1 /*- 2 * Copyright (c) 2006 Peter Wemm 3 * Copyright (c) 2008 Semihalf, Grzegorz Bernacki 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 19 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 20 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 21 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 22 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 23 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 24 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 25 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 * 27 * from: FreeBSD: src/sys/i386/i386/minidump_machdep.c,v 1.6 2008/08/17 23:27:27 28 */ 29 30 #include <sys/cdefs.h> 31 __FBSDID("$FreeBSD$"); 32 33 #include "opt_watchdog.h" 34 35 #include <sys/param.h> 36 #include <sys/systm.h> 37 #include <sys/conf.h> 38 #include <sys/cons.h> 39 #include <sys/kernel.h> 40 #include <sys/kerneldump.h> 41 #include <sys/msgbuf.h> 42 #ifdef SW_WATCHDOG 43 #include <sys/watchdog.h> 44 #endif 45 #include <vm/vm.h> 46 #include <vm/pmap.h> 47 #include <machine/atomic.h> 48 #include <machine/cpu.h> 49 #include <machine/elf.h> 50 #include <machine/md_var.h> 51 #include <machine/minidump.h> 52 #include <machine/vmparam.h> 53 54 CTASSERT(sizeof(struct kerneldumpheader) == 512); 55 56 /* 57 * Don't touch the first SIZEOF_METADATA bytes on the dump device. This 58 * is to protect us from metadata and to protect metadata from us. 59 */ 60 #define SIZEOF_METADATA (64*1024) 61 62 uint32_t *vm_page_dump; 63 int vm_page_dump_size; 64 65 static struct kerneldumpheader kdh; 66 67 static off_t dumplo; 68 69 /* Handle chunked writes. */ 70 static size_t fragsz; 71 static void *dump_va; 72 static uint64_t counter, progress; 73 74 CTASSERT(sizeof(*vm_page_dump) == 4); 75 76 static int 77 is_dumpable(vm_paddr_t pa) 78 { 79 int i; 80 81 for (i = 0; dump_avail[i] != 0 || dump_avail[i + 1] != 0; i += 2) { 82 if (pa >= dump_avail[i] && pa < dump_avail[i + 1]) 83 return (1); 84 } 85 return (0); 86 } 87 88 #define PG2MB(pgs) (((pgs) + (1 << 8) - 1) >> 8) 89 90 static int 91 blk_flush(struct dumperinfo *di) 92 { 93 int error; 94 95 if (fragsz == 0) 96 return (0); 97 98 error = dump_write(di, dump_va, 0, dumplo, fragsz); 99 dumplo += fragsz; 100 fragsz = 0; 101 return (error); 102 } 103 104 static int 105 blk_write(struct dumperinfo *di, char *ptr, vm_paddr_t pa, size_t sz) 106 { 107 size_t len; 108 int error, i, c; 109 u_int maxdumpsz; 110 111 maxdumpsz = min(di->maxiosize, MAXDUMPPGS * PAGE_SIZE); 112 if (maxdumpsz == 0) /* seatbelt */ 113 maxdumpsz = PAGE_SIZE; 114 error = 0; 115 if (ptr != NULL && pa != 0) { 116 printf("cant have both va and pa!\n"); 117 return (EINVAL); 118 } 119 if (pa != 0) { 120 if ((sz % PAGE_SIZE) != 0) { 121 printf("size not page aligned\n"); 122 return (EINVAL); 123 } 124 if ((pa & PAGE_MASK) != 0) { 125 printf("address not page aligned\n"); 126 return (EINVAL); 127 } 128 } 129 if (ptr != NULL) { 130 /* Flush any pre-existing pa pages before a virtual dump. */ 131 error = blk_flush(di); 132 if (error) 133 return (error); 134 } 135 while (sz) { 136 len = maxdumpsz - fragsz; 137 if (len > sz) 138 len = sz; 139 counter += len; 140 progress -= len; 141 if (counter >> 22) { 142 printf(" %lld", PG2MB(progress >> PAGE_SHIFT)); 143 counter &= (1<<22) - 1; 144 } 145 146 #ifdef SW_WATCHDOG 147 wdog_kern_pat(WD_LASTVAL); 148 #endif 