/* CTF type deduplication.

   Copyright (C) 2019-2023 Free Software Foundation, Inc.

   This file is part of libctf.

   libctf is free software; you can redistribute it and/or modify it under

   the terms of the GNU General Public License as published by the Free

   Software Foundation; either version 3, or (at your option) any later

   version.

   This program 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 a copy of the GNU General Public License

   along with this program; see the file COPYING.  If not see

   <http://www.gnu.org/licenses/>.  */

#include <ctf-impl.h>

#include <string.h>

#include <errno.h>

#include <assert.h>

#include "hashtab.h"

/* (In the below, relevant functions are named in square brackets.)  */

/* Type deduplication is a three-phase process:

    [ctf_dedup, ctf_dedup_hash_type, ctf_dedup_rhash_type]

    1) come up with unambiguous hash values for all types: no two types may have

       the same hash value, and any given type should have only one hash value

       (for optimal deduplication).

    [ctf_dedup, ctf_dedup_detect_name_ambiguity,

     ctf_dedup_conflictify_unshared, ctf_dedup_mark_conflicting_hash]

    2) mark those distinct types with names that collide (and thus cannot be

       declared simultaneously in the same translation unit) as conflicting, and

       recursively mark all types that cite one of those types as conflicting as

       well.  Possibly mark all types cited in only one TU as conflicting, if

       the CTF_LINK_SHARE_DUPLICATED link mode is active.

    [ctf_dedup_emit, ctf_dedup_emit_struct_members, ctf_dedup_id_to_target]

    3) emit all the types, one hash value at a time.  Types not marked

       conflicting are emitted once, into the shared dictionary: types marked

       conflicting are emitted once per TU into a dictionary corresponding to

       each TU in which they appear.  Structs marked conflicting get at the very

       least a forward emitted into the shared dict so that other dicts can cite

       it if needed.

   [id_to_packed_id]

   This all works over an array of inputs (usually in the same order as the

   inputs on the link line).  We don't use the ctf_link_inputs hash directly

   because it is convenient to be able to address specific input types as a

   *global type ID* or 'GID', a pair of an array offset and a ctf_id_t.  Since

   both are already 32 bits or less or can easily be constrained to that range,

   we can pack them both into a single 64-bit hash word for easy lookups, which

   would be much more annoying to do with a ctf_dict_t * and a ctf_id_t.  (On

   32-bit platforms, we must do that anyway, since pointers, and thus hash keys

   and values, are only 32 bits wide).  We track which inputs are parents of

   which other inputs so that we can correctly recognize that types we have

   traversed in children may cite types in parents, and so that we can process

   the parents first.)

   Note that thanks to ld -r, the deduplicator can be fed its own output, so the

   inputs may themselves have child dicts.  Since we need to support this usage

   anyway, we can use it in one other place.  If the caller finds translation

   units to be too small a unit ambiguous types, links can be 'cu-mapped', where

   the caller provides a mapping of input TU names to output child dict names.

   This mapping can fuse many child TUs into one potential child dict, so that

   ambiguous types in any of those input TUs go into the same child dict.

   When a many:1 cu-mapping is detected, the ctf_dedup machinery is called

   repeatedly, once for every output name that has more than one input, to fuse

   all the input TUs associated with a given output dict into one, and once again

   as normal to deduplicate all those intermediate outputs (and any 1:1 inputs)

   together.  This has much higher memory usage than otherwise, because in the

   intermediate state, all the output TUs are in memory at once and cannot be

   lazily opened.  It also has implications for the emission code: if types

   appear ambiguously in multiple input TUs that are all mapped to the same

   child dict, we cannot put them in children in the cu-mapping link phase

   because this output is meant to *become* a child in the next link stage and

   parent/child relationships are only one level deep: so instead, we just hide

   all but one of the ambiguous types.

   There are a few other subtleties here that make this more complex than it

   seems.  Let's go over the steps above in more detail.

   1) HASHING.

   [ctf_dedup_hash_type, ctf_dedup_rhash_type]

   Hashing proceeds recursively, mixing in the properties of each input type

   (including its name, if any), and then adding the hash values of every type

   cited by that type.  The result is stashed in the cd_type_hashes so other

   phases can find the hash values of input types given their IDs, and so that

   if we encounter this type again while hashing we can just return its hash

   value: it is also stashed in the *output mapping*, a mapping from hash value

   to the set of GIDs corresponding to that type in all inputs.  We also keep

   track of the GID of the first appearance of the type in any input (in

   cd_output_first_gid), and the GID of structs, unions, and forwards that only

   appear in one TU (in cd_struct_origin).  See below for where these things are

   used.

