Windows 32-bit builds were broken as the ASN1CALL calling convention
macro was not included in the output of gen_template.c, only in the
generated header.
vis.h is not built on platforms (such as macOS) that already have
that header, which resulted in prototypes for Heimdal-specific
vis.h extensions being absent.
Move those prototypes to a separate header, vis-extras.h, which
must be explicitly included in order to use the Heimdal extensions.
Status:
- And it works!
- We have an extensive test based on decoding a rich EK certficate.
This test exercises all of:
- decoding
- encoding with and without decoded open types
- copying of decoded values with decoded open types
- freeing of decoded values with decoded open types
Valgrind finds no memory errors.
- Added a manual page for the compiler.
- rfc2459.asn1 now has all three primary PKIX types that we care about
defined as in RFC5912, with IOS constraints and parameterization:
- `Extension` (embeds open type in an `OCTET STRING`)
- `OtherName` (embeds open type in an `ANY`-like type)
- `SingleAttribute` (embeds open type in an `ANY`-like type)
- `AttributeSet` (embeds open type in a `SET OF ANY`-like type)
All of these use OIDs as the open type type ID field, but integer
open type type ID fields are also supported (and needed, for
Kerberos).
That will cover every typed hole pattern in all our ASN.1 modules.
With this we'll be able to automatically and recursively decode
through all subject DN attributes even when the subject DN is a
directoryName SAN, and subjectDirectoryAttributes, and all
extensions, and all SANs, and all authorization-data elements, and
PA-data, and...
We're not really using `SingleAttribute` and `AttributeSet` yet
because various changes are needed in `lib/hx509` for that.
- `asn1_compile` builds and recognizes the subset of X.681/682/683 that
we need for, and now use in, rfc2459.asn1. It builds the necessary
AST, generates the correct C types, and generates templating for
object sets and open types!
- See READMEs for details.
- Codegen backend not tested; I won't make it implement automatic open
type handling, but it should at least not crash by substituting
`heim_any` for open types not embedded in `OCTET STRING`.
- We're _really_ starting to have problems with the ITU-T ASN.1
grammar and our version of it...
Type names have to start with upper-case, value names with
lower-case, but it's not enough to disambiguate.
The fact the we've allowed value and type names to violate their
respective start-with case rules is causing us trouble now that we're
adding grammar from X.681/682/683, and we're going to have to undo
that.
In preparation for that I'm capitalizing the `heim_any` and
`heim_any_set` types, and doing some additional cleanup, which
requires changes to other parts of Heimdal (all in this same commit
for now).
Problems we have because of this:
- We cannot IMPORT values into modules because we have no idea if a
symbol being imported refers to a value or a type because the only
clue we would have is the symbol's name, so we assume IMPORTed
symbols are for types.
This means we can't import OIDs, for example, which is super
annoying.
One thing we might be able to do here is mark imported symbols as
being of an undetermined-but-not-undefined type, then coerce the
symbol's type the first time it's used in a context where its type
is inferred as type, value, object, object set, or class. (Though
since we don't generate C symbols for objects or classes, we won't
be able to import them, especially since we need to know them at
compile time and cannot defer their handling to link- or
run-time.)
- The `NULL` type name, and the `NULL` value name now cause two
reduce/reduce conflicts via the `FieldSetting` production.
- Various shift/reduce conflicts involving `NULL` values in
non-top-level contexts (in constraints, for example).
- Currently I have a bug where to disambiguate the grammar I have a
CLASS_IDENTIFIER token that is all caps, while TYPE_IDENTIFIER must
start with a capital but not be all caps, but this breaks Kerberos
since all its types are all capitalized -- oof!
To fix this I made it so class names have to be all caps and
start with an underscore (ick).
TBD:
- Check all the XXX comments and address them
- Apply this treatment to Kerberos! Automatic handling of authz-data
sounds useful :)
- Apply this treatment to PKCS#10 (CSRs) and other ASN.1 modules too.
- Replace various bits of code in `lib/hx509/` with uses of this
feature.
- Add JER.
- Enhance `hxtool` and `asn1_print`.
Getting there!
This adds enough support for SET { ... } types to the template backend
to allow the X.690 sample test to be run with the template backend.
