asn1: Circular types and Topo. sort declarations

Many external ASN.1 modules that we have imported over time define types
like this:

  Foo ::= SEQUENCE { bar Bar }
  Bar ::= SEQUENCE { aMember INTEGER }

and before this change one had to re-order the definitions so that the
one for `Bar` came first.  No more.

We can now have out of order definitions in ASN.1 modules and the
compiler will topologically sort output C type declarations so that one
no longer has to manually sort types in ASN.1 modules when importing
them.

Besides that, it is now possible to create circular data types using
OPTIONAL since we generate such fields as pointers (which can then be
pointers to incomplete struct declarations):

  Circular ::= SEQUENCE {
          name UTF8String,
          next Circular OPTIONAL
  }

Circular types aren't necessarily useful, but they have been used in the
past.  E.g., the rpc.mountd protocol uses a circular type as a linked
list -- it should just have used an array, of course, as that's
semantically equivalent but more space efficient in its encoding, but
the point is that such types exist out there.
This commit is contained in:
Nicolas Williams
2020-12-24 03:18:28 -06:00
parent 5b978c7437
commit 83d4c6ddb5
8 changed files with 367 additions and 30 deletions

View File

@@ -240,7 +240,6 @@ static int default_tag_env = TE_EXPLICIT;
ModuleDefinition: IDENTIFIER objid_opt kw_DEFINITIONS TagDefault ExtensionDefault
EEQUAL kw_BEGIN ModuleBody kw_END
{
checkundefined();
}
;
@@ -320,11 +319,14 @@ referencenames : IDENTIFIER ',' referencenames
TypeAssignment : IDENTIFIER EEQUAL Type
{
Symbol *s = addsym ($1);
Symbol *s = addsym($1);
s->stype = Stype;
s->type = $3;
fix_labels(s);
generate_type (s);
if (original_order)
generate_type(s);
else
generate_type_header_forwards(s);
}
;