   Everything in this phase is time-critical, because it is operating over

   non-deduplicated types and so may have hundreds or thousands of times the

   data volume to deal with than later phases.  Trace output is hidden behind

   ENABLE_LIBCTF_HASH_DEBUGGING to prevent the sheer number of calls to

   ctf_dprintf from slowing things down (tenfold slowdowns are observed purely

   from the calls to ctf_dprintf(), even with debugging switched off), and keep

   down the volume of output (hundreds of gigabytes of debug output are not

   uncommon on larger links).

   We have to do *something* about potential cycles in the type graph.  We'd

   like to avoid emitting forwards in the final output if possible, because

   forwards aren't much use: they have no members.  We are mostly saved from

   needing to worry about this at emission time by ctf_add_struct*()

   automatically replacing newly-created forwards when the real struct/union

   comes along.  So we only have to avoid getting stuck in cycles during the

   hashing phase, while also not confusing types that cite members that are

   structs with each other.  It is easiest to solve this problem by noting two

   things:

    - all cycles in C depend on the presence of tagged structs/unions

    - all tagged structs/unions have a unique name they can be disambiguated by

   [ctf_dedup_is_stub]

   This means that we can break all cycles by ceasing to hash in cited types at

   every tagged struct/union and instead hashing in a stub consisting of the

   struct/union's *decorated name*, which is the name preceded by "s " or "u "

   depending on the namespace (cached in cd_decorated_names).  Forwards are

   decorated identically (so a forward to "struct foo" would be represented as

   "s foo"): this means that a citation of a forward to a type and a citation of

   a concrete definition of a type with the same name ends up getting the same

   hash value.

   Of course, it is quite possible to have two TUs with structs with the same

   name and different definitions, but that's OK because when we scan for types

   with ambiguous names we will identify these and mark them conflicting.

   We populate one thing to help conflictedness marking.  No unconflicted type

   may cite a conflicted one, but this means that conflictedness marking must

   walk from types to the types that cite them, which is the opposite of the

   usual order.  We can make this easier to do by constructing a *citers* graph

   in cd_citers, which points from types to the types that cite them: because we

   emit forwards corresponding to every conflicted struct/union, we don't need

   to do this for citations of structs/unions by other types.  This is very

   convenient for us, because that's the only type we don't traverse

   recursively: so we can construct the citers graph at the same time as we

   hash, rather than needing to add an extra pass.  (This graph is a dynhash of

   *type hash values*, so it's small: in effect it is automatically

   deduplicated.)

   2) COLLISIONAL MARKING.

   [ctf_dedup_detect_name_ambiguity, ctf_dedup_mark_conflicting_hash]

   We identify types whose names collide during the hashing process, and count

   the rough number of uses of each name (caching may throw it off a bit: this

   doesn't need to be accurate).  We then mark the less-frequently-cited types

   with each names conflicting: the most-frequently-cited one goes into the

   shared type dictionary, while all others are duplicated into per-TU

   dictionaries, named after the input TU, that have the shared dictionary as a

   parent.  For structures and unions this is not quite good enough: we'd like

   to have citations of forwards to ambiguously named structures and unions

   *stay* as citations of forwards, so that the user can tell that the caller

   didn't actually know which structure definition was meant: but if we put one

   of those structures into the shared dictionary, it would supplant and replace

   the forward, leaving no sign.  So structures and unions do not take part in

   this popularity contest: if their names are ambiguous, they are just

   duplicated, and only a forward appears in the shared dict.

   [ctf_dedup_propagate_conflictedness]

   The process of marking types conflicted is itself recursive: we recursively

   traverse the cd_citers graph populated in the hashing pass above and mark

   everything that we encounter conflicted (without wasting time re-marking

   anything that is already marked).  This naturally terminates just where we

   want it to (at types that are cited by no other types, and at structures and

   unions) and suffices to ensure that types that cite conflicted types are

   always marked conflicted.

   [ctf_dedup_conflictify_unshared, ctf_dedup_multiple_input_dicts]

   When linking in CTF_LINK_SHARE_DUPLICATED mode, we would like all types that

   are used in only one TU to end up in a per-CU dict. The easiest way to do

   that is to mark them conflicted.  ctf_dedup_conflictify_unshared does this,

   traversing the output mapping and using ctf_dedup_multiple_input_dicts to

   check the number of input dicts each distinct type hash value came from:

   types that only came from one get marked conflicted.  One caveat here is that

   we need to consider both structs and forwards to them: a struct that appears

   in one TU and has a dozen citations to an opaque forward in other TUs should

   *not* be considered to be used in only one TU, because users would find it

   useful to be able to traverse into opaque structures of that sort: so we use

   cd_struct_origin to check both structs/unions and the forwards corresponding

   to them.

