summaryrefslogtreecommitdiffstats
path: root/libdwfl/linux-kernel-modules.c
blob: 6edb27f298629743992352e03cb6e6ae6b2b51df (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
/* Standard libdwfl callbacks for debugging the running Linux kernel.
   Copyright (C) 2005-2011, 2013, 2014, 2015 Red Hat, Inc.
   This file is part of elfutils.

   This file is free software; you can redistribute it and/or modify
   it under the terms of either

     * the GNU Lesser General Public License as published by the Free
       Software Foundation; either version 3 of the License, or (at
       your option) any later version

   or

     * the GNU General Public License as published by the Free
       Software Foundation; either version 2 of the License, or (at
       your option) any later version

   or both in parallel, as here.

   elfutils is distributed in the hope that it will be useful, but
   WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   General Public License for more details.

   You should have received copies of the GNU General Public License and
   the GNU Lesser General Public License along with this program.  If
   not, see <http://www.gnu.org/licenses/>.  */

/* In case we have a bad fts we include this before config.h because it
   can't handle _FILE_OFFSET_BITS.
   Everything we need here is fine if its declarations just come first.
   Also, include sys/types.h before fts. On some systems fts.h is not self
   contained. */
#ifdef BAD_FTS
  #include <sys/types.h>
  #include <fts.h>
#endif

#include <config.h>
#include <system.h>

#include "libelfP.h"
#include "libdwflP.h"
#include <inttypes.h>
#include <errno.h>
#include <stdio.h>
#include <stdio_ext.h>
#include <string.h>
#include <stdlib.h>
#include <sys/utsname.h>
#include <fcntl.h>
#include <unistd.h>

/* If fts.h is included before config.h, its indirect inclusions may not
   give us the right LFS aliases of these functions, so map them manually.  */
#ifdef BAD_FTS
  #ifdef _FILE_OFFSET_BITS
    #define open open64
    #define fopen fopen64
  #endif
#else
  #include <sys/types.h>
  #include <fts.h>
#endif


#define KERNEL_MODNAME	"kernel"

#define MODULEDIRFMT	"/lib/modules/%s"

#define KNOTESFILE	"/sys/kernel/notes"
#define	MODNOTESFMT	"/sys/module/%s/notes"
#define KSYMSFILE	"/proc/kallsyms"
#define MODULELIST	"/proc/modules"
#define	SECADDRDIRFMT	"/sys/module/%s/sections/"
#define MODULE_SECT_NAME_LEN 32	/* Minimum any linux/module.h has had.  */


static const char *vmlinux_suffixes[] =
  {
    ".gz",
#ifdef USE_BZLIB
    ".bz2",
#endif
#ifdef USE_LZMA
    ".xz",
#endif
  };

/* Try to open the given file as it is or under the debuginfo directory.  */
static int
try_kernel_name (Dwfl *dwfl, char **fname, bool try_debug)
{
  if (*fname == NULL)
    return -1;

  /* Don't bother trying *FNAME itself here if the path will cause it to be
     tried because we give its own basename as DEBUGLINK_FILE.  */
  int fd = ((((dwfl->callbacks->debuginfo_path
	       ? *dwfl->callbacks->debuginfo_path : NULL)
	      ?: DEFAULT_DEBUGINFO_PATH)[0] == ':') ? -1
	    : TEMP_FAILURE_RETRY (open (*fname, O_RDONLY)));

  if (fd < 0)
    {
      Dwfl_Module fakemod = { .dwfl = dwfl };

      if (try_debug)
	/* Passing NULL for DEBUGLINK_FILE searches for both the basenamer
	   "vmlinux" and the default of basename + ".debug", to look for
	   "vmlinux.debug" files.  */
	fd = INTUSE(dwfl_standard_find_debuginfo) (&fakemod, NULL, NULL, 0,
						   *fname, NULL, 0,
						   &fakemod.debug.name);
      else
	/* Try the file's unadorned basename as DEBUGLINK_FILE,
	   to look only for "vmlinux" files.  */
	fd = INTUSE(dwfl_standard_find_debuginfo) (&fakemod, NULL, NULL, 0,
						   *fname, basename (*fname),
						   0, &fakemod.debug.name);