Limitations:
- For DER encoding, the SET { ... } members must be manually sorted by
the module author.
- Decoding of out-of-order encodings (BER) is not supported at this
time.
These shortcomings will be addressed later.
Note that because we don't parse IMPORTed modules at this time, we can't
sort SET { ... } members at compile time if any of them out an
outer-most tag that the compiler cannot see without learning to parse
IMPORTed modules.
The template compiler was applying IMPLICIT tags to CHOICE types. This
is very wrong, as the tag of a CHOICE's taken choice cannot be replaced
without making it impossible to figure out what the choice was. An
example of this is GeneralName's directoryName, which is an IMPLICIT-
tagged CHOICE.
Separately, the non-template compiler was requiring inlining of
IMPLICIT-tagged CHOICEs, which also happens in GeneralName's
directoryName case:
```
205 Name ::= CHOICE {
206 rdnSequence RDNSequence
207 }
...
287 GeneralName ::= CHOICE {
288 otherName [0] IMPLICIT -- OtherName --
SEQUENCE {
289 type-id OBJECT IDENTIFIER,
290 value [0] EXPLICIT heim_any
291 },
292 rfc822Name [1] IMPLICIT IA5String,
293 dNSName [2] IMPLICIT IA5String,
294 -- x400Address [3] IMPLICIT ORAddress,--
--->295 directoryName [4] IMPLICIT -- Name -- CHOICE
{
296 rdnSequence RDNSequence
297 },
298 -- ediPartyName [5] IMPLICIT EDIPartyName, --
299 uniformResourceIdentifier [6] IMPLICIT IA5String,
300 iPAddress [7] IMPLICIT OCTET STRING,
301 registeredID [8] IMPLICIT OBJECT IDENTIFIER
302 }
```
Anyways, that's fixed now, though changing that will require making
corresponding changes to `lib/hx509/`.
We're getting closer to parity between the two compilers. The template
compiler is still missing support for `SET { ... }` types. Speaking of
`SET { ... }`, the regular compiler generates code that uses `qsort()`
to sort the encoded values values of the members of such a set, but this
seems silly because the order of members is knowable at compile time, as
for DER and CER the order by the tags of the members, from lowest to
highest (see X.690, section 9.3 and X.680, section 8.6). As it happens
using `qsort()` on the encodings of the members works, but it would be
be better to sort in `lib/asn1/asn1parse.y` and then not have to bother
anywhere else. Sorting SETs at definition time will help keep the
tamplate compiler simple. Not that we _need_ `SET { ... }` for anything
in-tree other than the X.690 sample...
While we're at it, let's note that the core of PKIX from the RFC
2459/3280/5280/5912 consists of *two* ASN.1 modules, one with
default-EXPLICIT tags, and one with default-IMPLICIT tags, and
Heimdal has these merged as a default-EXPLICIT tags module in
`lib/asn1/rfc2459.asn1`, with `IMPLICIT` added in by hand in all the
tags in the default-IMPLICIT tagged module. This fixes one recently
added type from PKIX that didn't have `IMPLICIT` added in manually!
Finally. We're almost at parity for the template compiler.
Now we have a build option to use templating:
`./configure --enable-asn1-templating`
Tests fail if you build `rfc2459.asn1` with `--template`.
TBD: Figure out what differences remain between the two compilers, and
fix the templating compiler accordingly, adding tests along the
way.
Making IMPLICIT tags work in the templating compiler turned out to be a
simple fix: don't attempt to do anything clever about IMPLICIT tags in
the template generator in the compiler other than denoting them --
instead leave all the smarts about IMPLICIT tags to the interpreter.
This might be a very slight pessimization, but also a great
simplification.
The result is very elegant: when the interpreter finds an IMPLICIT
tag it then recurses to find the template for the body of the type
so-tagged, and evaluates that. Much more elegant than the code
generated by the non-template compiler, not least for not needing
any additional temporary memory allocation.
With this we finally have parity in basic testing of the template
compiler. Indeed, for IMPLICIT tags the template compiler and
interpreter might even be better because they support IMPLICIT tags
with BER lengths, whereas the non-template compiler doesn't (mostly
because `der_replace_tag()` needs to be changed to support it.