   3) EMISSION.

   [ctf_dedup_walk_output_mapping, ctf_dedup_rwalk_output_mapping,

    ctf_dedup_rwalk_one_output_mapping]

   Emission involves another walk of the entire output mapping, this time

   traversing everything other than struct members, recursively.  Types are

   emitted from leaves to trunk, emitting all types a type cites before emitting

   the type itself.  We sort the output mapping before traversing it, for

   reproducibility and also correctness: the input dicts may have parent/child

   relationships, so we simply sort all types that first appear in parents

   before all children, then sort types that first appear in dicts appearing

   earlier on the linker command line before those that appear later, then sort

   by input ctf_id_t.  (This is where we use cd_output_first_gid, collected

   above.)

   The walking is done using a recursive traverser which arranges to not revisit

   any type already visited and to call its callback once per input GID for

   input GIDs corresponding to conflicted output types.  The traverser only

   finds input types and calls a callback for them as many times as the output

   needs to appear: it doesn't try to figure out anything about where the output

   might go.  That's done by the callback based on whether the type is

   marked conflicted or not.

   [ctf_dedup_emit_type, ctf_dedup_id_to_target, ctf_dedup_synthesize_forward]

   ctf_dedup_emit_type is the (sole) callback for ctf_dedup_walk_output_mapping.

   Conflicted types have all necessary dictionaries created, and then we emit

   the type into each dictionary in turn, working over each input CTF type

   corresponding to each hash value and using ctf_dedup_id_to_target to map each

   input ctf_id_t into the corresponding type in the output (dealing with input

   ctf_id_t's with parents in the process by simply chasing to the parent dict

   if the type we're looking up is in there).  Emitting structures involves

   simply noting that the members of this structure need emission later on:

   because you cannot cite a single structure member from another type, we avoid

   emitting the members at this stage to keep recursion depths down a bit.

   At this point, if we have by some mischance decided that two different types

   with child types that hash to different values have in fact got the same hash

   value themselves and *not* marked it conflicting, the type walk will walk

   only *one* of them and in all likelihood we'll find that we are trying to

   emit a type into some child dictionary that references a type that was never

   emitted into that dictionary and assertion-fail.  This always indicates a bug

   in the conflictedness marking machinery or the hashing code, or both.

   ctf_dedup_id_to_target calls ctf_dedup_synthesize_forward to do one extra

   thing, alluded to above: if this is a conflicted tagged structure or union,

   and the target is the shared dict (i.e., the type we're being asked to emit

   is not itself conflicted so can't just point straight at the conflicted

   type), we instead synthesise a forward with the same name, emit it into the

   shared dict, record it in cd_output_emission_conflicted_forwards so that we

   don't re-emit it, and return it.  This means that cycles that contain

   conflicts do not cause the entire cycle to be replicated in every child: only

   that piece of the cycle which takes you back as far as the closest tagged

   struct/union needs to be replicated.  This trick means that no part of the

   deduplicator needs a cycle detector: every recursive walk can stop at tagged

   structures.

   [ctf_dedup_emit_struct_members]

   The final stage of emission is to walk over all structures with members

   that need emission and emit all of them. Every type has been emitted at

   this stage, so emission cannot fail.

   [ctf_dedup_populate_type_mappings, ctf_dedup_populate_type_mapping]

   Finally, we update the input -> output type ID mappings used by the ctf-link

   machinery to update all the other sections.  This is surprisingly expensive

   and may be replaced with a scheme which lets the ctf-link machinery extract

   the needed info directly from the deduplicator.  */

/* Possible future optimizations are flagged with 'optimization opportunity'

   below.  */

/* Global optimization opportunity: a GC pass, eliminating types with no direct

   or indirect citations from the other sections in the dictionary.  */

/* Internal flag values for ctf_dedup_hash_type.  */

/* Child call: consider forwardable types equivalent to forwards or stubs below

   this point.  */

#define CTF_DEDUP_HASH_INTERNAL_CHILD         0x01

/* Transform references to single ctf_id_ts in passed-in inputs into a number

   that will fit in a uint64_t.  Needs rethinking if CTF_MAX_TYPE is boosted.