View File

@@ -752,9 +752,12 @@ define_type (int level, const char *name, const char *basename, Type *t, int typ
switch (t->type) {
case TType:
space(level);
fprintf (headerfile, "%s %s;\n", t->symbol->gen_name, name);
fprintf(headerfile, "%s %s;\n", t->symbol->gen_name, name);
break;
case TInteger:
if (t->symbol && t->symbol->emitted_definition)
break;
space(level);
if(t->members) {
Member *m;
@@ -768,11 +771,11 @@ define_type (int level, const char *name, const char *basename, Type *t, int typ
label_prefix, label_prefix_sep,
m->gen_name, m->val, last_member_p(m));
}
fprintf (headerfile, "} %s;\n", name);
fprintf(headerfile, "} %s;\n", name);
} else if (t->range == NULL) {
fprintf (headerfile, "heim_integer %s;\n", name);
fprintf(headerfile, "heim_integer %s;\n", name);
} else if (t->range->min < INT_MIN && t->range->max <= INT64_MAX) {
fprintf (headerfile, "int64_t %s;\n", name);
fprintf(headerfile, "int64_t %s;\n", name);
} else if (t->range->min >= 0 && t->range->max > UINT_MAX) {
fprintf (headerfile, "uint64_t %s;\n", name);
} else if (t->range->min >= INT_MIN && t->range->max <= INT_MAX) {
@@ -798,6 +801,10 @@ define_type (int level, const char *name, const char *basename, Type *t, int typ
size_t max_memno = 0;
size_t bitset_size;
if (t->symbol && t->symbol->emitted_definition)
break;
memset(&i, 0, sizeof(i));
/*
* range.max implies the size of the base unsigned integer used for the
* bitfield members. If it's less than or equal to UINT_MAX, then that
@@ -827,7 +834,7 @@ define_type (int level, const char *name, const char *basename, Type *t, int typ
fprintf (headerfile, "heim_bit_string %s;\n", name);
else {
int pos = 0;
getnewbasename(&newbasename, typedefp, basename, name);
getnewbasename(&newbasename, typedefp || level == 0, basename, name);
fprintf (headerfile, "struct %s {\n", newbasename);
HEIM_TAILQ_FOREACH(m, t->members, members) {
@@ -876,6 +883,9 @@ define_type (int level, const char *name, const char *basename, Type *t, int typ
case TEnumerated: {
Member *m;
if (t->symbol && t->symbol->emitted_definition)
break;
label_prefix = prefix_enum ? name : (enum_prefix ? enum_prefix : "");
label_prefix_sep = prefix_enum ? "_" : "";
space(level);
@@ -897,7 +907,7 @@ define_type (int level, const char *name, const char *basename, Type *t, int typ
case TSequence: {
Member *m;
getnewbasename(&newbasename, typedefp, basename, name);
getnewbasename(&newbasename, typedefp || level == 0, basename, name);
space(level);
fprintf (headerfile, "struct %s {\n", newbasename);
@@ -927,12 +937,11 @@ define_type (int level, const char *name, const char *basename, Type *t, int typ
Type i;
struct range range = { 0, UINT_MAX };
getnewbasename(&newbasename, typedefp, basename, name);
getnewbasename(&newbasename, typedefp || level == 0, basename, name);
memset(&i, 0, sizeof(i));
i.type = TInteger;
i.range = &range;
i.members = NULL;
i.constraint = NULL;
space(level);
fprintf (headerfile, "struct %s {\n", newbasename);
@@ -955,13 +964,13 @@ define_type (int level, const char *name, const char *basename, Type *t, int typ
fprintf (headerfile, "heim_general_string %s;\n", name);
break;
case TTag:
define_type (level, name, basename, t->subtype, typedefp, preservep);
define_type (level, name, basename, t->subtype, typedefp, preservep);
break;
case TChoice: {
int first = 1;
Member *m;
getnewbasename(&newbasename, typedefp, basename, name);
getnewbasename(&newbasename, typedefp || level == 0, basename, name);
space(level);
fprintf (headerfile, "struct %s {\n", newbasename);
@@ -1050,24 +1059,229 @@ define_type (int level, const char *name, const char *basename, Type *t, int typ
default:
abort ();
}
if (newbasename)
free(newbasename);
free(newbasename);
}
static void
declare_type(const Symbol *s, Type *t, int typedefp)
{
char *newbasename = NULL;
if (typedefp)
fprintf(headerfile, "typedef ");
switch (t->type) {
case TType:
define_type(0, s->gen_name, s->gen_name, s->type, TRUE, TRUE);
if (template_flag)
generate_template_type_forward(s->gen_name);
emitted_declaration(s);
return;
case TInteger:
case TBoolean:
case TOctetString:
case TBitString:
case TEnumerated:
case TGeneralizedTime:
case TGeneralString:
case TTeletexString:
case TUTCTime:
case TUTF8String:
case TPrintableString:
case TIA5String:
case TBMPString:
