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 <sys/thread2.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 /* 20 phys_avail entry pairs correspond to 10 md_pa's */
  76 static struct md_pa dump_map[10];
  77
  78 static void
  79 md_pa_init(void)
  80 {
  81         int n, idx;
  82
  83         bzero(dump_map, sizeof(dump_map));
  84         for (n = 0; n < NELEM(dump_map); n++) {
  85                 idx = n * 2;
  86                 if (dump_avail[idx] == 0 && dump_avail[idx + 1] == 0)
  87                         break;
  88                 dump_map[n].md_start = dump_avail[idx];
  89                 dump_map[n].md_size = dump_avail[idx + 1] - dump_avail[idx];
  90         }
  91 }
  92
  93 static struct md_pa *
  94 md_pa_first(void)
  95 {
  96
  97         return (&dump_map[0]);
  98 }
  99
 100 static struct md_pa *
 101 md_pa_next(struct md_pa *mdp)
 102 {
 103
 104         mdp++;
 105         if (mdp->md_size == 0)
 106                 mdp = NULL;
 107         return (mdp);
 108 }
 109
 110 static int
 111 buf_write(struct dumperinfo *di, char *ptr, size_t sz)
 112 {
 113         size_t len;
 114         int error;
 115
 116         while (sz) {
 117                 len = DEV_BSIZE - fragsz;
 118                 if (len > sz)
 119                         len = sz;
 120                 bcopy(ptr, buffer + fragsz, len);
 121                 fragsz += len;
 122                 ptr += len;
 123                 sz -= len;
 124                 if (fragsz == DEV_BSIZE) {
 125                         error = dev_ddump(di->priv, buffer, 0, dumplo,
 126                             DEV_BSIZE);
 127                         if (error)
 128                                 return error;
 129                         dumplo += DEV_BSIZE;
 130                         fragsz = 0;
 131                 }
 132         }
 133
 134         return (0);
 135 }
 136
 137 static int
 138 buf_flush(struct dumperinfo *di)
 139 {
 140         int error;
 141
 142         if (fragsz == 0)
 143                 return (0);
 144
 145         error = dev_ddump(di->priv, buffer, 0, dumplo, DEV_BSIZE);
 146         dumplo += DEV_BSIZE;
 147         fragsz = 0;
 148         return (error);
 149 }
 150
 151 #define PG2MB(pgs) ((pgs + (1 << 8) - 1) >> 8)
 152
 153 static int
 154 cb_dumpdata(struct md_pa *mdp, int seqnr, void *arg)
 155 {
 156         struct dumperinfo *di = (struct dumperinfo*)arg;
 157         vm_paddr_t a, pa;
 158         void *va;
 159         uint64_t pgs;
 160         size_t counter, sz, chunk;
 161         int i, c, error;
 162         int max_iosize;
 163
 164         error = 0;      /* catch case in which chunk size is 0 */
 165         counter = 0;    /* Update twiddle every 16MB */
 166         va = NULL;
 167         pgs = mdp->md_size / PAGE_SIZE;
 168         pa = mdp->md_start;
 169         max_iosize = min(MAXPHYS, di->maxiosize);
 170
 171         kprintf("  chunk %d: %ldMB (%ld pages)", seqnr, PG2MB(pgs), pgs);
 172
 173         cnpoll(TRUE);
 174         while (pgs) {
 175                 chunk = pgs;
 176                 if (chunk > (max_iosize/PAGE_SIZE))
 177                         chunk = max_iosize/PAGE_SIZE;
 178                 sz = chunk << PAGE_SHIFT;
 179                 counter += sz;
 180                 if (counter >> 24) {
 181                         kprintf(" %ld", PG2MB(pgs));
 182                         counter &= (1<<24) - 1;
 183                 }
 184                 for (i = 0; i < chunk; i++) {
 185                         a = pa + i * PAGE_SIZE;
 186                         va = pmap_kenter_temporary(trunc_page(a), i);
 187                 }
 188                 smp_invltlb();
 189                 error = dev_ddump(di->priv, va, 0, dumplo, sz);
 190                 if (error)
 191                         break;
 192                 dumplo += sz;
 193                 pgs -= chunk;
 194                 pa += sz;
 195
 196                 /* Check for user abort. */
 197                 c = cncheckc();
 198                 if (c == 0x03) {
 199                         error = ECANCELED;
 200                         goto done;
 201                 }
 202                 if (c != -1 && c != NOKEY)
 203                         kprintf(" (CTRL-C to abort) ");
 204         }
 205         kprintf(" ... %s\n", (error) ? "fail" : "ok");
 206 done:
 207         cnpoll(FALSE);
 208         return (error);
 209 }
 210
 211 static int
 212 cb_dumphdr(struct md_pa *mdp, int seqnr, void *arg)
 213 {
 214         struct dumperinfo *di = (struct dumperinfo*)arg;
 215         Elf_Phdr phdr;
 216         uint64_t size;
 217         int error;
 218
 219         size = mdp->md_size;
 220         bzero(&phdr, sizeof(phdr));
 221         phdr.p_type = PT_LOAD;
 222         phdr.p_flags = PF_R;                    /* XXX */
 223         phdr.p_offset = fileofs;
 224         phdr.p_vaddr = mdp->md_start;
 225         phdr.p_paddr = mdp->md_start;
 226         phdr.p_filesz = size;
 227         phdr.p_memsz = size;
 228         phdr.p_align = PAGE_SIZE;
 229
 230         error = buf_write(di, (char*)&phdr, sizeof(phdr));
 231         fileofs += phdr.p_filesz;
 232         return (error);
 233 }
 234
 235 static int
 236 cb_size(struct md_pa *mdp, int seqnr, void *arg)
 237 {
 238         uint64_t *sz = (uint64_t*)arg;
 239
 240         *sz += (uint64_t)mdp->md_size;
 241         return (0);
 242 }
 243
 244 static int
 245 foreach_chunk(callback_t cb, void *arg)
 246 {
 247         struct md_pa *mdp;
 248         int error, seqnr;
 249
 250         seqnr = 0;
 251         mdp = md_pa_first();
 252         while (mdp != NULL) {
 253                 error = (*cb)(mdp, seqnr++, arg);
 254                 if (error)
 255                         return (-error);
 256                 mdp = md_pa_next(mdp);
 257         }
 258         return (seqnr);
 259 }
 260
 261 void
 262 md_dumpsys(struct dumperinfo *di)
 263 {
 264         Elf_Ehdr ehdr;
 265         uint64_t dumpsize;
 266         off_t hdrgap;
 267         size_t hdrsz;
 268         int error;
 269
 270         /*
 271          * Save context if dump called without panic.
