Quick DER, or if you like, "Quick and Easy DER", is a library for handling DER, which is a widely used binary representation of ASN.1 syntax in binary formats. The library describes ASN.1 syntax in a parser table that can be fed into library routines, resulting in pointer/length descriptors for the individual data fragments.
See also:
- Building and installing Quick DER
- Extended usage examples
- Embedded systems use
- Python bindings
- Syntax definitions for walks and packing
- Changelog (high-level history since 1.1)
- License (BSD 2-Clause)
Quick DER is a three-pronged fork:
- you can use the library to process DER data at a low level (e.g. as a sequence of nested tags with a data payload),
- you can use the data structures provided by the library for a wide variety of RFCs to process data from those RFCs at a higher level (e.g. LDAP messages with named fields),
- the tools provided with the library can be used to translate ASN.1 syntax into (C language) data structures and Python bindings for use as above.
Since Quick DER handles binary data (rather than character strings), all data
is described in so-called dercursor structures; each containing a pointer
and a size of the data pointed at.
The routine der_header() can be used to examine the component at the
start of the DER data. This provides information on the tag and payload
length of the component.
There are routines der_iterate_first() and der_iterate_next() routines
to manually iterate over a DER structure's components. This can be used to
analyse structures that have not been unpacked (yet). The der_countelements()
routine can be used to predict the number of iterations.
There are also routines to manually walk through packaged DER structures,
namely der_enter() and der_skip() to enter into a nested structure and to
find the next element in a concatenation of such elements.
A more advanced form of such walks through a DER structure exists in the
form of der_walk(), which is fed a sequence of enter/skip statements with
tags that will be validated.
The preferred approach of using this library is translating the ASN.1 syntax into a parser table. This is done when building your software, as a phase preceding the compilation of your Quick DER using program. The translation from ASN.1 to C code can be done manually or integrated into the buildsystem for your project (see below).
The resulting path walks are used in calls like der_unpack() to transform
DER into C-style structures, and der_pack() to make the opposite transformation.
The C-style structures are derived from the ASN.1 syntax, and permits access to
the information with no further need of understanding the depths of processing
DER.
The output from mapping ASN.1 to a parser table is an include file for the
C programming language. This defines the various path walks that can be
used to unpack and pack DER data. The output and input of these routines
takes the form of an array of dercursor values.
To simplify use of the unpacked data, there are overlay structures for the
dercursor array. These overlay structures use the same labels that are
used in the ASN.1 syntax, so it is possible to walk around in the structures.
Some parts of the syntax indicate OPTIONAL elements. Such elements result
in the respective dercursor variables to be NULL values; specifically, the
function der_isnull() returns a true value for these elements.
The storage structure for the SEQUENCE OF and SET OF elements is not
the usual dercursor field with uint8_t *derptr and size_t derlen
attributes, but another field type dernode which can take a number of forms,
namely a dercursor with uint8_t *wire.derptr and size_t wire.derlen
or a derarray with an array of similar entries under info.derray and
an array length under info.dercnt. It is the caller's responsibility to
keep track of which parsing-phase is active, and so which kind of data
is stored in the dernode.
We have provided a simple asn2quickder compiler based on
asn1ate
to generate both the overlay data structures and the
byte codes for der_pack() and der_unpack(). For most everyday ASN.1
this should suffice to produce a header file from an ASN.1 module.
To use the compiler (and the associated MacroASN1Module), you will
need to build the Python bindings and tools. See the
Python install documentation.
The compiler generates annotations for external callers that wish to unpack
SEQUENCE OF and SET OF syntax elements. These annotations can be used
for invoking the sub-unpacking and sub-packing procedures in a reasonably
general manner. We think this is the best trade-off between simplicity
on the side of Quick DER, and ability on the side of a potentially more
demanding ASN.1 application.
The tool can generate C language headers and Python bindings for an
ASN.1 description. An additional tool, asn1literate produces documentation
from an ASN.1 description.
For users of CMake, the macros in cmake/MacroASN1Module.cmake may be
an inspiration to automate ASN.1 compilation in those projects. The macros
should not be used verbatim, since they are tied up with compiling ASN.1
before the tools are installed.
Quick DER accepts a liberal superset of DER, while making no memory allocations (although it can use extra memory for handling variable- length structures if you help it). It provides a library of pre- compiled headers and bindings to common protocols that use DER.
The entire library has been designed to operate without dynamic memory allocation. This means that there will never be a memory leak as a result of using Quick DER.
When DER information is unpacked, it is assumed to be loaded into a memory buffer
by the calling program, and the dercursor structures point to portions of that
buffer. The data is stored in dercursor arrays which the user program may
overlay with meaningful, ASN.1-labelled structures. In many applications, such
structures can be allocated on the stack.
