sys/platform/pc64/x86_64/dump_machdep.c

   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  * $FreeBSD: src/sys/amd64/amd64/dump_machdep.c,v 1.18 2009/05/29 21:27:12 jamie Exp $
  27  */
  28
  29 #include <sys/param.h>
  30 #include <sys/systm.h>
  31 #include <sys/conf.h>
  32 #include <sys/cons.h>
  33 #include <sys/sysctl.h>
  34 #include <sys/device.h>
  35 #include <sys/kernel.h>
  36 #include <sys/kerneldump.h>
  37 #include <sys/kbio.h>
  38 #include <vm/vm.h>
  39 #include <vm/pmap.h>
  40 #include <machine/elf.h>
  41 #include <machine/md_var.h>
  42 #include <machine/thread.h>
  43 #include <machine/vmparam.h>
  44
  45 CTASSERT(sizeof(struct kerneldumpheader) == 512);
  46
  47 int do_minidump = 1;
  48 TUNABLE_INT("debug.minidump", &do_minidump);
  49 SYSCTL_INT(_debug, OID_AUTO, minidump, CTLFLAG_RW, &do_minidump, 0,
  50     "Enable mini crash dumps");
  51
  52 /*
  53  * Don't touch the first SIZEOF_METADATA bytes on the dump device. This
  54  * is to protect us from metadata and to protect metadata from us.
  55  */
  56 #define SIZEOF_METADATA         (64*1024)
  57
  58 #define MD_ALIGN(x)     (((off_t)(x) + PAGE_MASK) & ~PAGE_MASK)
  59 #define DEV_ALIGN(x)    roundup2((off_t)(x), DEV_BSIZE)
  60
  61 struct md_pa {
  62         vm_paddr_t md_start;
  63         vm_paddr_t md_size;
  64 };
  65
  66 typedef int callback_t(struct md_pa *, int, void *);
  67
  68 static struct kerneldumpheader kdh;
  69 static off_t dumplo, fileofs;
  70
  71 /* Handle buffered writes. */
  72 static char buffer[DEV_BSIZE];
  73 static size_t fragsz;
  74
  75 static struct md_pa dump_map[VM_PHYSSEG_MAX+1];
  76
  77 static void
  78 md_pa_init(void)
  79 {
  80         int n;
  81
  82         bzero(dump_map, sizeof(dump_map));
  83         for (n = 0; n < NELEM(dump_map); n++) {
  84                 if (dump_avail[n].phys_beg == 0 && dump_avail[n].phys_end == 0)
  85                         break;
  86                 dump_map[n].md_start = dump_avail[n].phys_beg;
  87                 dump_map[n].md_size = dump_avail[n].phys_end -
  88                                       dump_avail[n].phys_beg;
  89         }
  90 }
  91
  92 static struct md_pa *
  93 md_pa_first(void)
  94 {
  95
  96         return (&dump_map[0]);
  97 }
  98
  99 static struct md_pa *
 100 md_pa_next(struct md_pa *mdp)
 101 {
 102
 103         mdp++;
 104         if (mdp->md_size == 0)
 105                 mdp = NULL;
 106         return (mdp);
 107 }
 108
 109 static int
 110 buf_write(struct dumperinfo *di, char *ptr, size_t sz)
 111 {
 112         size_t len;
 113         int error;
 114
 115         while (sz) {
 116                 len = DEV_BSIZE - fragsz;
 117                 if (len > sz)
 118                         len = sz;
 119                 bcopy(ptr, buffer + fragsz, len);
 120                 fragsz += len;
 121                 ptr += len;
 122                 sz -= len;
 123                 if (fragsz == DEV_BSIZE) {
 124                         error = dev_ddump(di->priv, buffer, 0, dumplo,
 125                             DEV_BSIZE);
 126                         if (error)
 127                                 return error;
 128                         dumplo += DEV_BSIZE;
 129                         fragsz = 0;
 130                 }
 131         }
 132
 133         return (0);
 134 }
 135
 136 static int
 137 buf_flush(struct dumperinfo *di)
 138 {
 139         int error;
 140
 141         if (fragsz == 0)
 142                 return (0);
 143
 144         error = dev_ddump(di->priv, buffer, 0, dumplo, DEV_BSIZE);
 145         dumplo += DEV_BSIZE;
 146         fragsz = 0;
 147         return (error);
 148 }
 149
 150 #define PG2MB(pgs) ((pgs + (1 << 8) - 1) >> 8)
 151
 152 static int
 153 cb_dumpdata(struct md_pa *mdp, int seqnr, void *arg)
 154 {
 155         struct dumperinfo *di = (struct dumperinfo*)arg;
 156         vm_paddr_t a, pa;
 157         void *va;
 158         uint64_t pgs;
 159         size_t counter, sz, chunk;
 160         int i, c, error;
 161         int max_iosize;
 162
 163         error = 0;      /* catch case in which chunk size is 0 */
 164         counter = 0;    /* Update twiddle every 16MB */
 165         va = NULL;
 166         pgs = mdp->md_size / PAGE_SIZE;
 167         pa = mdp->md_start;
 168         max_iosize = min(MAXPHYS, di->maxiosize);
 169
 170         kprintf("  chunk %d: %ldMB (%ld pages)", seqnr, PG2MB(pgs), pgs);
 171
 172         cnpoll(TRUE);