 272          */
 273         if (dumpthread == NULL) {
 274                 savectx(&dumppcb);
 275                 dumpthread = curthread;
 276         }
 277
 278         if (do_minidump) {
 279                 minidumpsys(di);
 280                 return;
 281         }
 282         bzero(&ehdr, sizeof(ehdr));
 283         ehdr.e_ident[EI_MAG0] = ELFMAG0;
 284         ehdr.e_ident[EI_MAG1] = ELFMAG1;
 285         ehdr.e_ident[EI_MAG2] = ELFMAG2;
 286         ehdr.e_ident[EI_MAG3] = ELFMAG3;
 287         ehdr.e_ident[EI_CLASS] = ELF_CLASS;
 288 #if BYTE_ORDER == LITTLE_ENDIAN
 289         ehdr.e_ident[EI_DATA] = ELFDATA2LSB;
 290 #else
 291         ehdr.e_ident[EI_DATA] = ELFDATA2MSB;
 292 #endif
 293         ehdr.e_ident[EI_VERSION] = EV_CURRENT;
 294         ehdr.e_ident[EI_OSABI] = ELFOSABI_STANDALONE;   /* XXX big picture? */
 295         ehdr.e_type = ET_CORE;
 296         ehdr.e_machine = EM_X86_64;
 297         ehdr.e_phoff = sizeof(ehdr);
 298         ehdr.e_flags = 0;
 299         ehdr.e_ehsize = sizeof(ehdr);
 300         ehdr.e_phentsize = sizeof(Elf_Phdr);
 301         ehdr.e_shentsize = sizeof(Elf_Shdr);
 302
 303         md_pa_init();
 304
 305         /* Calculate dump size. */
 306         dumpsize = 0L;
 307         ehdr.e_phnum = foreach_chunk(cb_size, &dumpsize);
 308         hdrsz = ehdr.e_phoff + ehdr.e_phnum * ehdr.e_phentsize;
 309         fileofs = MD_ALIGN(hdrsz);
 310         dumpsize += fileofs;
 311         hdrgap = fileofs - DEV_ALIGN(hdrsz);
 312
 313         /* Determine dump offset on device. */
 314         if (di->mediasize < SIZEOF_METADATA + dumpsize + sizeof(kdh) * 2) {
 315                 error = ENOSPC;
 316                 goto fail;
 317         }
 318         dumplo = di->mediaoffset + di->mediasize - dumpsize;
 319         dumplo -= sizeof(kdh) * 2;
 320
 321         mkdumpheader(&kdh, KERNELDUMPMAGIC, KERNELDUMP_AMD64_VERSION,
 322             dumpsize, di->blocksize);
 323
 324         kprintf("Dumping %llu MB (%d chunks)\n", (long long)dumpsize >> 20,
 325             ehdr.e_phnum);
 326
 327         /* Dump leader */
 328         error = dev_ddump(di->priv, &kdh, 0, dumplo, sizeof(kdh));
 329         if (error)
 330                 goto fail;
 331         dumplo += sizeof(kdh);
 332
 333         /* Dump ELF header */
 334         error = buf_write(di, (char*)&ehdr, sizeof(ehdr));
 335         if (error)
 336                 goto fail;
 337
 338         /* Dump program headers */
 339         error = foreach_chunk(cb_dumphdr, di);
 340         if (error < 0)
 341                 goto fail;
 342         buf_flush(di);
 343
 344         /*
 345          * All headers are written using blocked I/O, so we know the
 346          * current offset is (still) block aligned. Skip the alignement
 347          * in the file to have the segment contents aligned at page
 348          * boundary. We cannot use MD_ALIGN on dumplo, because we don't
 349          * care and may very well be unaligned within the dump device.
 350          */
 351         dumplo += hdrgap;
 352
 353         /* Dump memory chunks (updates dumplo) */
 354         error = foreach_chunk(cb_dumpdata, di);
 355         if (error < 0)
 356                 goto fail;
 357
 358         /* Dump trailer */
 359         error = dev_ddump(di->priv, &kdh, 0, dumplo, sizeof(kdh));
 360         if (error)
 361                 goto fail;
 362
 363         /* Signal completion, signoff and exit stage left. */
 364         dev_ddump(di->priv, NULL, 0, 0, 0);
 365         kprintf("\nDump complete\n");
 366         return;
 367
 368  fail:
 369         if (error < 0)
 370                 error = -error;
 371
 372         if (error == ECANCELED)
 373                 kprintf("\nDump aborted\n");
 374         else if (error == ENOSPC)
 375                 kprintf("\nDump failed. Partition too small.\n");
 376         else
 377                 kprintf("\n** DUMP FAILED (ERROR %d) **\n", error);
 378 }