   On 32-bit platforms, we pack things together differently: see the note

   above.  */

#if UINTPTR_MAX < UINT64_MAX

# define IDS_NEED_ALLOCATION 1

# define CTF_DEDUP_GID(fp, input, type) id_to_packed_id (fp, input, type)

# define CTF_DEDUP_GID_TO_INPUT(id) packed_id_to_input (id)

# define CTF_DEDUP_GID_TO_TYPE(id) packed_id_to_type (id)

#else

# define CTF_DEDUP_GID(fp, input, type) \

  (void *) (((uint64_t) input) << 32 | (type))

# define CTF_DEDUP_GID_TO_INPUT(id) ((int) (((uint64_t) id) >> 32))

# define CTF_DEDUP_GID_TO_TYPE(id) (ctf_id_t) (((uint64_t) id) & ~(0xffffffff00000000ULL))

#endif

#ifdef IDS_NEED_ALLOCATION

 /* This is the 32-bit path, which stores GIDs in a pool and returns a pointer

    into the pool.  It is notably less efficient than the 64-bit direct storage

    approach, but with a smaller key, this is all we can do.  */

static void *

id_to_packed_id (ctf_dict_t *fp, int input_num, ctf_id_t type)

{

  const void *lookup;

  ctf_type_id_key_t *dynkey = NULL;

  ctf_type_id_key_t key = { input_num, type };

  if (!ctf_dynhash_lookup_kv (fp->ctf_dedup.cd_id_to_dict_t,

            &key, &lookup, NULL))

    {

      if ((dynkey = malloc (sizeof (ctf_type_id_key_t))) == NULL)

  goto oom;

      memcpy (dynkey, &key, sizeof (ctf_type_id_key_t));

      if (ctf_dynhash_insert (fp->ctf_dedup.cd_id_to_dict_t, dynkey, NULL) < 0)

  goto oom;

      ctf_dynhash_lookup_kv (fp->ctf_dedup.cd_id_to_dict_t,

           dynkey, &lookup, NULL);

    }

  /* We use a raw assert() here because there isn't really a way to get any sort

     of error back from this routine without vastly complicating things for the

     much more common case of !IDS_NEED_ALLOCATION.  */

  assert (lookup);

  return (void *) lookup;

 oom:

  free (dynkey);

  ctf_set_errno (fp, ENOMEM);

  return NULL;

}

static int

packed_id_to_input (const void *id)

{

  const ctf_type_id_key_t *key = (ctf_type_id_key_t *) id;

  return key->ctii_input_num;

}

static ctf_id_t

packed_id_to_type (const void *id)

{

  const ctf_type_id_key_t *key = (ctf_type_id_key_t *) id;

  return key->ctii_type;

}

#endif

/* Make an element in a dynhash-of-dynsets, or return it if already present.  */

static ctf_dynset_t *

make_set_element (ctf_dynhash_t *set, const void *key)

{

  ctf_dynset_t *element;

  if ((element = ctf_dynhash_lookup (set, key)) == NULL)

    {

      if ((element = ctf_dynset_create (htab_hash_string,

          htab_eq_string,

          NULL)) == NULL)

  return NULL;

      if (ctf_dynhash_insert (set, (void *) key, element) < 0)

  {

    ctf_dynset_destroy (element);

    return NULL;

  }

    }

  return element;

}

/* Initialize the dedup atoms table.  */

int

ctf_dedup_atoms_init (ctf_dict_t *fp)

{

  if (fp->ctf_dedup_atoms)

    return 0;

  if (!fp->ctf_dedup_atoms_alloc)

    {

      if ((fp->ctf_dedup_atoms_alloc

     = ctf_dynset_create (htab_hash_string, htab_eq_string,

        free)) == NULL)

  return ctf_set_errno (fp, ENOMEM);

    }

  fp->ctf_dedup_atoms = fp->ctf_dedup_atoms_alloc;

  return 0;

}

/* Intern things in the dedup atoms table.  */

static const char *

intern (ctf_dict_t *fp, char *atom)

{

  const void *foo;

  if (atom == NULL)

    return NULL;

  if (!ctf_dynset_exists (fp->ctf_dedup_atoms, atom, &foo))

    {

      if (ctf_dynset_insert (fp->ctf_dedup_atoms, atom) < 0)

  {

    ctf_set_errno (fp, ENOMEM);

    return NULL;

  }

      foo = atom;

    }

  else

    free (atom);

  return (const char *) foo;

}

/* Add an indication of the namespace to a type name in a way that is not valid

   for C identifiers.  Used to maintain hashes of type names to other things

   while allowing for the four C namespaces (normal, struct, union, enum).