      if (fakemod.debug.name != NULL)
	{
	  free (*fname);
	  *fname = fakemod.debug.name;
	}
    }

  if (fd < 0)
    for (size_t i = 0;
	 i < sizeof vmlinux_suffixes / sizeof vmlinux_suffixes[0] && fd < 0;
	 ++i)
      {
	char *zname;
	if (asprintf (&zname, "%s%s", *fname, vmlinux_suffixes[i]) > 0)
	  {
	    fd = TEMP_FAILURE_RETRY (open (zname, O_RDONLY));
	    if (fd < 0)
	      free (zname);
	    else
	      {
		free (*fname);
		*fname = zname;
	      }
	  }
      }

  if (fd < 0)
    {
      free (*fname);
      *fname = NULL;
    }

  return fd;
}

static inline const char *
kernel_release (void)
{
#ifdef __linux__
  /* Cache the `uname -r` string we'll use.  */
  static struct utsname utsname;
  if (utsname.release[0] == '\0' && uname (&utsname) != 0)
    return NULL;
  return utsname.release;
#else
  /* Used for finding the running linux kernel, which isn't supported
     on non-linux kernel systems.  */
  errno = ENOTSUP;
  return NULL;
#endif
}

static int
find_kernel_elf (Dwfl *dwfl, const char *release, char **fname)
{
  /* First try to find an uncompressed vmlinux image.  Possibly
     including debuginfo.  */
  if (release == NULL
      || ((release[0] == '/'
	   ? asprintf (fname, "%s/vmlinux", release)
	   : asprintf (fname, "/boot/vmlinux-%s", release)) < 0))
    return -1;

  int fd = try_kernel_name (dwfl, fname, true);
  if (fd < 0 && release[0] != '/')
    {
      free (*fname);
      if (asprintf (fname, MODULEDIRFMT "/vmlinux", release) < 0)
	return -1;
      fd = try_kernel_name (dwfl, fname, true);
    }

  /* There might be a compressed vmlinuz image.  Probably without
     debuginfo, but try to find it under the debug path also, just in
     case.  */
  if (fd < 0)
    {
      free (*fname);
      if ((release[0] == '/'
           ? asprintf (fname, "%s/vmlinuz", release)
           : asprintf (fname, "/boot/vmlinuz-%s", release)) < 0)
        return -1;

      fd = try_kernel_name (dwfl, fname, true);
      if (fd < 0 && release[0] != '/')
	{
	  free (*fname);
	  if (asprintf (fname, MODULEDIRFMT "/vmlinuz", release) < 0)
	    return -1;
	  fd = try_kernel_name (dwfl, fname, true);
	}
    }

  return fd;
}

static int
get_release (Dwfl *dwfl, const char **release)
{
  if (dwfl == NULL)
    return -1;

  const char *release_string = release == NULL ? NULL : *release;
  if (release_string == NULL)
    {
      release_string = kernel_release ();
      if (release_string == NULL)
	return errno;
      if (release != NULL)
	*release = release_string;
    }

  return 0;
}

static int
report_kernel (Dwfl *dwfl, const char **release,
	       int (*predicate) (const char *module, const char *file))
{
  int result = get_release (dwfl, release);
  if (unlikely (result != 0))
    return result;

  if (release == NULL || *release == NULL)
    return EINVAL;

  char *fname;
  int fd = find_kernel_elf (dwfl, *release, &fname);

  if (fd < 0)
    result = ((predicate != NULL && !(*predicate) (KERNEL_MODNAME, NULL))
	      ? 0 : errno ?: ENOENT);
  else
    {
      bool report = true;

      if (predicate != NULL)
	{
	  /* Let the predicate decide whether to use this one.  */
	  int want = (*predicate) (KERNEL_MODNAME, fname);
	  if (want < 0)
	    result = errno;
	  report = want > 0;
	}

