sys/platform/pc32/i386/minidump_machdep.c

   1 /*-
   2  * Copyright (c) 2006 Peter Wemm
   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/i386/i386/minidump_machdep.c,v 1.9 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/device.h>
  34 #include <sys/globaldata.h>
  35 #include <sys/kernel.h>
  36 #include <sys/kerneldump.h>
  37 #include <sys/msgbuf.h>
  38 #include <vm/vm.h>
  39 #include <vm/vm_kern.h>
  40 #include <vm/pmap.h>
  41 #include <machine/atomic.h>
  42 #include <machine/elf.h>
  43 #include <machine/globaldata.h>
  44 #include <machine/md_var.h>
  45 #include <machine/vmparam.h>
  46 #include <machine/minidump.h>
  47
  48 CTASSERT(sizeof(struct kerneldumpheader) == 512);
  49
  50 /*
  51  * Don't touch the first SIZEOF_METADATA bytes on the dump device. This
  52  * is to protect us from metadata and to protect metadata from us.
  53  */
  54 #define SIZEOF_METADATA         (64*1024)
  55
  56 #define MD_ALIGN(x)     (((off_t)(x) + PAGE_MASK) & ~PAGE_MASK)
  57 #define DEV_ALIGN(x)    (((off_t)(x) + (DEV_BSIZE-1)) & ~(DEV_BSIZE-1))
  58
  59 uint32_t *vm_page_dump;
  60 int vm_page_dump_size;
  61
  62 static struct kerneldumpheader kdh;
  63 static off_t dumplo;
  64
  65 /* Handle chunked writes. */
  66 static size_t fragsz;
  67 static void *dump_va;
  68 static uint64_t counter, progress;
  69
  70 CTASSERT(sizeof(*vm_page_dump) == 4);
  71
  72 static int
  73 is_dumpable(vm_paddr_t pa)
  74 {
  75         int i;
  76
  77         for (i = 0; dump_avail[i] != 0 || dump_avail[i + 1] != 0; i += 2) {
  78                 if (pa >= dump_avail[i] && pa < dump_avail[i + 1])
  79                         return (1);
  80         }
  81         return (0);
  82 }
  83
  84 #define PG2MB(pgs) (((pgs) + (1 << 8) - 1) >> 8)
  85
  86 static int
  87 blk_flush(struct dumperinfo *di)
  88 {
  89         int error;
  90
  91         if (fragsz == 0)
  92                 return (0);
  93
  94         error = dev_ddump(di->priv, dump_va, 0, dumplo, fragsz);
  95         dumplo += fragsz;
  96         fragsz = 0;
  97         return (error);
  98 }
  99
 100 static int
 101 blk_write(struct dumperinfo *di, char *ptr, vm_paddr_t pa, size_t sz)
 102 {
 103         size_t len;
 104         int error, i, c;
 105
 106         error = 0;
 107         if ((sz % PAGE_SIZE) != 0) {
 108                 kprintf("size not page aligned\n");
 109                 return (EINVAL);
 110         }
 111         if (ptr != NULL && pa != 0) {
 112                 kprintf("cant have both va and pa!\n");
 113                 return (EINVAL);
 114         }
 115         if (pa != 0 && (((uintptr_t)ptr) % PAGE_SIZE) != 0) {
 116                 kprintf("address not page aligned\n");
 117                 return (EINVAL);
 118         }
 119         if (ptr != NULL) {
 120                 /* If we're doing a virtual dump, flush any pre-existing pa pages */
 121                 error = blk_flush(di);
 122                 if (error)
 123                         return (error);
 124         }
 125         while (sz) {
 126                 len = (MAXDUMPPGS * PAGE_SIZE) - fragsz;
 127                 if (len > sz)
 128                         len = sz;
 129                 counter += len;
 130                 progress -= len;
 131                 if (counter >> 24) {
 132                         kprintf(" %lld", PG2MB(progress >> PAGE_SHIFT));
 133                         counter &= (1<<24) - 1;
 134                 }
 135                 if (ptr) {
 136                         error = dev_ddump(di->priv, ptr, 0, dumplo, len);
 137                         if (error)
 138                                 return (error);
 139                         dumplo += len;
 140                         ptr += len;
 141                         sz -= len;
 142                 } else {
 143                         for (i = 0; i < len; i += PAGE_SIZE)
 144                                 dump_va = pmap_kenter_temporary(pa + i, (i + fragsz) >> PAGE_SHIFT);
 145                         fragsz += len;
 146                         pa += len;
 147                         sz -= len;
 148                         if (fragsz == (MAXDUMPPGS * PAGE_SIZE)) {
 149                                 error = blk_flush(di);
 150                                 if (error)
 151                                         return (error);
 152                         }
 153                 }
 154
 155                 /* Check for user abort. */
 156                 c = cncheckc();
 157                 if (c == 0x03)
 158                         return (ECANCELED);
 159                 if (c != -1)
 160                         kprintf(" (CTRL-C to abort) ");
 161         }