 173         while (pgs) {
 174                 chunk = pgs;
 175                 if (chunk > (max_iosize/PAGE_SIZE))
 176                         chunk = max_iosize/PAGE_SIZE;
 177                 sz = chunk << PAGE_SHIFT;
 178                 counter += sz;
 179                 if (counter >> 24) {
 180                         kprintf(" %ld", PG2MB(pgs));
 181                         counter &= (1<<24) - 1;
 182                 }
 183                 for (i = 0; i < chunk; i++) {
 184                         a = pa + i * PAGE_SIZE;
 185                         va = pmap_kenter_temporary(trunc_page(a), i);
 186                 }
 187                 smp_invltlb();
 188                 error = dev_ddump(di->priv, va, 0, dumplo, sz);
 189                 if (error)
 190                         break;
 191                 dumplo += sz;
 192                 pgs -= chunk;
 193                 pa += sz;
 194
 195                 /* Check for user abort. */
 196                 c = cncheckc();
 197                 if (c == 0x03) {
 198                         error = ECANCELED;
 199                         goto done;
 200                 }
 201                 if (c != -1 && c != NOKEY)
 202                         kprintf(" (CTRL-C to abort) ");
 203         }
 204         kprintf(" ... %s\n", (error) ? "fail" : "ok");
 205 done:
 206         cnpoll(FALSE);
 207         return (error);
 208 }
 209
 210 static int
 211 cb_dumphdr(struct md_pa *mdp, int seqnr, void *arg)
 212 {
 213         struct dumperinfo *di = (struct dumperinfo*)arg;
 214         Elf_Phdr phdr;
 215         uint64_t size;
 216         int error;
 217
 218         size = mdp->md_size;
 219         bzero(&phdr, sizeof(phdr));
 220         phdr.p_type = PT_LOAD;
 221         phdr.p_flags = PF_R;                    /* XXX */
 222         phdr.p_offset = fileofs;
 223         phdr.p_vaddr = mdp->md_start;
 224         phdr.p_paddr = mdp->md_start;
 225         phdr.p_filesz = size;
 226         phdr.p_memsz = size;
 227         phdr.p_align = PAGE_SIZE;
 228
 229         error = buf_write(di, (char*)&phdr, sizeof(phdr));
 230         fileofs += phdr.p_filesz;
 231         return (error);
 232 }
 233
 234 static int
 235 cb_size(struct md_pa *mdp, int seqnr, void *arg)
 236 {
 237         uint64_t *sz = (uint64_t*)arg;
 238
 239         *sz += (uint64_t)mdp->md_size;
 240         return (0);
 241 }
 242
 243 static int
 244 foreach_chunk(callback_t cb, void *arg)
 245 {
 246         struct md_pa *mdp;
 247         int error, seqnr;
 248
 249         seqnr = 0;
 250         mdp = md_pa_first();
 251         while (mdp != NULL) {
 252                 error = (*cb)(mdp, seqnr++, arg);
 253                 if (error)
 254                         return (-error);
 255                 mdp = md_pa_next(mdp);
 256         }
 257         return (seqnr);
 258 }
 259
 260 void
 261 md_dumpsys(struct dumperinfo *di)
 262 {
 263         Elf_Ehdr ehdr;
 264         uint64_t dumpsize;
 265         off_t hdrgap;
 266         size_t hdrsz;
 267         int error;
 268
 269         /*
 270          * Save context if dump called without panic.
 271          */
 272         if (dumpthread == NULL) {
 273                 savectx(&dumppcb);
 274                 dumpthread = curthread;
 275         }
 276
 277         if (do_minidump) {
 278                 minidumpsys(di);
 279                 return;
 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_X86_64;
 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_AMD64_VERSION,
 321             dumpsize, di->blocksize);
 322
 323         kprintf("Dumping %llu MB (%d chunks)\n", (long long)dumpsize >> 20,
 324             ehdr.e_phnum);
 325
 326         /* Dump leader */
 327         error = dev_ddump(di->priv, &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 = dev_ddump(di->priv, &kdh, 0, dumplo, sizeof(kdh));
 359         if (error)
 360                 goto fail;
 361
 362         /* Signal completion, signoff and exit stage left. */
 363         dev_ddump(di->priv, NULL, 0, 0, 0);
 364         kprintf("\nDump complete\n");
 365         return;
 366
 367  fail:
 368         if (error < 0)
 369                 error = -error;
 370
 371         if (error == ECANCELED)
 372                 kprintf("\nDump aborted\n");
 373         else if (error == ENOSPC)
 374                 kprintf("\nDump failed. Partition too small.\n");
 375         else
 376                 kprintf("\n** DUMP FAILED (ERROR %d) **\n", error);
 377 }