      if (report)
	{
	  /* Note that on some architectures (e.g. x86_64) the vmlinux
	     is ET_EXEC, while on others (e.g. ppc64) it is ET_DYN.
	     In both cases the phdr p_vaddr load address will be non-zero.
	     We want the image to be placed as if it was ET_DYN, so
	     pass true for add_p_vaddr which will do the right thing
	     (in combination with a zero base) in either case.  */
	  Dwfl_Module *mod = INTUSE(dwfl_report_elf) (dwfl, KERNEL_MODNAME,
						      fname, fd, 0, true);
	  if (mod == NULL)
	    result = -1;
	  else
	    /* The kernel is ET_EXEC, but always treat it as relocatable.  */
	    mod->e_type = ET_DYN;
	}

      free (fname);

      if (!report || result < 0)
	close (fd);
    }

  return result;
}

/* Look for a kernel debug archive.  If we find one, report all its modules.
   If not, return ENOENT.  */
static int
report_kernel_archive (Dwfl *dwfl, const char **release,
		       int (*predicate) (const char *module, const char *file))
{
  int result = get_release (dwfl, release);
  if (unlikely (result != 0))
    return result;

  if (release == NULL || *release == NULL)
    return EINVAL;

  char *archive;
  int res = (((*release)[0] == '/')
	     ? asprintf (&archive, "%s/debug.a", *release)
	     : asprintf (&archive, MODULEDIRFMT "/debug.a", *release));
  if (unlikely (res < 0))
    return ENOMEM;

  int fd = try_kernel_name (dwfl, &archive, false);
  if (fd < 0)
    result = errno ?: ENOENT;
  else
    {
      /* We have the archive file open!  */
      Dwfl_Module *last = __libdwfl_report_offline (dwfl, NULL, archive, fd,
						    true, predicate);
      if (unlikely (last == NULL))
	result = -1;
      else
	{
	  /* Find the kernel and move it to the head of the list.  */
	  Dwfl_Module **tailp = &dwfl->modulelist, **prevp = tailp;
	  for (Dwfl_Module *m = *prevp; m != NULL; m = *(prevp = &m->next))
	    if (!m->gc && m->e_type != ET_REL && !strcmp (m->name, "kernel"))
	      {
		*prevp = m->next;
		m->next = *tailp;
		*tailp = m;
		break;
	      }
	}
    }

  free (archive);
  return result;
}

static size_t
check_suffix (const FTSENT *f, size_t namelen)
{
#define TRY(sfx)							\
  if ((namelen ? f->fts_namelen == namelen + sizeof sfx - 1		\
       : f->fts_namelen >= sizeof sfx)					\
      && !memcmp (f->fts_name + f->fts_namelen - (sizeof sfx - 1),	\
		  sfx, sizeof sfx))					\
    return sizeof sfx - 1

  TRY (".ko");
  TRY (".ko.gz");
#if USE_BZLIB
  TRY (".ko.bz2");
#endif
#if USE_LZMA
  TRY (".ko.xz");
#endif
#if USE_ZSTD
  TRY (".ko.zst");
#endif

  return 0;

#undef	TRY
}

/* Report a kernel and all its modules found on disk, for offline use.
   If RELEASE starts with '/', it names a directory to look in;
   if not, it names a directory to find under /lib/modules/;
   if null, /lib/modules/`uname -r` is used.
   Returns zero on success, -1 if dwfl_report_module failed,
   or an errno code if finding the files on disk failed.  */
int
dwfl_linux_kernel_report_offline (Dwfl *dwfl, const char *release,
				  int (*predicate) (const char *module,
						    const char *file))
{
  int result = report_kernel_archive (dwfl, &release, predicate);
  if (result != ENOENT)
    return result;

