203 if (error) 204 break; 205 dumplo += sz; 206 pgs -= chunk; 207 pa += sz; 208 209 /* Check for user abort. */ 210 c = cncheckc(); 211 if (c == 0x03) 212 return (ECANCELED); 213 if (c != -1) 214 printf(" (CTRL-C to abort) "); 215 } 216 printf(" ... %s\n", (error) ? "fail" : "ok"); 217 return (error); 218 } 219 220 static int 221 cb_dumphdr(struct md_pa *mdp, int seqnr, void *arg) 222 { 223 struct dumperinfo *di = (struct dumperinfo*)arg; 224 Elf_Phdr phdr; 225 uint64_t size; 226 int error; 227 228 size = mdp->md_size; 229 bzero(&phdr, sizeof(phdr)); 230 phdr.p_type = PT_LOAD; 231 phdr.p_flags = PF_R; /* XXX */ 232 phdr.p_offset = fileofs; 233 phdr.p_vaddr = mdp->md_start; 234 phdr.p_paddr = mdp->md_start; 235 phdr.p_filesz = size; 236 phdr.p_memsz = size; 237 phdr.p_align = PAGE_SIZE; 238 239 error = buf_write(di, (char*)&phdr, sizeof(phdr)); 240 fileofs += phdr.p_filesz; 241 return (error); 242 } 243 244 static int 245 cb_size(struct md_pa *mdp, int seqnr, void *arg) 246 { 247 uint64_t *sz = (uint64_t*)arg; 248 249 *sz += (uint64_t)mdp->md_size; 250 return (0); 251 } 252 253 static int 254 foreach_chunk(callback_t cb, void *arg) 255 { 256 struct md_pa *mdp; 257 int error, seqnr; 258 259 seqnr = 0; 260 mdp = md_pa_first(); 261 while (mdp != NULL) { 262 error = (*cb)(mdp, seqnr++, arg); 263 if (error) 264 return (-error); 265 mdp = md_pa_next(mdp); 266 } 267 return (seqnr); 268 } 269 270 void 271 dumpsys(struct dumperinfo *di) 272 { 273 Elf_Ehdr ehdr; 274 uint64_t dumpsize; 275 off_t hdrgap; 276 size_t hdrsz; 277 int error; 278 279 if (do_minidump) { 280 minidumpsys(di); 281 return; 282 } 283 bzero(&ehdr, sizeof(ehdr)); 284 ehdr.e_ident[EI_MAG0] = ELFMAG0; 285 ehdr.e_ident[EI_MAG1] = ELFMAG1; 286 ehdr.e_ident[EI_MAG2] = ELFMAG2; 287 ehdr.e_ident[EI_MAG3] = ELFMAG3; 288 ehdr.e_ident[EI_CLASS] = ELF_CLASS; 289 #if BYTE_ORDER == LITTLE_ENDIAN 290 ehdr.e_ident[EI_DATA] = ELFDATA2LSB; 291 #else 292 ehdr.e_ident[EI_DATA] = ELFDATA2MSB; 293 #endif 294 ehdr.e_ident[EI_VERSION] = EV_CURRENT; 295 ehdr.e_ident[EI_OSABI] = ELFOSABI_STANDALONE; /* XXX big picture? */ 296 ehdr.e_type = ET_CORE; 297 ehdr.e_machine = EM_VALUE; 298 ehdr.e_phoff = sizeof(ehdr); 299 ehdr.e_flags = 0; 300 ehdr.e_ehsize = sizeof(ehdr); 301 ehdr.e_phentsize = sizeof(Elf_Phdr); 302 ehdr.e_shentsize = sizeof(Elf_Shdr); 303 304 md_pa_init(); 305 306 /* Calculate dump size. */ 307 dumpsize = 0L; 308 ehdr.e_phnum = foreach_chunk(cb_size, &dumpsize); 309 hdrsz = ehdr.e_phoff + ehdr.e_phnum * ehdr.e_phentsize; 310 fileofs = MD_ALIGN(hdrsz); 311 dumpsize += fileofs; 312 hdrgap = fileofs - DEV_ALIGN(hdrsz); 313 314 /* Determine dump offset on device. */ 315 if (di->mediasize < SIZEOF_METADATA + dumpsize + sizeof(kdh) * 2) { 316 error = ENOSPC; 317 goto fail; 318 } 319 dumplo = di->mediaoffset + di->mediasize - dumpsize; 320 dumplo -= sizeof(kdh) * 2; 321 322 mkdumpheader(&kdh, KERNELDUMPMAGIC, KERNELDUMP_VERSION, dumpsize, 323 di->blocksize); 324 325 printf("Dumping %llu MB (%d chunks)\n", (long long)dumpsize >> 20, 326 ehdr.e_phnum); 327 328 /* Dump leader */ 329 error = dump_write(di, &kdh, 0, dumplo, sizeof(kdh)); 330 if (error) 331 goto fail; 332 dumplo += sizeof(kdh); 333 334 /* Dump ELF header */ 335 error = buf_write(di, (char*)&ehdr, sizeof(ehdr)); 336 if (error) 337 goto fail; 338 339 /* Dump program headers */ 340 error = foreach_chunk(cb_dumphdr, di); 341 if (error < 0) 342 goto fail; 343 buf_flush(di); 344 345 /* 346 * All headers are written using blocked I/O, so we know the 347 * current offset is (still) block aligned. Skip the alignement 348 * in the file to have the segment contents aligned at page 349 * boundary. We cannot use MD_ALIGN on dumplo, because we don't 350 * care and may very well be unaligned within the dump device. 351 */ 352 dumplo += hdrgap; 353 354 /* Dump memory chunks (updates dumplo) */ 355 error = foreach_chunk(cb_dumpdata, di); 356 if (error < 0) 357 goto fail; 358 359 /* Dump trailer */ 360 error = dump_write(di, &kdh, 0, dumplo, sizeof(kdh)); 361 if (error) 362 goto fail; 363 364 /* Signal completion, signoff and exit stage left. */ 365 dump_write(di, NULL, 0, 0, 0); 366 printf("\nDump complete\n"); 367 return; 368 369 fail: 370 if (error < 0) 371 error = -error; 372 373 if (error == ECANCELED) 374 printf("\nDump aborted\n"); 375 else if (error == ENOSPC) 376 printf("\nDump failed. Partition too small.\n"); 377 else 378 printf("\n** DUMP FAILED (ERROR %d) **\n", error); 379 } 380