And, of course, the template compiler is simply superior in that it
produces smaller code and is *much* easier to work with because the
functions to interpret templates are small and simple. Which means we
can add more functions to deal with other encoding rules fairly
trivially. It should be possible to add all of these with very little
work, almost all of it localized to `lib/asn1/template.c`:
- PER Packed Encoding Rules [X.691]
- XER XML Encoding Rules [X.693]
- OER Octet Encoding Rules [X.696] (intended to replace PER)
- JER JSON Encoding Rules [X.697] (doubles as visual representation)
- GSER Generic String E.R.s [RFC3641] (a visual representation)
- XDR External Data Repr. [STD67][RFC4506]
(XDR is *not* an ASN.1 encoding rules specification, but it's a
*lot* like PER/OER but with 4-octet alignment, and is specified
for the syntax equivalent (XDR) of only a subset of ASN.1 syntax
and semantics.)
All we'd have to do is add variants of `_asn1_{length,encode,decode}()`
for each set of rules, then generate per-type stub functions that call
them (as we already do for DER).
We could then have an encoding rule transliteration program that takes a
`TypeName` and some representation of a value encoded by some encoding
rules, and outputs the same thing encoded by a different set of rules.
This would double as a pretty-printer and parser if we do add support
for JER and/or GSER. It would find the template for the given type
using `dlsym()` against some shared object (possibly `libasn1` itself).
Whereas generating source code for C (or whatever language) for
additional ERs requires much more work. Plus, templates are much
smaller, and the interpreter is tiny, which yields much smaller text and
much smaller CPU icache/dcache footprint, which yields better
performance in many cases.
As well, the template system should be much easier to port to other
languages. Though in the cases of, e.g., Rust, it would require use of
`unsafe` in the interpreter, so in fact the inverse might be true: that
it's easier to generate safe Rust code than to implement a template
interpreter in Rust. Similarly for Haskell, OCAML, etc. But wherever
the template interpreter is easy to implement, it's a huge win.
Note that implementing OER and PER using the templates as they are
currently would be a bit of a challenge, as the interpreter would have
to first do a pass of each SEQUENCE/SET to determine the size and
layout of the OER/PER sequence/set preamble by counting the number of
OPTIONAL/DEFAULT members, BOOLEAN members, and extensibility markers
with extensions present. We could always generate more entries to
encode precomputed preamble metadata. We would also need to add a
template entry type for extensibility markers, which currently we do
not.
This commit _mostly_ fixes the Heimdal ASN.1 compiler to properly
support IMPLICIT tagging in most if not all the many cases where it
didn't already, as you could see in lib/asn1/canthandle.asn1 prior to
this commit.
This fix is a bit of a hack in that a proper fix would change the
function prototypes of the encode/decode/length functions generated by
the compiler to take an optional IMPLICIT tag to tag with instead of the
type they code. That fix would not be localized to lib/asn1/ however,
and would change the API and ABI of generated code (which is mostly not
an ABI for Heimdal, but still, some external projects would have to make
changes).
Instead, for IMPLICIT tags we currently depend on the IMPLICIT tag and
the sub-type's tag having the same size -- this can be fixed with extra
allocation on the encoder side as we do on the decoder side, but we
might leave it for later.
The issue we're fixing manifested as:
-- The [CONTEXT 0] tag in Bar below was turned into an EXPLICIT tag
-- instead of an IMPLICIT one, netting the DER encoding for the `foo`
-- member as:
-- [CONTEXT 0] [UNIVERSAL Seq] [UNIVERSAL Int] <encoding of i>
-- instead of the correct:
-- [CONTEXT 0] [UNIVERSAL Int] <encoding of i>
Foo ::= SEQUENCE { i INTEGER }
Bar ::= SEQUENCE { foo [0] IMPLICIT Foo }
or
Foo ::= INTEGER
Bar ::= SEQUENCE { foo [0] IMPLICIT Foo } -- tag context 0 marked
-- constructed!
I've reviewed this in part by reviewing the output of the compiler
before and after this change using this procedure:
- Run an earlier version of the ASN.1 compiler output for all
modules in lib/asn1/. Save these in a different location.
- Run this (or later) version of the ASN.1 compiler output for
the same modules, adding --original-order for modules that
have been manually sorted already (e.g., rfc2459.asn1).
