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