  /* First report the kernel.  */
  result = report_kernel (dwfl, &release, predicate);
  if (result == 0)
    {
      /* Do "find /lib/modules/RELEASE -name *.ko".  */

      char *modulesdir[] = { NULL, NULL };
      if (release[0] == '/')
	modulesdir[0] = (char *) release;
      else
	{
	  if (asprintf (&modulesdir[0], MODULEDIRFMT, release) < 0)
	    return errno;
	}

      FTS *fts = fts_open (modulesdir, FTS_NOSTAT | FTS_LOGICAL, NULL);
      if (modulesdir[0] == (char *) release)
	modulesdir[0] = NULL;
      if (fts == NULL)
	{
	  free (modulesdir[0]);
	  return errno;
	}

      FTSENT *f;
      while ((f = fts_read (fts)) != NULL)
	{
	  /* Skip a "source" subtree, which tends to be large.
	     This insane hard-coding of names is what depmod does too.  */
	  if (f->fts_namelen == sizeof "source" - 1
	      && !strcmp (f->fts_name, "source"))
	    {
	      fts_set (fts, f, FTS_SKIP);
	      continue;
	    }

	  switch (f->fts_info)
	    {
	    case FTS_F:
	    case FTS_SL:
	    case FTS_NSOK:;
	      /* See if this file name matches "*.ko".  */
	      const size_t suffix = check_suffix (f, 0);
	      if (suffix)
		{
		  /* We have a .ko file to report.  Following the algorithm
		     by which the kernel makefiles set KBUILD_MODNAME, we
		     replace all ',' or '-' with '_' in the file name and
		     call that the module name.  Modules could well be
		     built using different embedded names than their file
		     names.  To handle that, we would have to look at the
		     __this_module.name contents in the module's text.  */

		  char *name = strndup (f->fts_name, f->fts_namelen - suffix);
		  if (unlikely (name == NULL))
		    {
		      __libdwfl_seterrno (DWFL_E_NOMEM);
		      result = -1;
		      break;
		    }
		  for (size_t i = 0; i < f->fts_namelen - suffix; ++i)
		    if (name[i] == '-' || name[i] == ',')
		      name[i] = '_';

		  if (predicate != NULL)
		    {
		      /* Let the predicate decide whether to use this one.  */
		      int want = (*predicate) (name, f->fts_path);
		      if (want < 0)
			{
			  result = -1;
			  free (name);
			  break;
			}
		      if (!want)
			{
			  free (name);
			  continue;
			}
		    }

		  if (dwfl_report_offline (dwfl, name, f->fts_path, -1) == NULL)
		    {
		      free (name);
		      result = -1;
		      break;
		    }
		  free (name);
		}
	      continue;

	    case FTS_ERR:
	    case FTS_DNR:
	    case FTS_NS:
	      result = f->fts_errno;
	      break;

	    case FTS_SLNONE:
	    default:
	      continue;
	    }

	  /* We only get here in error cases.  */
	  break;
	}
      fts_close (fts);
      free (modulesdir[0]);
    }

  return result;
}
INTDEF (dwfl_linux_kernel_report_offline)


/* State of read_address used by intuit_kernel_bounds. */
struct read_address_state {
  FILE *f;
  char *line;
  size_t linesz;
  size_t n;
  char *p;
  const char *type;
};

static inline bool
read_address (struct read_address_state *state, Dwarf_Addr *addr)
{
  if ((state->n = getline (&state->line, &state->linesz, state->f)) < 1 ||
      state->line[state->n - 2] == ']')
    return false;
  *addr = strtoull (state->line, &state->p, 16);
  state->p += strspn (state->p, " \t");
  state->type = strsep (&state->p, " \t\n");
  if (state->type == NULL)
    return false;
  return state->p != NULL && state->p != state->line;
}


/* Grovel around to guess the bounds of the runtime kernel image.  */
static int
intuit_kernel_bounds (Dwarf_Addr *start, Dwarf_Addr *end, Dwarf_Addr *notes)
{
  struct read_address_state state = { NULL, NULL, 0, 0, NULL, NULL };

  *notes = 0;

  state.f = fopen (KSYMSFILE, "r");
  if (state.f == NULL)
    return errno;