   Return a pointer into the cd_decorated_names atoms table.  */

static const char *

ctf_decorate_type_name (ctf_dict_t *fp, const char *name, int kind)

{

  ctf_dedup_t *d = &fp->ctf_dedup;

  const char *ret;

  const char *k;

  char *p;

  size_t i;

  switch (kind)

    {

    case CTF_K_STRUCT:

      k = "s ";

      i = 0;

      break;

    case CTF_K_UNION:

      k = "u ";

      i = 1;

      break;

    case CTF_K_ENUM:

      k = "e ";

      i = 2;

      break;

    default:

      k = "";

      i = 3;

    }

  if ((ret = ctf_dynhash_lookup (d->cd_decorated_names[i], name)) == NULL)

    {

      char *str;

      if ((str = malloc (strlen (name) + strlen (k) + 1)) == NULL)

  goto oom;

      p = stpcpy (str, k);

      strcpy (p, name);

      ret = intern (fp, str);

      if (!ret)

  goto oom;

      if (ctf_dynhash_cinsert (d->cd_decorated_names[i], name, ret) < 0)

  goto oom;

    }

  return ret;

 oom:

  ctf_set_errno (fp, ENOMEM);

  return NULL;

}

/* Hash a type, possibly debugging-dumping something about it as well.  */

static inline void

ctf_dedup_sha1_add (ctf_sha1_t *sha1, const void *buf, size_t len,

        const char *description _libctf_unused_,

        unsigned long depth _libctf_unused_)

{

  ctf_sha1_add (sha1, buf, len);

#ifdef ENABLE_LIBCTF_HASH_DEBUGGING

  ctf_sha1_t tmp;

  char tmp_hval[CTF_SHA1_SIZE];

  tmp = *sha1;

  ctf_sha1_fini (&tmp, tmp_hval);

  ctf_dprintf ("%lu: after hash addition of %s: %s\n", depth, description,

         tmp_hval);

#endif

}

static const char *

ctf_dedup_hash_type (ctf_dict_t *fp, ctf_dict_t *input,

         ctf_dict_t **inputs, uint32_t *parents,

         int input_num, ctf_id_t type, int flags,

         unsigned long depth,

         int (*populate_fun) (ctf_dict_t *fp,

            ctf_dict_t *input,

            ctf_dict_t **inputs,

            int input_num,

            ctf_id_t type,

            void *id,

            const char *decorated_name,

            const char *hash));

/* Determine whether this type is being hashed as a stub (in which case it is

   unsafe to cache it).  */

static int

ctf_dedup_is_stub (const char *name, int kind, int fwdkind, int flags)

{

  /* We can cache all types unless we are recursing to children and are hashing

     in a tagged struct, union or forward, all of which are replaced with their

     decorated name as a stub and will have different hash values when hashed at

     the top level.  */

  return ((flags & CTF_DEDUP_HASH_INTERNAL_CHILD) && name

    && (kind == CTF_K_STRUCT || kind == CTF_K_UNION

        || (kind == CTF_K_FORWARD && (fwdkind == CTF_K_STRUCT

              || fwdkind == CTF_K_UNION))));

}

/* Populate struct_origin if need be (not already populated, or populated with

   a different origin), in which case it must go to -1, "shared".)

   Only called for forwards or forwardable types with names, when the link mode

   is CTF_LINK_SHARE_DUPLICATED.  */

static int

ctf_dedup_record_origin (ctf_dict_t *fp, int input_num, const char *decorated,

       void *id)

{

  ctf_dedup_t *d = &fp->ctf_dedup;

  void *origin;

  int populate_origin = 0;

  if (ctf_dynhash_lookup_kv (d->cd_struct_origin, decorated, NULL, &origin))

    {

      if (CTF_DEDUP_GID_TO_INPUT (origin) != input_num

    && CTF_DEDUP_GID_TO_INPUT (origin) != -1)

  {

    populate_origin = 1;

    origin = CTF_DEDUP_GID (fp, -1, -1);

  }

    }

  else

    {

      populate_origin = 1;

      origin = id;

    }

  if (populate_origin)

    if (ctf_dynhash_cinsert (d->cd_struct_origin, decorated, origin) < 0)

      return ctf_set_errno (fp, errno);

  return 0;

}

/* Do the underlying hashing and recursion for ctf_dedup_hash_type (which it

   calls, recursively).  */

static const char *

ctf_dedup_rhash_type (ctf_dict_t *fp, ctf_dict_t *input, ctf_dict_t **inputs,

          uint32_t *parents, int input_num, ctf_id_t type,

          void *type_id, const ctf_type_t *tp, const char *name,

          const char *decorated, int kind, int flags,

          unsigned long depth,

          int (*populate_fun) (ctf_dict_t *fp,

             ctf_dict_t *input,

             ctf_dict_t **inputs,

             int input_num,

             ctf_id_t type,

             void *id,

             const char *decorated_name,

             const char *hash))