case TUniversalString:
case TVisibleString:
case TOID :
case TNull:
define_type(0, s->gen_name, s->gen_name, s->type, TRUE, TRUE);
if (template_flag)
generate_template_type_forward(s->gen_name);
emitted_declaration(s);
emitted_definition(s);
return;
case TTag:
declare_type(s, t->subtype, FALSE);
emitted_declaration(s);
return;
default:
break;
}
switch (t->type) {
case TSet:
case TSequence:
getnewbasename(&newbasename, TRUE, s->gen_name, s->gen_name);
fprintf(headerfile, "struct %s %s;\n", newbasename, s->gen_name);
break;
case TSetOf:
case TSequenceOf:
getnewbasename(&newbasename, TRUE, s->gen_name, s->gen_name);
fprintf(headerfile, "struct %s %s;\n", newbasename, s->gen_name);
break;
case TChoice:
getnewbasename(&newbasename, TRUE, s->gen_name, s->gen_name);
fprintf(headerfile, "struct %s %s;\n", newbasename, s->gen_name);
break;
default:
abort ();
}
free(newbasename);
emitted_declaration(s);
}
static void generate_subtypes_header_helper(const Member *m);
static void generate_type_header(const Symbol *);
static void
generate_subtypes_header_helper(const Member *m)
{
Member *sm;
Symbol *s;
if (m->ellipsis)
return;
if (m->type->symbol && (s = getsym(m->type->symbol->name)) &&
!s->emitted_definition) {
/* A field of some named type; recurse */
if (!m->optional && !m->defval)
generate_type_header(s);
return;
}
if (!m->type->subtype && !m->type->members)
return;
if (m->type->type == TTag &&
m->type->subtype && m->type->subtype->symbol &&
(s = getsym(m->type->subtype->symbol->name))) {
if (!m->optional && !m->defval)
generate_type_header(s);
return;
}
if (m->type->subtype) {
switch (m->type->subtype->type) {
case TSet:
case TSequence:
case TChoice:
break;
default:
return;
}
/* A field of some anonymous (inlined) structured type */
HEIM_TAILQ_FOREACH(sm, m->type->subtype->members, members) {
generate_subtypes_header_helper(sm);
}
}
if (m->type->members) {
HEIM_TAILQ_FOREACH(sm, m->type->members, members) {
generate_subtypes_header_helper(sm);
}
}
}
static void
generate_subtypes_header(const Symbol *s)
{
Type *t = s->type;
Member *m;
/*
* Recurse down structured types to make sure top-level types get
* defined before they are referenced.
*
* We'll take care to skip OPTIONAL member fields of constructed types so
* that we can have circular types like:
*
* Foo ::= SEQUENCE {
* bar Bar OPTIONAL
* }
*
* Bar ::= SEQUENCE {
* foo Foo OPTIONAL
* }
*
* not unlike XDR, which uses `*' to mean "optional", except in XDR it's
* called a "pointer". With some care we should be able to eventually
* support the silly XDR linked list example:
*
* ListOfFoo ::= SEQUENCE {
* someField SomeType,
* next ListOfFoo OPTIONAL
* }
*
* Not that anyone needs it -- just use a SEQUENCE OF and be done.
*/
while (t->type == TTag && t->subtype) {
switch (t->subtype->type) {
case TTag:
case TSet:
case TSequence:
case TChoice:
t = t->subtype;
continue;
default:
break;
}
break;
}
switch (t->type) {
default: return;
case TType:
if (t->symbol && (s = getsym(t->symbol->name)))
generate_type_header(s);
return;
case TSet:
case TSequence:
case TChoice:
break;
}
HEIM_TAILQ_FOREACH(m, t->members, members) {
generate_subtypes_header_helper(m);
}
}
static void
generate_type_header (const Symbol *s)
{
int preservep = preserve_type(s->name) ? TRUE : FALSE;
fprintf (headerfile, "/*\n");
fprintf (headerfile, "%s ::= ", s->name);
/*
* Recurse down the types of member fields of `s' to make sure that
* referenced types have had their definitions emitted already if the
* member fields are not OPTIONAL/DEFAULTed.
*/
if (s->type)
generate_subtypes_header(s);
if (!s->type)
return;
fprintf(headerfile, "/*\n");
fprintf(headerfile, "%s ::= ", s->name);
define_asn1 (0, s->type);
fprintf (headerfile, "\n*/\n\n");
fprintf(headerfile, "\n*/\n\n");
fprintf (headerfile, "typedef ");
define_type (0, s->gen_name, s->gen_name, s->type, TRUE, preservep);
if (s->emitted_definition)
return;
fprintf (headerfile, "\n");
fprintf(headerfile, "typedef ");
define_type(0, s->gen_name, s->gen_name, s->type, TRUE,
preserve_type(s->name) ? TRUE : FALSE);
fprintf(headerfile, "\n");
if (template_flag)
generate_template_type_forward(s->gen_name);
emitted_definition(s);
}
void
generate_type_header_forwards(const Symbol *s)
{
declare_type(s, s->type, TRUE);
fprintf(headerfile, "\n");
}
void