  (void) __fsetlocking (state.f, FSETLOCKING_BYCALLER);

  int result;
  do
    result = read_address (&state, start) ? 0 : -1;
  while (result == 0 && strchr ("TtRr", *state.type) == NULL);

  if (result == 0)
    {
      Dwarf_Addr addr;
      *end = *start;
      while (read_address (&state, &addr) && addr >= *end)
	{
	  *end = addr;
	  if (*notes == 0 && !strcmp (state.p, "__start_notes\n"))
	    *notes = *end;
	}

      Dwarf_Addr round_kernel = sysconf (_SC_PAGESIZE);
      *start &= -(Dwarf_Addr) round_kernel;
      *end += round_kernel - 1;
      *end &= -(Dwarf_Addr) round_kernel;
      if (*start >= *end || *end - *start < round_kernel)
	result = -1;
    }
  free (state.line);

  if (result == -1)
    result = ferror_unlocked (state.f) ? errno : ENOEXEC;

  fclose (state.f);

  return result;
}


/* Look for a build ID note in NOTESFILE and associate the ID with MOD.  */
static int
check_notes (Dwfl_Module *mod, const char *notesfile,
	     Dwarf_Addr vaddr, const char *secname)
{
  int fd = open (notesfile, O_RDONLY);
  if (fd < 0)
    return 1;

  assert (sizeof (Elf32_Nhdr) == sizeof (GElf_Nhdr));
  assert (sizeof (Elf64_Nhdr) == sizeof (GElf_Nhdr));
  union
  {
    GElf_Nhdr nhdr;
    unsigned char data[8192];
  } buf;

  ssize_t n = read (fd, buf.data, sizeof buf);
  close (fd);

  if (n <= 0)
    return 1;

  unsigned char *p = buf.data;
  size_t len = 0;
  while (p < &buf.data[n])
    {
      /* No translation required since we are reading the native kernel.  */
      GElf_Nhdr *nhdr = (void *) p;
      len += sizeof *nhdr;
      p += len;
      unsigned char *name = p;
      unsigned char *bits;
      /* This is somewhat ugly, GNU Property notes use different padding,
	 but all we have is the file content, so we have to actually check
	 the name and type.  */
      if (nhdr->n_type == NT_GNU_PROPERTY_TYPE_0
          && nhdr->n_namesz == sizeof "GNU"
          && name + nhdr->n_namesz < &buf.data[n]
          && !memcmp (name, "GNU", sizeof "GNU"))
	{
	  len += nhdr->n_namesz;
	  len = NOTE_ALIGN8 (len);
	  p = buf.data + len;
	  bits = p;
	  len += nhdr->n_descsz;
	  len = NOTE_ALIGN8 (len);
	  p = buf.data + len;
	}
      else
	{
	  len += nhdr->n_namesz;
	  len = NOTE_ALIGN4 (len);
	  p = buf.data + len;
	  bits = p;
	  len += nhdr->n_descsz;
	  len = NOTE_ALIGN4 (len);
	  p = buf.data + len;
	}

      if (p <= &buf.data[n]
	  && nhdr->n_type == NT_GNU_BUILD_ID
	  && nhdr->n_namesz == sizeof "GNU"
	  && !memcmp (name, "GNU", sizeof "GNU"))
	{
	  /* Found it.  For a module we must figure out its VADDR now.  */

	  if (secname != NULL
	      && (INTUSE(dwfl_linux_kernel_module_section_address)
		  (mod, NULL, mod->name, 0, secname, 0, NULL, &vaddr) != 0
		  || vaddr == (GElf_Addr) -1l))
	    vaddr = 0;

	  if (vaddr != 0)
	    vaddr += bits - buf.data;
	  return INTUSE(dwfl_module_report_build_id) (mod, bits,
						      nhdr->n_descsz, vaddr);
	}
    }

  return 0;
}

/* Look for a build ID for the kernel.  */
static int
check_kernel_notes (Dwfl_Module *kernelmod, GElf_Addr vaddr)
{
  return check_notes (kernelmod, KNOTESFILE, vaddr, NULL) < 0 ? -1 : 0;
}