Some portions of the data may be dynamically sized, notably the SEQUENCE OF
and SET OF structures, which indicate that the structural description following
it may be repeated in the binary data. Such data portions will be stored and
not yet unpacked by der_unpack(). Based on the stored DER data in a dercursor,
the calling application can choose to use iterators, der_walk() and so on to
avoid actually unpacking it; or it may allocate memory dynamically, and use that
to repeatedly call der_unpack() to find the individual entries.
In short, the Quick DER library never needs to perform memory allocation, and it provides the calling program with a lot of control to avoid it too. This is ideal for embedded applications, but is also beneficial for a secure programming style.
Software built around Quick DER, such as
LillyDAP
for example, may well have its own regime of memory management.
If this is the case, it is quite possible to unpack SEQUENCE OF and
SET OF elements. The dernode type can be overlaid on a regular
dercursor, and the caller can subsequently allocate memory sufficient
for handling the array of dercursors.
The structure inserting the extra symbols wire for wire format and
info for the array is union named dernode; der_unpack() will fill it
with the wire form, but it is assumed that the caller's post-parser will
invoke der_unpack() again on the components, and deliver the result in
info, thereby overwriting the original wire format. Before sending out an
info structure with der_pack(), the calling application should revert to
wire form again through preparing calls to der_pack().
The library can be built for static linking, and can be optimized for code size when linking. See the embedded Quick DER file for details.
We have started a collection of RFC's, stripped down to ASN.1 modules, that can be mapped to header files that can simply be include in a C program:
#include <quick-der/rfc5280.h>
#include <quick-der/rfc4120.h>
#include <quick-der/rfc4511.h>
These examples are included with Quick DER:
- X.509 certificate and CRL structures,
- Kerberos packet formats and messaging,
- LDAP as a protocol description,
- .. a dozen other RFC protocol descriptions.
Most ASN.1 descriptions do things that confuse the compiler, so small changes
have been made, and clearly marked with asn1ate and an explanation in the
ASN.1 files in the rfc/, itu/ and arpa2/ directories. The issues found
were usually quite simple:
- Using uppercase initial characters in type names
- Removing complex constraints (Quick DER does not implement them anyway)
- Cutting an infinitely circular reference for
Filter.notin LDAP.
There are some specifications that go far beyond the abilities of asn1ate,
by using such concepts as macros. This applies to the SNMP RFC's (which are
not included).
There are Python (both Python 2 and Python 3) bindings that can be generated when building the library.
See Python and Quick DER for a plan that greatly simplifies ASN.1 processing with Python.
This is for ASN.1 experts; others can safely skip this subsection.
The Quick DER library is designed to process DER, although it will also accept
some of the more general BER format. If a length or value is written in more
than the minimal space, the library is still likely to accept it. Note that
the BIT STRING type is somewhat likely to run into overflow problems, so
there the full restrictiveness of DER is applied.
Integer types longer than four octets (i.e. 32 bit) are not supported directly.
The library comes with a collection of tests which may also be used as
examples of the library's use. The test / example ldapsearch is
extensively documented; it reads a binary dump (DER) of an LDAP
search request, and prints out the filter expression that is found
in that search.
The example works with the low level interface: it explicitly calls
routines like der_enter() to explore the DER data.
To validate the structures written in DER, both der_unpack() and der_walk()
can be used. Most other routines are coded for flexibility and should not be
assumed to validate DER in more detail than strictly required.
The der_unpack() routine runs through the entire structure, and validates
the tags it runs by, as well as the complete nesting structure it encounters.
It is a complete validation of the structures.
The der_walk() routine performs lazy validation, meaning that it will
carefully check tags and nesting inasfar as it is needed to get from its
starting point to its end point. Anything but the paths explored will be
accepted without question.
The LillyDAP framework builds on top of Quick DER to process LDAP messages. It uses the headers for LDAP (RFC 4511) from Quick DER and the higher level API to translate DER data into C structures.
LillyDAP, in turn, is a framework for building miniature LDAP services. These can help you centralise your data storage under own control; for instance, your PGP key ring or your vCard collection are good candidates for sharing locally.
There are a few things that this library can use:
- More testing. The current
testdirectory is far too small; we can take a PKIX certificate apart and re-compose it, so we're definately doing something good, but this is nowhere near thorough testing. If you run into a problem case, then please share it so we can solve it and extend our test base. - Compiler output testing. The compiler output is currently compiled to see if there are C syntactical problems, but that is all. They have been visually compared to manually crafted code, too. More exhaustive tests, including a full application, would be in order.
And of course, there are many useful things we may do with this library:
- Kerberos in PKIX. Certificates wrapping Kerberos Tickets for use with TLS-KDH,
- Remote PKCS #11. The main issue in doing this well is proper encapsulation, but with Quick and Easy DER, and the emphasis of both on security and well-defined sizes, it appears to be a perfect match to wrap PKCS #11 arguments in DER structures.