View File

@@ -103,13 +103,16 @@ typedef struct asn1_module {
/* CLI options and flags needed everywhere: */
getarg_strings preserve;
getarg_strings seq;
const char *enum_prefix;
unsigned int one_code_file:1;
unsigned int support_ber:1;
unsigned int parse_units_flag:1;
unsigned int prefix_enum:1; /* Should be a getarg_strings of bitrsting types to do this for */
unsigned int rfc1510_bitstring:1; /* Should be a getarg_strings of bitrsting types to do this for */
} *asn1_module;
void generate_type (const Symbol *);
void generate_type_header_forwards(const Symbol *);
void generate_constant (const Symbol *);
void generate_type_encode (const Symbol *);
void generate_type_decode (const Symbol *);
@@ -143,6 +146,7 @@ void close_codefile(void);
int is_template_compat (const Symbol *);
void generate_template(const Symbol *);
void generate_template_type_forward(const char *);
void gen_template_import(const Symbol *);
@@ -152,6 +156,7 @@ extern int support_ber;
extern int template_flag;
extern int rfc1510_bitstring;
extern int one_code_file;
extern int original_order;
extern int parse_units_flag;
extern char *type_file_string;

View File

@@ -879,6 +879,12 @@ gen_template_import(const Symbol *s)
gen_extern_stubs(f, s->gen_name);
}
void
generate_template_type_forward(const char *name)
{
fprintf(get_code_file(), "extern const struct asn1_template asn1_%s[];\n", name);
}
static void
generate_template_type(const char *varname,
const char **dupname,
@@ -900,8 +906,6 @@ generate_template_type(const char *varname,
if (type->type == TTag)
implicit = (type->tag.tagenv == TE_IMPLICIT);
fprintf(get_code_file(), "extern const struct asn1_template asn1_%s[];\n", tl->name);
template_members(&tl->template, basetype, name, type, optional, implicit, isstruct, need_offset);
/* if its a sequence or set type, check if there is a ellipsis */

View File

@@ -71,6 +71,7 @@ int one_code_file;
char *option_file;
int parse_units_flag = 1;
char *type_file_string = "krb5-types.h";
int original_order;
int version_flag;
int help_flag;
struct getargs args[] = {
@@ -88,7 +89,15 @@ struct getargs args[] = {
{ "preserve-binary", 0, arg_strings, &preserve, NULL, NULL },
{ "sequence", 0, arg_strings, &seq, NULL, NULL },
{ "one-code-file", 0, arg_flag, &one_code_file, NULL, NULL },
{ "option-file", 0, arg_string, &option_file, NULL, NULL },
{ "gen-name", 0, arg_string, &name,
"Name of generated module", "NAME" },
{ "option-file", 0, arg_string, &option_file,
"File with additional compiler CLI options", "FILE" },
{ "original-order", 0, arg_flag, &original_order,
"Define C types and functions in the order in which they appear in "
"the ASN.1 module instead of topologically sorting types. This "
"is useful for comparing output to earlier compiler versions.",
NULL },
{ "parse-units", 0, arg_negative_flag, &parse_units_flag, NULL, NULL },
{ "type-file", 0, arg_string, &type_file_string, NULL, NULL },
{ "version", 0, arg_flag, &version_flag, NULL, NULL },
@@ -116,11 +125,11 @@ main(int argc, char **argv)
int len = 0, i;
setprogname(argv[0]);
if(getarg(args, num_args, argc, argv, &optidx))
if (getarg(args, num_args, argc, argv, &optidx))
usage(1);
if(help_flag)
if (help_flag)
usage(0);
if(version_flag) {
if (version_flag) {
print_version(NULL);
exit(0);
}
@@ -204,7 +213,7 @@ main(int argc, char **argv)
}
init_generate (file, name);
init_generate(file, name);
if (one_code_file)
generate_header_of_codefile(name);
@@ -213,6 +222,8 @@ main(int argc, char **argv)
ret = yyparse ();
if(ret != 0 || error_flag != 0)
exit(1);
if (!original_order)
generate_types();
close_generate ();
if (argc != optidx)
fclose(yyin);