149 if (ptr) { 150 error = dump_write(di, ptr, 0, dumplo, len); 151 if (error) 152 return (error); 153 dumplo += len; 154 ptr += len; 155 sz -= len; 156 } else { 157 for (i = 0; i < len; i += PAGE_SIZE) 158 dump_va = pmap_kenter_temporary(pa + i, 159 (i + fragsz) >> PAGE_SHIFT); 160 fragsz += len; 161 pa += len; 162 sz -= len; 163 if (fragsz == maxdumpsz) { 164 error = blk_flush(di); 165 if (error) 166 return (error); 167 } 168 } 169 170 /* Check for user abort. */ 171 c = cncheckc(); 172 if (c == 0x03) 173 return (ECANCELED); 174 if (c != -1) 175 printf(" (CTRL-C to abort) "); 176 } 177 178 return (0); 179 } 180 181 /* A buffer for general use. Its size must be one page at least. */ 182 static char dumpbuf[PAGE_SIZE]; 183 CTASSERT(sizeof(dumpbuf) % sizeof(pt2_entry_t) == 0); 184 185 int 186 minidumpsys(struct dumperinfo *di) 187 { 188 struct minidumphdr mdhdr; 189 uint64_t dumpsize; 190 uint32_t ptesize; 191 uint32_t bits; 192 uint32_t pa, prev_pa = 0, count = 0; 193 vm_offset_t va; 194 int i, bit, error; 195 char *addr; 196 197 /* 198 * Flush caches. Note that in the SMP case this operates only on the 199 * current CPU's L1 cache. Before we reach this point, code in either 200 * the system shutdown or kernel debugger has called stop_cpus() to stop 201 * all cores other than this one. Part of the ARM handling of 202 * stop_cpus() is to call wbinv_all() on that core's local L1 cache. So 203 * by time we get to here, all that remains is to flush the L1 for the 204 * current CPU, then the L2. 205 */ 206 dcache_wbinv_poc_all(); 207 208 counter = 0; 209 /* Walk page table pages, set bits in vm_page_dump */ 210 ptesize = 0; 211 for (va = KERNBASE; va < kernel_vm_end; va += PAGE_SIZE) { 212 pa = pmap_dump_kextract(va, NULL); 213 if (pa != 0 && is_dumpable(pa)) 214 dump_add_page(pa); 215 ptesize += sizeof(pt2_entry_t); 216 } 217 218 /* Calculate dump size. */ 219 dumpsize = ptesize; 220 dumpsize += round_page(msgbufp->msg_size); 221 dumpsize += round_page(vm_page_dump_size); 222 223 for (i = 0; i < vm_page_dump_size / sizeof(*vm_page_dump); i++) { 224 bits = vm_page_dump[i]; 225 while (bits) { 226 bit = ffs(bits) - 1; 227 pa = (((uint64_t)i * sizeof(*vm_page_dump) * NBBY) + 228 bit) * PAGE_SIZE; 229 /* Clear out undumpable pages now if needed */ 230 if (is_dumpable(pa)) 231 dumpsize += PAGE_SIZE; 232 else 233 dump_drop_page(pa); 234 bits &= ~(1ul << bit); 235 } 236 } 237 238 dumpsize += PAGE_SIZE; 239 240 /* Determine dump offset on device. */ 241 if (di->mediasize < SIZEOF_METADATA + dumpsize + di->blocksize * 2 + 242 kerneldumpcrypto_dumpkeysize(di->kdc)) { 243 error = ENOSPC; 244 goto fail; 245 } 246 247 dumplo = di->mediaoffset + di->mediasize - dumpsize; 248 dumplo -= di->blocksize * 2; 249 dumplo -= kerneldumpcrypto_dumpkeysize(di->kdc); 250 progress = dumpsize; 251 252 /* Initialize kernel dump crypto. */ 253 error = kerneldumpcrypto_init(di->kdc); 254 if (error) 255 goto fail; 256 257 /* Initialize mdhdr */ 258 bzero(&mdhdr, sizeof(mdhdr)); 259 strcpy(mdhdr.magic, MINIDUMP_MAGIC); 260 mdhdr.version = MINIDUMP_VERSION; 261 mdhdr.msgbufsize = msgbufp->msg_size; 262 mdhdr.bitmapsize = vm_page_dump_size; 263 mdhdr.ptesize = ptesize; 264 mdhdr.kernbase = KERNBASE; 265 mdhdr.arch = __ARM_ARCH; 266 #if __ARM_ARCH >= 6 267 mdhdr.mmuformat = MINIDUMP_MMU_FORMAT_V6; 268 #else 269 mdhdr.mmuformat = MINIDUMP_MMU_FORMAT_V4; 270 #endif 271 mkdumpheader(&kdh, KERNELDUMPMAGIC, KERNELDUMP_ARM_VERSION, dumpsize, 272 kerneldumpcrypto_dumpkeysize(di->kdc), di->blocksize); 273 274 printf("Physical memory: %u MB\n", ptoa((uintmax_t)physmem) / 1048576); 275 printf("Dumping %llu MB:", (long long)dumpsize >> 20); 276 277 /* Dump leader */ 278 error = dump_write_header(di, &kdh, 0, dumplo); 279 if (error) 280 goto fail; 281 dumplo += di->blocksize; 282 283 /* Dump key */ 284 error = dump_write_key(di, 0, dumplo); 285 if (error) 286 goto fail; 287 dumplo += kerneldumpcrypto_dumpkeysize(di->kdc); 288 289 /* Dump my header */ 290 bzero(dumpbuf, sizeof(dumpbuf)); 291 bcopy(&mdhdr, dumpbuf, sizeof(mdhdr)); 292 error = blk_write(di, dumpbuf, 0, PAGE_SIZE); 293 if (error) 294 goto fail; 295 296 /* Dump msgbuf up front */ 297 error = blk_write(di, (char *)msgbufp->msg_ptr, 0, 298 round_page(msgbufp->msg_size)); 299 if (error) 300 goto fail; 301 302 /* Dump bitmap */ 303 error = blk_write(di, (char *)vm_page_dump, 0, 304 round_page(vm_page_dump_size)); 305 if (error) 306 goto fail; 307 308 /* Dump kernel page table pages */ 309 addr = dumpbuf; 310 for (va = KERNBASE; va < kernel_vm_end; va += PAGE_SIZE) { 311 pmap_dump_kextract(va, (pt2_entry_t *)addr); 312 addr += sizeof(pt2_entry_t); 313 if (addr == dumpbuf + sizeof(dumpbuf)) { 314 error = blk_write(di, dumpbuf, 0, sizeof(dumpbuf)); 315 if (error != 0) 316 goto fail; 317 addr = dumpbuf; 318 } 319 } 320 if (addr != dumpbuf) { 321 error = blk_write(di, dumpbuf, 0, addr - dumpbuf); 322 if (error != 0) 323 goto fail; 324 } 325 326 /* Dump memory chunks */ 327 for (i = 0; i < vm_page_dump_size / sizeof(*vm_page_dump); i++) { 328 bits = vm_page_dump[i]; 329 while (bits) { 330 bit = ffs(bits) - 1; 331 pa = (((uint64_t)i * sizeof(*vm_page_dump) * NBBY) + 332 bit) * PAGE_SIZE; 333 if (!count) { 334 prev_pa = pa; 335 count++; 336 } else { 337 if (pa == (prev_pa + count * PAGE_SIZE)) 338 count++; 339 else { 340 error = blk_write(di, NULL, prev_pa, 341 count * PAGE_SIZE); 342 if (error) 343 goto fail; 344 count = 1; 345 prev_pa = pa; 346 } 347 } 348 bits &= ~(1ul << bit); 349 } 350 } 351 if (count) { 352 error = blk_write(di, NULL, prev_pa, count * PAGE_SIZE); 353 if (error) 354 goto fail; 355 count = 0; 356 prev_pa = 0; 357 } 358 359 error = blk_flush(di); 360 if (error) 361 goto fail; 362 363 /* Dump trailer */ 364 error = dump_write_header(di, &kdh, 0, dumplo); 365 if (error) 366 goto fail; 367 dumplo += di->blocksize; 368 369 /* Signal completion, signoff and exit stage left. */ 370 dump_write(di, NULL, 0, 0, 0); 371 printf("\nDump complete\n"); 372 return (0); 373 374 fail: 375 if (error < 0) 376 error = -error; 377 378 if (error == ECANCELED) 379 printf("\nDump aborted\n"); 380 else if (error == ENOSPC) 381 printf("\nDump failed. Partition too small.\n"); 382 else 383 printf("\n** DUMP FAILED (ERROR %d) **\n", error); 384 return (error); 385 } 386 387 void 388 dump_add_page(vm_paddr_t pa) 389 { 390 int idx, bit; 391 392 pa >>= PAGE_SHIFT; 393 idx = pa >> 5; /* 2^5 = 32 */ 394 bit = pa & 31; 395 atomic_set_int(&vm_page_dump[idx], 1ul << bit); 396 } 397 398 void 399 dump_drop_page(vm_paddr_t pa) 400 { 401 int idx, bit; 402 403 pa >>= PAGE_SHIFT; 404 idx = pa >> 5; /* 2^5 = 32 */ 405 bit = pa & 31; 406 atomic_clear_int(&vm_page_dump[idx], 1ul << bit); 407 } 408