- Run clang-format on the saved and newest generated C source
and header files.
- Diff the generated output. Substantial differences will
relate to handling of IMPLICIT tagging. These are
particularly evident in the tcg.asn1 module, which uses a lot
of those.
Later commits add test data (certificates with extensions that use
IMPLICIT tagging) taken from external specifications as well, which
exercise this fix.
Non-urgent brokenness yet to be fixed:
- When the IMPLICIT tag and the tag of the underlying type require
differing numbers of bytes to encode, the encoding and decoding will
fail. The prototypes of generated length_*() functions make it
impossible to do much better.
- SET OF <primitive> still crashes the compiler (not a new bug).
Futures:
- Unwind hackery in cms.asn1 that worked around our lack of proper
IMPLICIT tagging support.
Here are some of the generated code deltas one expects to see around
this commit:
$ git checkout $earlier_version
$ ./autogen.sh
$ mkdir build
$ cd build
$ ../configure ...
$ make -j4
$ make check
$ cd lib/asn1
$ for i in *.c; do
[[ $i = asn1parse.? || $i = lex.? || $i = *.h ]] && continue
clang-format -i $i $i
cmp /tmp/save/$i $i && echo NO DIFFS: $i && continue; echo DIFF: $i
done
NO DIFFS: asn1_cms_asn1.c
NO DIFFS: asn1_digest_asn1.c
NO DIFFS: asn1_err.c
NO DIFFS: asn1_krb5_asn1.c
/tmp/save/asn1_kx509_asn1.c asn1_kx509_asn1.c differ: byte 6433, line 264
DIFF: asn1_kx509_asn1.c
NO DIFFS: asn1_ocsp_asn1.c
NO DIFFS: asn1_pkcs10_asn1.c
/tmp/save/asn1_pkcs12_asn1.c asn1_pkcs12_asn1.c differ: byte 12934, line 455
DIFF: asn1_pkcs12_asn1.c
NO DIFFS: asn1_pkcs8_asn1.c
NO DIFFS: asn1_pkcs9_asn1.c
NO DIFFS: asn1_pkinit_asn1.c
/tmp/save/asn1_rfc2459_asn1.c asn1_rfc2459_asn1.c differ: byte 20193, line 532
DIFF: asn1_rfc2459_asn1.c
NO DIFFS: asn1_rfc4043_asn1.c
/tmp/save/asn1_rfc4108_asn1.c asn1_rfc4108_asn1.c differ: byte 595, line 26
DIFF: asn1_rfc4108_asn1.c
/tmp/save/asn1_tcg_asn1.c asn1_tcg_asn1.c differ: byte 31835, line 1229
DIFF: asn1_tcg_asn1.c
/tmp/save/asn1_test_asn1.c asn1_test_asn1.c differ: byte 384, line 21
DIFF: asn1_test_asn1.c
/tmp/save/test_template_asn1-template.c test_template_asn1-template.c differ: byte 650, line 20
DIFF: test_template_asn1-template.c
$
$ cd ../..