 162
 163         return (0);
 164 }
 165
 166 /* A fake page table page, to avoid having to handle both 4K and 2M pages */
 167 static pt_entry_t fakept[NPTEPG];
 168
 169 void
 170 minidumpsys(struct dumperinfo *di)
 171 {
 172         uint64_t dumpsize;
 173         uint32_t ptesize;
 174         vm_offset_t va;
 175         vm_offset_t kern_end;
 176         int error;
 177         uint32_t bits;
 178         uint64_t pa;
 179         pd_entry_t *pd;
 180         pt_entry_t *pt;
 181         int i, j, k, bit;
 182         struct minidumphdr mdhdr;
 183         struct mdglobaldata *md;
 184
 185         counter = 0;
 186         ptesize = 0;
 187
 188         md = (struct mdglobaldata *)globaldata_find(0);
 189
 190         kern_end = kernel_vm_end;
 191         if (kern_end < (vm_offset_t)&(md[ncpus]))
 192                 kern_end = (vm_offset_t)&(md[ncpus]);
 193 #if 0
 194         kern_end = 0xFFFFF000;
 195 #endif
 196
 197         /* Walk page table pages, set bits in vm_page_dump */
 198         for (va = KERNBASE; va < kern_end; va += NBPDR) {
 199                 /*
 200                  * We always write a page, even if it is zero. Each
 201                  * page written corresponds to 2MB of space
 202                  */
 203                 ptesize += PAGE_SIZE;
 204                 pd = (pd_entry_t *)((uintptr_t)IdlePTD + KERNBASE);     /* always mapped! */
 205                 j = va >> PDRSHIFT;
 206                 if ((pd[j] & (PG_PS | PG_V)) == (PG_PS | PG_V))  {
 207                         /* This is an entire 2M page. */
 208                         pa = pd[j] & PG_FRAME & ~PDRMASK;
 209                         for (k = 0; k < NPTEPG; k++) {
 210                                 if (is_dumpable(pa))
 211                                         dump_add_page(pa);
 212                                 pa += PAGE_SIZE;
 213                         }
 214                         continue;
 215                 }
 216                 if ((pd[j] & PG_V) == PG_V) {
 217                         /* set bit for each valid page in this 2MB block */
 218                         pt = pmap_kenter_temporary(pd[j] & PG_FRAME, 0);
 219                         for (k = 0; k < NPTEPG; k++) {
 220                                 if ((pt[k] & PG_V) == PG_V) {
 221                                         pa = pt[k] & PG_FRAME;
 222                                         if (is_dumpable(pa))
 223                                                 dump_add_page(pa);
 224                                 }
 225                         }
 226                 } else {
 227                         /* nothing, we're going to dump a null page */
 228                 }
 229         }
 230
 231         /* Calculate dump size. */
 232         dumpsize = ptesize;
 233         dumpsize += round_page(msgbufp->msg_size);
 234         dumpsize += round_page(vm_page_dump_size);
 235         for (i = 0; i < vm_page_dump_size / sizeof(*vm_page_dump); i++) {
 236                 bits = vm_page_dump[i];
 237                 while (bits) {
 238                         bit = bsfl(bits);
 239                         pa = (((uint64_t)i * sizeof(*vm_page_dump) * NBBY) + bit) * PAGE_SIZE;
 240                         /* Clear out undumpable pages now if needed */
 241                         if (is_dumpable(pa)) {
 242                                 dumpsize += PAGE_SIZE;
 243                         } else {
 244                                 dump_drop_page(pa);
 245                         }
 246                         bits &= ~(1ul << bit);
 247                 }
 248         }
 249         dumpsize += PAGE_SIZE;
 250
 251         /* Determine dump offset on device. */
 252         if (di->mediasize < SIZEOF_METADATA + dumpsize + sizeof(kdh) * 2) {
 253                 error = ENOSPC;
 254                 goto fail;
 255         }
 256         dumplo = di->mediaoffset + di->mediasize - dumpsize;
 257         dumplo -= sizeof(kdh) * 2;
 258         progress = dumpsize;
 259
 260         /* Initialize mdhdr */
 261         bzero(&mdhdr, sizeof(mdhdr));
 262         strcpy(mdhdr.magic, MINIDUMP_MAGIC);
 263         mdhdr.version = MINIDUMP_VERSION;
 264         mdhdr.msgbufsize = msgbufp->msg_size;
 265         mdhdr.bitmapsize = vm_page_dump_size;
 266         mdhdr.ptesize = ptesize;
 267         mdhdr.kernbase = KERNBASE;
 268
 269         mkdumpheader(&kdh, KERNELDUMPMAGIC, KERNELDUMP_I386_VERSION,
 270             dumpsize, di->blocksize);
 271
 272         kprintf("Physical memory: %jd MB\n", (intmax_t)ptoa(physmem) / 1048576);
 273         kprintf("Dumping %jd MB:", (intmax_t)dumpsize >> 20);
 274
 275         /* Dump leader */