{

  ctf_dedup_t *d = &fp->ctf_dedup;

  ctf_next_t *i = NULL;

  ctf_sha1_t hash;

  ctf_id_t child_type;

  char hashbuf[CTF_SHA1_SIZE];

  const char *hval = NULL;

  const char *whaterr;

  int err = 0;

  const char *citer = NULL;

  ctf_dynset_t *citers = NULL;

  /* Add a citer to the citers set.  */

#define ADD_CITER(citers, hval)           \

  do                  \

    {                 \

      whaterr = N_("error updating citers");        \

      if (!citers)             \

  if ((citers = ctf_dynset_create (htab_hash_string,    \

           htab_eq_string,    \

           NULL)) == NULL)   \

    goto oom;             \

      if (ctf_dynset_cinsert (citers, hval) < 0)     \

  goto oom;             \

    }                 \

  while (0)

  /* If this is a named struct or union or a forward to one, and this is a child

     traversal, treat this type as if it were a forward -- do not recurse to

     children, ignore all content not already hashed in, and hash in the

     decorated name of the type instead.  */

  if (ctf_dedup_is_stub (name, kind, tp->ctt_type, flags))

    {

#ifdef ENABLE_LIBCTF_HASH_DEBUGGING

      ctf_dprintf ("Struct/union/forward citation: substituting forwarding "

       "stub with decorated name %s\n", decorated);

#endif

      ctf_sha1_init (&hash);

      ctf_dedup_sha1_add (&hash, decorated, strlen (decorated) + 1,

        "decorated struct/union/forward name", depth);

      ctf_sha1_fini (&hash, hashbuf);

      if ((hval = intern (fp, strdup (hashbuf))) == NULL)

  {

    ctf_err_warn (fp, 0, 0, _("%s (%i): out of memory during forwarding-"

            "stub hashing for type with GID %p"),

      ctf_link_input_name (input), input_num, type_id);

    return NULL;        /* errno is set for us.  */

  }

      /* In share-duplicated link mode, make sure the origin of this type is

   recorded, even if this is a type in a parent dict which will not be

   directly traversed.  */

      if (d->cd_link_flags & CTF_LINK_SHARE_DUPLICATED

    && ctf_dedup_record_origin (fp, input_num, decorated, type_id) < 0)

  return NULL;       /* errno is set for us.  */

      return hval;

    }

  /* Now ensure that subsequent recursive calls (but *not* the top-level call)

     get this treatment.  */

  flags |= CTF_DEDUP_HASH_INTERNAL_CHILD;

  /* If this is a struct, union, or forward with a name, record the unique

     originating input TU, if there is one.  */

  if (decorated && (ctf_forwardable_kind (kind) || kind != CTF_K_FORWARD))

    if (d->cd_link_flags & CTF_LINK_SHARE_DUPLICATED

  && ctf_dedup_record_origin (fp, input_num, decorated, type_id) < 0)

      return NULL;       /* errno is set for us.  */

#ifdef ENABLE_LIBCTF_HASH_DEBUGGING

  ctf_dprintf ("%lu: hashing thing with ID %i/%lx (kind %i): %s.\n",

         depth, input_num, type, kind, name ? name : "");

#endif

  /* Some type kinds don't have names: the API provides no way to set the name,

     so the type the deduplicator outputs will be nameless even if the input

     somehow has a name, and the name should not be mixed into the hash.  */

  switch (kind)

    {

    case CTF_K_POINTER:

    case CTF_K_ARRAY:

    case CTF_K_FUNCTION:

    case CTF_K_VOLATILE:

    case CTF_K_CONST:

    case CTF_K_RESTRICT:

    case CTF_K_SLICE:

      name = NULL;

    }

  /* Mix in invariant stuff, transforming the type kind if needed.  Note that

     the vlen is *not* hashed in: the actual variable-length info is hashed in

     instead, piecewise.  The vlen is not part of the type, only the

     variable-length data is: identical types with distinct vlens are quite

     possible.  Equally, we do not want to hash in the isroot flag: both the

     compiler and the deduplicator set the nonroot flag to indicate clashes with

     *other types in the same TU* with the same name: so two types can easily

     have distinct nonroot flags, yet be exactly the same type.*/

  ctf_sha1_init (&hash);

  if (name)

    ctf_dedup_sha1_add (&hash, name, strlen (name) + 1, "name", depth);

  ctf_dedup_sha1_add (&hash, &kind, sizeof (uint32_t), "kind", depth);

  /* Hash content of this type.  */

  switch (kind)

    {

    case CTF_K_UNKNOWN:

      /* No extra state.  */

      break;

    case CTF_K_FORWARD:

      /* Add the forwarded kind, stored in the ctt_type.  */

      ctf_dedup_sha1_add (&hash, &tp->ctt_type, sizeof (tp->ctt_type),

        "forwarded kind", depth);

      break;

    case CTF_K_INTEGER:

    case CTF_K_FLOAT:

      {

  ctf_encoding_t ep;

  memset (&ep, 0, sizeof (ctf_encoding_t));

  ctf_dedup_sha1_add (&hash, &tp->ctt_size, sizeof (uint32_t), "size",

          depth);

  if (ctf_type_encoding (input, type, &ep) < 0)

    {

      whaterr = N_("error getting encoding");

      goto input_err;

    }

  ctf_dedup_sha1_add (&hash, &ep, sizeof (ctf_encoding_t), "encoding",

          depth);

  break;

      }

      /* Types that reference other types.  */

    case CTF_K_TYPEDEF:

    case CTF_K_VOLATILE:

    case CTF_K_CONST:

    case CTF_K_RESTRICT:

    case CTF_K_POINTER:

      /* Hash the referenced type, if not already hashed, and mix it in.  */

      child_type = ctf_type_reference (input, type);

      if ((hval = ctf_dedup_hash_type (fp, input, inputs, parents, input_num,

               child_type, flags, depth,

               populate_fun)) == NULL)

  {

    whaterr = N_("error doing referenced type hashing");

    goto err;

  }

      ctf_dedup_sha1_add (&hash, hval, strlen (hval) + 1, "referenced type",

        depth);

      citer = hval;

      break;

      /* The slices of two types hash identically only if the type they overlay

   also has the same encoding.  This is not ideal, but in practice will work

   well enough.  We work directly rather than using the CTF API because

   we do not want the slice's normal automatically-shine-through

   semantics to kick in here.  */

    case CTF_K_SLICE:

      {

  const ctf_slice_t *slice;

  const ctf_dtdef_t *dtd;

  ssize_t size;

  ssize_t increment;

  child_type = ctf_type_reference (input, type);

  ctf_get_ctt_size (input, tp, &size, &increment);

  ctf_dedup_sha1_add (&hash, &size, sizeof (ssize_t), "size", depth);

  if ((hval = ctf_dedup_hash_type (fp, input, inputs, parents, input_num,

           child_type, flags, depth,

           populate_fun)) == NULL)

    {

      whaterr = N_("error doing slice-referenced type hashing");

      goto err;

    }

  ctf_dedup_sha1_add (&hash, hval, strlen (hval) + 1, "sliced type",

          depth);

  citer = hval;

  if ((dtd = ctf_dynamic_type (input, type)) != NULL)

    slice = (ctf_slice_t *) dtd->dtd_vlen;

  else

    slice = (ctf_slice_t *) ((uintptr_t) tp + increment);

  ctf_dedup_sha1_add (&hash, &slice->cts_offset,

          sizeof (slice->cts_offset), "slice offset", depth);

  ctf_dedup_sha1_add (&hash, &slice->cts_bits,

          sizeof (slice->cts_bits), "slice bits", depth);

  break;

      }

    case CTF_K_ARRAY:

      {

  ctf_arinfo_t ar;

  if (ctf_array_info (input, type, &ar) < 0)

    {

      whaterr = N_("error getting array info");

      goto input_err;

    }

  if ((hval = ctf_dedup_hash_type (fp, input, inputs, parents, input_num,

           ar.ctr_contents, flags, depth,

           populate_fun)) == NULL)

    {

      whaterr = N_("error doing array contents type hashing");

      goto err;

    }

  ctf_dedup_sha1_add (&hash, hval, strlen (hval) + 1, "array contents",

          depth);

  ADD_CITER (citers, hval);

  if ((hval = ctf_dedup_hash_type (fp, input, inputs, parents, input_num,

           ar.ctr_index, flags, depth,

           populate_fun)) == NULL)

    {

      whaterr = N_("error doing array index type hashing");

      goto err;

    }

  ctf_dedup_sha1_add (&hash, hval, strlen (hval) + 1, "array index",

          depth);

  ctf_dedup_sha1_add (&hash, &ar.ctr_nelems, sizeof (ar.ctr_nelems),

          "element count", depth);

  ADD_CITER (citers, hval);

  break;

      }

    case CTF_K_FUNCTION:

      {

  ctf_funcinfo_t fi;

  ctf_id_t *args;

  uint32_t j;

  if (ctf_func_type_info (input, type, &fi) < 0)

    {

      whaterr = N_("error getting func type info");

      goto input_err;

    }

  if ((hval = ctf_dedup_hash_type (fp, input, inputs, parents, input_num,

           fi.ctc_return, flags, depth,

           populate_fun)) == NULL)