/* Look for a build ID for a loaded kernel module.  */
static int
check_module_notes (Dwfl_Module *mod)
{
  char *dirs[2] = { NULL, NULL };
  if (asprintf (&dirs[0], MODNOTESFMT, mod->name) < 0)
    return ENOMEM;

  FTS *fts = fts_open (dirs, FTS_NOSTAT | FTS_LOGICAL, NULL);
  if (fts == NULL)
    {
      free (dirs[0]);
      return 0;
    }

  int result = 0;
  FTSENT *f;
  while ((f = fts_read (fts)) != NULL)
    {
      switch (f->fts_info)
	{
	case FTS_F:
	case FTS_SL:
	case FTS_NSOK:
	  result = check_notes (mod, f->fts_accpath, 0, f->fts_name);
	  if (result > 0)	/* Nothing found.  */
	    {
	      result = 0;
	      continue;
	    }
	  break;

	case FTS_ERR:
	case FTS_DNR:
	  result = f->fts_errno;
	  break;

	case FTS_NS:
	case FTS_SLNONE:
	default:
	  continue;
	}

      /* We only get here when finished or in error cases.  */
      break;
    }
  fts_close (fts);
  free (dirs[0]);

  return result;
}

int
dwfl_linux_kernel_report_kernel (Dwfl *dwfl)
{
  Dwarf_Addr start = 0;
  Dwarf_Addr end = 0;

  #define report() \
    (INTUSE(dwfl_report_module) (dwfl, KERNEL_MODNAME, start, end))

  /* This is a bit of a kludge.  If we already reported the kernel,
     don't bother figuring it out again--it never changes.  */
  for (Dwfl_Module *m = dwfl->modulelist; m != NULL; m = m->next)
    if (!strcmp (m->name, KERNEL_MODNAME))
      {
	start = m->low_addr;
	end = m->high_addr;
	return report () == NULL ? -1 : 0;
      }

  /* Try to figure out the bounds of the kernel image without
     looking for any vmlinux file.  */
  Dwarf_Addr notes;
  int result = intuit_kernel_bounds (&start, &end, &notes);
  if (result == 0)
    {
      Dwfl_Module *mod = report ();
      return unlikely (mod == NULL) ? -1 : check_kernel_notes (mod, notes);
    }
  if (result != ENOENT)
    return result;

  /* Find the ELF file for the running kernel and dwfl_report_elf it.  */
  return report_kernel (dwfl, NULL, NULL);
}
INTDEF (dwfl_linux_kernel_report_kernel)


static inline bool
subst_name (char from, char to,
            const char * const module_name,
            char * const alternate_name,
            const size_t namelen)
{
  const char *n = memchr (module_name, from, namelen);
  if (n == NULL)
    return false;
  char *a = mempcpy (alternate_name, module_name, n - module_name);
  *a++ = to;
  ++n;
  const char *p;
  while ((p = memchr (n, from, namelen - (n - module_name))) != NULL)
    {
      a = mempcpy (a, n, p - n);
      *a++ = to;
      n = p + 1;
    }
  memcpy (a, n, namelen - (n - module_name) + 1);
  return true;
}

/* Dwfl_Callbacks.find_elf for the running Linux kernel and its modules.  */

int
dwfl_linux_kernel_find_elf (Dwfl_Module *mod,
			    void **userdata __attribute__ ((unused)),
			    const char *module_name,
			    Dwarf_Addr base __attribute__ ((unused)),
			    char **file_name, Elf **elfp)
{
  if (mod->build_id_len > 0)
    {
      int fd = INTUSE(dwfl_build_id_find_elf) (mod, NULL, NULL, 0,
					       file_name, elfp);
      if (fd >= 0 || mod->main.elf != NULL || errno != 0)
	return fd;
    }

  const char *release = kernel_release ();
  if (release == NULL)
    return errno;

  if (!strcmp (module_name, KERNEL_MODNAME))
    return find_kernel_elf (mod->dwfl, release, file_name);