$ git checkout $newer_version
$ make -j4 && make check
$ cd lib/asn1
$ for i in *.[ch]; do
[[ $i = asn1parse.? || $i = lex.? || $i = *.h ]] && continue
clang-format -i $i $i
cmp /tmp/save/$i $i && echo NO DIFFS: $i && continue
diff -ubw /tmp/save/$i $i
done | $PAGER
and one should see deltas such as the following:
- a small enhancement to handling of OPTIONAL members:
(data)->macData = calloc(1, sizeof(*(data)->macData));
if ((data)->macData == NULL)
goto fail;
e = decode_PKCS12_MacData(p, len, (data)->macData, &l);
- if (e) {
+ if (e == ASN1_MISSING_FIELD) {
free((data)->macData);
(data)->macData = NULL;
+ } else if (e) {
+ goto fail;
} else {
p += l;
len -= l;
ret += l;
- more complete handling of DEFAULTed members:
e = decode_FWReceiptVersion(p, len, &(data)->version, &l);
- if (e)
+ if (e == ASN1_MISSING_FIELD) {
+ (data)->version = 1;
+ } else if (e) {
goto fail;
- p += l;
- len -= l;
- ret += l;
+ } else {
+ p += l;
+ len -= l;
+ ret += l;
+ }
{
- replacement of tags with implicit tags (encode side):
/* targetUri */
if ((data)->targetUri) {
size_t Top_tag_oldret HEIMDAL_UNUSED_ATTRIBUTE = ret;
ret = 0;
e = encode_URIReference(p, len, (data)->targetUri, &l);
if (e)
return e;
p -= l;
len -= l;
ret += l;
- e = der_put_length_and_tag(p, len, ret, ASN1_C_CONTEXT, PRIM, 4, &l);
+ e = der_replace_tag(p, len, ASN1_C_CONTEXT, CONS, 4);
if (e)
return e;
p -= l;
len -= l;
ret += l;
ret += Top_tag_oldret;
}
- replacement of tags with implicit tags (decode side):
strengthOfFunction_oldlen = len;
if (strengthOfFunction_datalen > len) {
e = ASN1_OVERRUN;
goto fail;
}
len = strengthOfFunction_datalen;
- e = decode_StrengthOfFunction(p, len, (data)->strengthOfFunction, &l);
- if (e)
- goto fail;
- p += l;
- len -= l;
- ret += l;
+ {
+ unsigned char *pcopy;
+ pcopy = calloc(1, len);
+ if (pcopy == 0) {
+ e = ENOMEM;
+ goto fail;
+ }
+ memcpy(pcopy, p, len);
+ e = der_replace_tag(pcopy, len, ASN1_C_UNIV, PRIM, 0);
+ if (e)
+ goto fail;
+ e = decode_StrengthOfFunction(p, len, (data)->strengthOfFunction, &l);
+ if (e)
+ goto fail;
+ p += l;
+ len -= l;
+ ret += l;
+ free(pcopy);
+ }
len = strengthOfFunction_oldlen - strengthOfFunction_datalen;
}
}
{
size_t profileOid_datalen, profileOid_oldlen;
- correct determination of implicit tag constructed vs no for IMPLICT-
tagged named primitive types:
{
size_t profileUri_datalen, profileUri_oldlen;
Der_type profileUri_type;
e = der_match_tag_and_length(p, len, ASN1_C_CONTEXT, &profileUri_type, 2,
&profileUri_datalen, &l);
- if (e == 0 && profileUri_type != PRIM) {
+ if (e == 0 && profileUri_type != CONS) {
e = ASN1_BAD_ID;
}
if (e) {
(data)->profileUri = NULL;
} else {
(data)->profileUri = calloc(1, sizeof(*(data)->profileUri));
if ((data)->profileUri == NULL) {
e = ENOMEM;
goto fail;
}
- correct determination of length of IMPLICT-tagged OIDs:
if ((data)->profileOid) {
size_t Top_tag_oldret = ret;
ret = 0;
ret += der_length_oid((data)->profileOid);
+ ret += 1 + der_length_len(ret);
ret += Top_tag_oldret;
}
These deltas should be examined with the corresponding ASN.1 module at
hand, cross-referencing the source code to the ASN.1 type definitions
and manually applying X.690 rules to double-check the choices of
primitive vs. constructed tag, and the choices of when to replace tags
and when not.
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.
cc-1028 cc: ERROR File = ../source4/heimdal/lib/asn1/gen_template.c, Line = 548
The expression used must have a constant value.
struct templatehead template = { 0L, &(template). tqh_first };
^
If this really fixes the IRIX build, we'll propose this for heimdal upstream.
metze
Signed-off-by: Love Hornquist Astrand <lha@h5l.org>
Looks like they defined basename() in string.h and ntohs/htonl are
implemented in terms of __bswap16() which is a macro with tmp
variables and so one cannot embed one call to ntohs/htons in another.
Not good but we workaround this limitation in glibc.
ASN.1 INTEGERs will now compile to C int64_t or uint64_t, depending
on whether the constraint ranges include numbers that cannot be
represented in 32-bit ints and whether they include negative
numbers.
Template backend support included. check-template is now built with
--template, so we know we're testing it.
Tests included.
most of these warnings are not problems because of ample
use of abort() calls. However, the large number of warnings
makes it difficult to identify real problems. Initialize
the variables to shut up the compilers.
Change-Id: I8477c11b17c7b6a7d9074c721fdd2d7303b186a8