    {

      whaterr = N_("error getting func return type");

      goto err;

    }

  ctf_dedup_sha1_add (&hash, hval, strlen (hval) + 1, "func return",

          depth);

  ctf_dedup_sha1_add (&hash, &fi.ctc_argc, sizeof (fi.ctc_argc),

          "func argc", depth);

  ctf_dedup_sha1_add (&hash, &fi.ctc_flags, sizeof (fi.ctc_flags),

          "func flags", depth);

  ADD_CITER (citers, hval);

  if ((args = calloc (fi.ctc_argc, sizeof (ctf_id_t))) == NULL)

    {

      err = ENOMEM;

      whaterr = N_("error doing memory allocation");

      goto err;

    }

  if (ctf_func_type_args (input, type, fi.ctc_argc, args) < 0)

    {

      free (args);

      whaterr = N_("error getting func arg type");

      goto input_err;

    }

  for (j = 0; j < fi.ctc_argc; j++)

    {

      if ((hval = ctf_dedup_hash_type (fp, input, inputs, parents,

               input_num, args[j], flags, depth,

               populate_fun)) == NULL)

        {

    free (args);

    whaterr = N_("error doing func arg type hashing");

    goto err;

        }

      ctf_dedup_sha1_add (&hash, hval, strlen (hval) + 1, "func arg type",

        depth);

      ADD_CITER (citers, hval);

    }

  free (args);

  break;

      }

    case CTF_K_ENUM:

      {

  int val;

  const char *ename;

  ctf_dedup_sha1_add (&hash, &tp->ctt_size, sizeof (uint32_t),

          "enum size", depth);

  while ((ename = ctf_enum_next (input, type, &i, &val)) != NULL)

    {

      ctf_dedup_sha1_add (&hash, ename, strlen (ename) + 1, "enumerator",

        depth);

      ctf_dedup_sha1_add (&hash, &val, sizeof (val), "enumerand", depth);

    }

  if (ctf_errno (input) != ECTF_NEXT_END)

    {

      whaterr = N_("error doing enum member iteration");

      goto input_err;

    }

  break;

      }

    /* Top-level only.  */

    case CTF_K_STRUCT:

    case CTF_K_UNION:

      {

  ssize_t offset;

  const char *mname;

  ctf_id_t membtype;

  ssize_t size;

  ctf_get_ctt_size (input, tp, &size, NULL);

  ctf_dedup_sha1_add (&hash, &size, sizeof (ssize_t), "struct size",

          depth);

  while ((offset = ctf_member_next (input, type, &i, &mname, &membtype,

            0)) >= 0)

    {

      if (mname == NULL)

        mname = "";

      ctf_dedup_sha1_add (&hash, mname, strlen (mname) + 1,

        "member name", depth);

#ifdef ENABLE_LIBCTF_HASH_DEBUGGING

      ctf_dprintf ("%lu: Traversing to member %s\n", depth, mname);

#endif

      if ((hval = ctf_dedup_hash_type (fp, input, inputs, parents,

               input_num, membtype, flags, depth,

               populate_fun)) == NULL)

        {

    whaterr = N_("error doing struct/union member type hashing");

    goto iterr;

        }

      ctf_dedup_sha1_add (&hash, hval, strlen (hval) + 1, "member hash",

        depth);

      ctf_dedup_sha1_add (&hash, &offset, sizeof (offset), "member offset",

        depth);

      ADD_CITER (citers, hval);

    }

  if (ctf_errno (input) != ECTF_NEXT_END)

    {

      whaterr = N_("error doing struct/union member iteration");

      goto input_err;

    }

  break;

      }

    default:

      whaterr = N_("error: unknown type kind");

      goto err;

    }

  ctf_sha1_fini (&hash, hashbuf);

  if ((hval = intern (fp, strdup (hashbuf))) == NULL)

    {

      whaterr = N_("cannot intern hash");

      goto oom;

    }

  /* Populate the citers for this type's subtypes, now the hash for the type

     itself is known.  */

  whaterr = N_("error tracking citers");

  if (citer)

    {

      ctf_dynset_t *citer_hashes;

      if ((citer_hashes = make_set_element (d->cd_citers, citer)) == NULL)

  goto oom;

      if (ctf_dynset_cinsert (citer_hashes, hval) < 0)

  goto oom;

    }

  else if (citers)

    {

      const void *k;

      while ((err = ctf_dynset_cnext (citers, &i, &k)) == 0)

  {

    ctf_dynset_t *citer_hashes;

    citer = (const char *) k;

    if ((citer_hashes = make_set_element (d->cd_citers, citer)) == NULL)

      goto oom;