  /* Do "find /lib/modules/`uname -r` -name MODULE_NAME.ko".  */

  char *modulesdir[] = { NULL, NULL };
  if (asprintf (&modulesdir[0], MODULEDIRFMT, release) < 0)
    return -1;

  FTS *fts = fts_open (modulesdir, FTS_NOSTAT | FTS_LOGICAL, NULL);
  if (fts == NULL)
    {
      free (modulesdir[0]);
      return -1;
    }

  size_t namelen = strlen (module_name);

  /* This is a kludge.  There is no actual necessary relationship between
     the name of the .ko file installed and the module name the kernel
     knows it by when it's loaded.  The kernel's only idea of the module
     name comes from the name embedded in the object's magic
     .gnu.linkonce.this_module section.

     In practice, these module names match the .ko file names except for
     some using '_' and some using '-'.  So our cheap kludge is to look for
     two files when either a '_' or '-' appears in a module name, one using
     only '_' and one only using '-'.  */

  char *alternate_name = malloc (namelen + 1);
  if (unlikely (alternate_name == NULL))
    {
      free (modulesdir[0]);
      return ENOMEM;
    }
  if (!subst_name ('-', '_', module_name, alternate_name, namelen) &&
      !subst_name ('_', '-', module_name, alternate_name, namelen))
    alternate_name[0] = '\0';

  FTSENT *f;
  int error = ENOENT;
  while ((f = fts_read (fts)) != NULL)
    {
      /* Skip a "source" subtree, which tends to be large.
	 This insane hard-coding of names is what depmod does too.  */
      if (f->fts_namelen == sizeof "source" - 1
	  && !strcmp (f->fts_name, "source"))
	{
	  fts_set (fts, f, FTS_SKIP);
	  continue;
	}

      error = ENOENT;
      switch (f->fts_info)
	{
	case FTS_F:
	case FTS_SL:
	case FTS_NSOK:
	  /* See if this file name is "MODULE_NAME.ko".  */
	  if (check_suffix (f, namelen)
	      && (!memcmp (f->fts_name, module_name, namelen)
		  || !memcmp (f->fts_name, alternate_name, namelen)))
	    {
	      int fd = open (f->fts_accpath, O_RDONLY);
	      *file_name = strdup (f->fts_path);
	      fts_close (fts);
	      free (modulesdir[0]);
	      free (alternate_name);
	      if (fd < 0)
		free (*file_name);
	      else if (*file_name == NULL)
		{
		  close (fd);
		  fd = -1;
		}
	      return fd;
	    }
	  break;

	case FTS_ERR:
	case FTS_DNR:
	case FTS_NS:
	  error = f->fts_errno;
	  break;

	case FTS_SLNONE:
	default:
	  break;
	}
    }

  fts_close (fts);
  free (modulesdir[0]);
  free (alternate_name);
  errno = error;
  return -1;
}
INTDEF (dwfl_linux_kernel_find_elf)


/* Dwfl_Callbacks.section_address for kernel modules in the running Linux.
   We read the information from /sys/module directly.  */

int
dwfl_linux_kernel_module_section_address
(Dwfl_Module *mod __attribute__ ((unused)),
 void **userdata __attribute__ ((unused)),
 const char *modname, Dwarf_Addr base __attribute__ ((unused)),
 const char *secname, Elf32_Word shndx __attribute__ ((unused)),
 const GElf_Shdr *shdr __attribute__ ((unused)),
 Dwarf_Addr *addr)
{
  char *sysfile;
  if (asprintf (&sysfile, SECADDRDIRFMT "%s", modname, secname) < 0)
    return DWARF_CB_ABORT;

  FILE *f = fopen (sysfile, "r");
  free (sysfile);

  if (f == NULL)
    {
      if (errno == ENOENT)
	{
	  /* The .modinfo and .data.percpu sections are never kept
	     loaded in the kernel.  If the kernel was compiled without
	     CONFIG_MODULE_UNLOAD, the .exit.* sections are not
	     actually loaded at all.