View File

@@ -33,9 +33,12 @@
#include "gen_locl.h"
#include "lex.h"
#include "lex.h"
static Hashtab *htab;
struct symhead symbols;
static int
cmp(void *a, void *b)
{
@@ -84,10 +87,28 @@ addsym(char *name)
output_name(s->gen_name);
s->stype = SUndefined;
hashtabadd(htab, s);
//HEIM_TAILQ_INSERT_TAIL(&symbols, s, symlist);
do {
if (((s)->symlist.tqe_next = (&symbols)->tqh_first) != NULL)
(&symbols)->tqh_first->symlist.tqe_prev = &(s)->symlist.tqe_next;
else
(&symbols)->tqh_last = &(s)->symlist.tqe_next;
(&symbols)->tqh_first = (s);
(s)->symlist.tqe_prev = &(&symbols)->tqh_first;
} while (0);
}
return s;
}
Symbol *
getsym(char *name)
{
Symbol key;
key.name = name;
return (Symbol *) hashtabsearch(htab, (void *) &key);
}
static int
checkfunc(void *ptr, void *arg)
{
@@ -106,3 +127,41 @@ checkundefined(void)
hashtabforeach(htab, checkfunc, &f);
return f;
}
#if 0
static int
generate_1type(void *ptr, void *arg)
{
Symbol *s = ptr;
if (s->stype == Stype && s->type)
generate_type(s);
return 0;
}
#endif
void
generate_types(void)
{
Symbol *s;
if (checkundefined())
errx(1, "Some types are undefined");
HEIM_TAILQ_FOREACH_REVERSE(s, &symbols, symhead, symlist) {
if (s->stype == Stype && s->type)
generate_type(s);
}
//hashtabforeach(htab, generate_1type, NULL);
}
void
emitted_declaration(const Symbol *s)
{
((Symbol *)(uintptr_t)s)->emitted_declaration = 1;
}
void
emitted_definition(const Symbol *s)
{
((Symbol *)(uintptr_t)s)->emitted_definition = 1;
}

View File

@@ -156,12 +156,27 @@ struct symbol {
enum { SUndefined, SValue, Stype } stype;
struct value *value;
Type *type;
HEIM_TAILQ_ENTRY(symbol) symlist;
unsigned int emitted_declaration:1;
unsigned int emitted_definition:1;
};
typedef struct symbol Symbol;
//HEIM_TAILQ_HEAD(symhead, symbol);
struct symhead {
struct symbol *tqh_first;
struct symbol **tqh_last;
};
extern struct symhead symbols;
void initsym (void);
Symbol *addsym (char *);
Symbol *getsym(char *name);
void output_name (char *);
int checkundefined(void);
void generate_types(void);
void emitted_declaration(const Symbol *);
void emitted_definition(const Symbol *);
#endif

View File

@@ -6,6 +6,33 @@ BEGIN
IMPORTS heim_any FROM heim;
-- Check that we handle out of order definitions.
-- The compiler should emit the definition of TESTOutOfOrderBar before that of
-- TESTOutOfOrderFoo.
TESTOutOfOrderFoo ::= SEQUENCE {
bar TESTOutOfOrderBar
}
TESTOutOfOrderBar ::= SEQUENCE {
aMember INTEGER
}
-- Check that we can handle rpc.mountd style "lists". This is unnecessarily
-- inefficient in its encoding, and there's no point to using this over
-- SEQUENCE OF (arrays), but it's neat that we can do this now that we can do
-- out of order definitions.
--
-- This could be useful if we ever extend asn1_compile to also handle XDR,
-- which we well might since XDR's syntax is a dual of a strict subset of
-- ASN.1, and since XDR the encoding is fairly straightforward.
--
-- Note that the `next' member has to be OPTIONAL or DEFAULTed for this to
-- work.
TESTCircular ::= SEQUENCE {
name UTF8String,
next TESTCircular OPTIONAL
}
TESTuint32 ::= INTEGER (0..4294967295)
TESTuint64 ::= INTEGER(0..9223372036854775807)
TESTint64 ::= INTEGER(-9223372036854775808..9223372036854775807)