 276         error = dev_ddump(di->priv, &kdh, 0, dumplo, sizeof(kdh));
 277         if (error)
 278                 goto fail;
 279         dumplo += sizeof(kdh);
 280
 281         /* Dump my header */
 282         bzero(&fakept, sizeof(fakept));
 283         bcopy(&mdhdr, &fakept, sizeof(mdhdr));
 284         error = blk_write(di, (char *)&fakept, 0, PAGE_SIZE);
 285         if (error)
 286                 goto fail;
 287
 288         /* Dump msgbuf up front */
 289         error = blk_write(di, (char *)msgbufp->msg_ptr, 0, round_page(msgbufp->msg_size));
 290         if (error)
 291                 goto fail;
 292
 293         /* Dump bitmap */
 294         error = blk_write(di, (char *)vm_page_dump, 0, round_page(vm_page_dump_size));
 295         if (error)
 296                 goto fail;
 297
 298         /* Dump kernel page table pages */
 299         for (va = KERNBASE; va < kern_end; va += NBPDR) {
 300                 /* We always write a page, even if it is zero */
 301                 pd = (pd_entry_t *)((uintptr_t)IdlePTD + KERNBASE);     /* always mapped! */
 302                 j = va >> PDRSHIFT;
 303                 if ((pd[j] & (PG_PS | PG_V)) == (PG_PS | PG_V))  {
 304                         /* This is a single 2M block. Generate a fake PTP */
 305                         pa = pd[j] & PG_FRAME & ~PDRMASK;
 306                         for (k = 0; k < NPTEPG; k++) {
 307                                 fakept[k] = (pa + (k * PAGE_SIZE)) | PG_V | PG_RW | PG_A | PG_M;
 308                         }
 309                         error = blk_write(di, (char *)&fakept, 0, PAGE_SIZE);
 310                         if (error)
 311                                 goto fail;
 312                         /* flush, in case we reuse fakept in the same block */
 313                         error = blk_flush(di);
 314                         if (error)
 315                                 goto fail;
 316                         continue;
 317                 }
 318                 if ((pd[j] & PG_V) == PG_V) {
 319                         pa = pd[j] & PG_FRAME;
 320                         error = blk_write(di, 0, pa, PAGE_SIZE);
 321                         if (error)
 322                                 goto fail;
 323                 } else {
 324                         bzero(fakept, sizeof(fakept));
 325                         error = blk_write(di, (char *)&fakept, 0, PAGE_SIZE);
 326                         if (error)
 327                                 goto fail;
 328                         /* flush, in case we reuse fakept in the same block */
 329                         error = blk_flush(di);
 330                         if (error)
 331                                 goto fail;
 332                 }
 333         }
 334
 335         /* Dump memory chunks */
 336         /* XXX cluster it up and use blk_dump() */
 337         for (i = 0; i < vm_page_dump_size / sizeof(*vm_page_dump); i++) {
 338                 bits = vm_page_dump[i];
 339                 while (bits) {
 340                         bit = bsfl(bits);
 341                         pa = (((uint64_t)i * sizeof(*vm_page_dump) * NBBY) + bit) * PAGE_SIZE;
 342                         error = blk_write(di, 0, pa, PAGE_SIZE);
 343                         if (error)
 344                                 goto fail;
 345                         bits &= ~(1ul << bit);
 346                 }
 347         }
 348
 349         error = blk_flush(di);
 350         if (error)
 351                 goto fail;
 352
 353         /* Dump trailer */
 354         error = dev_ddump(di->priv, &kdh, 0, dumplo, sizeof(kdh));
 355         if (error)
 356                 goto fail;
 357         dumplo += sizeof(kdh);
 358
 359         /* Signal completion, signoff and exit stage left. */
 360         dev_ddump(di->priv, NULL, 0, 0, 0);
 361         kprintf("\nDump complete\n");
 362         return;
 363
 364  fail:
 365         if (error < 0)
 366                 error = -error;
 367
 368         if (error == ECANCELED)
 369                 kprintf("\nDump aborted\n");
 370         else if (error == ENOSPC)
 371                 kprintf("\nDump failed. Partition too small.\n");
 372         else
 373                 kprintf("\n** DUMP FAILED (ERROR %d) **\n", error);
 374 }
 375
 376 void
 377 dump_add_page(vm_paddr_t pa)
 378 {
 379         int idx, bit;
 380
 381         pa >>= PAGE_SHIFT;
 382         idx = pa >> 5;          /* 2^5 = 32 */
 383         bit = pa & 31;
 384         atomic_set_int(&vm_page_dump[idx], 1ul << bit);
 385 }
 386
 387 void
 388 dump_drop_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_clear_int(&vm_page_dump[idx], 1ul << bit);
 396 }