	     Setting *ADDR to -1 tells the caller this section is
	     actually absent from memory.  */

	  if (!strcmp (secname, ".modinfo")
	      || !strcmp (secname, ".data.percpu")
	      || !strncmp (secname, ".exit", 5))
	    {
	      *addr = (Dwarf_Addr) -1l;
	      return DWARF_CB_OK;
	    }

	  /* The goofy PPC64 module_frob_arch_sections function tweaks
	     the section names as a way to control other kernel code's
	     behavior, and this cruft leaks out into the /sys information.
	     The file name for ".init*" may actually look like "_init*".  */

	  const bool is_init = !strncmp (secname, ".init", 5);
	  if (is_init)
	    {
	      if (asprintf (&sysfile, SECADDRDIRFMT "_%s",
			    modname, &secname[1]) < 0)
		return ENOMEM;
	      f = fopen (sysfile, "r");
	      free (sysfile);
	      if (f != NULL)
		goto ok;
	    }

	  /* The kernel truncates section names to MODULE_SECT_NAME_LEN - 1.
	     In case that size increases in the future, look for longer
	     truncated names first.  */
	  size_t namelen = strlen (secname);
	  if (namelen >= MODULE_SECT_NAME_LEN)
	    {
	      int len = asprintf (&sysfile, SECADDRDIRFMT "%s",
				  modname, secname);
	      if (len < 0)
		return DWARF_CB_ABORT;
	      char *end = sysfile + len;
	      do
		{
		  *--end = '\0';
		  f = fopen (sysfile, "r");
		  if (is_init && f == NULL && errno == ENOENT)
		    {
		      sysfile[len - namelen] = '_';
		      f = fopen (sysfile, "r");
		      sysfile[len - namelen] = '.';
		    }
		}
	      while (f == NULL && errno == ENOENT
		     && end - &sysfile[len - namelen] >= MODULE_SECT_NAME_LEN);
	      free (sysfile);

	      if (f != NULL)
		goto ok;
	    }
	}

      return DWARF_CB_ABORT;
    }

 ok:
  (void) __fsetlocking (f, FSETLOCKING_BYCALLER);

  int result = (fscanf (f, "%" PRIx64 "\n", addr) == 1 ? 0
		: ferror_unlocked (f) ? errno : ENOEXEC);
  fclose (f);

  if (result == 0)
    return DWARF_CB_OK;

  errno = result;
  return DWARF_CB_ABORT;
}
INTDEF (dwfl_linux_kernel_module_section_address)

int
dwfl_linux_kernel_report_modules (Dwfl *dwfl)
{
  FILE *f = fopen (MODULELIST, "r");
  if (f == NULL)
    return errno;

  (void) __fsetlocking (f, FSETLOCKING_BYCALLER);

  int result = 0;
  Dwarf_Addr modaddr;
  unsigned long int modsz;
  char modname[128];
  char *line = NULL;
  size_t linesz = 0;
  /* We can't just use fscanf here because it's not easy to distinguish \n
     from other whitespace so as to take the optional word following the
     address but always stop at the end of the line.  */
  while (getline (&line, &linesz, f) > 0
	 && sscanf (line, "%128s %lu %*s %*s %*s %" PRIx64 " %*s\n",
		    modname, &modsz, &modaddr) == 3)
    {
      Dwfl_Module *mod = INTUSE(dwfl_report_module) (dwfl, modname,
						     modaddr, modaddr + modsz);
      if (mod == NULL)
	{
	  result = -1;
	  break;
	}

      result = check_module_notes (mod);
    }
  free (line);

  if (result == 0)
    result = ferror_unlocked (f) ? errno : feof_unlocked (f) ? 0 : ENOEXEC;

  fclose (f);

  return result;
}
INTDEF (dwfl_linux_kernel_report_modules)