/* tlv-parser.c - Parse BER encoded objects
* Copyright (C) 2023 g10 Code GmbH
*
* This file is part of GnuPG.
*
* This file is free software; you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as
* published by the Free Software Foundation; either version 2.1 of
* the License, or (at your option) any later version.
*
* This file 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 Lesser General Public License
* along with this program; if not, see <https://www.gnu.org/licenses/>.
* SPDX-License-Identifier: LGPL-2.1-or-later
*/
#include <config.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <gpg-error.h>
#include "util.h"
#include "tlv.h"
#define TLV_MAX_DEPTH 25
struct bufferlist_s
{
struct bufferlist_s *next;
char *buffer;
};
/* An object to control the ASN.1 parsing. */
struct tlv_parser_s
{
/* The orginal buffer with the entire pkcs#12 object and its length. */
const unsigned char *origbuffer;
size_t origbufsize;
/* The current buffer we are working on and its length. */
const unsigned char *buffer;
size_t bufsize;
int in_ndef; /* Flag indicating that we are in a NDEF. */
int pending; /* The last tlv_next has not yet been processed. */
struct tag_info ti; /* The current tag. */
gpg_error_t lasterr; /* Last error from tlv function. */
const char *lastfunc;/* Name of last called function. */
int verbosity; /* Arg from tlv_parser_new. */
struct bufferlist_s *bufferlist; /* To keep track of malloced buffers. */
unsigned int stacklen; /* Used size of the stack. */
struct {
const unsigned char *buffer; /* Saved value of BUFFER. */
size_t bufsize; /* Saved value of BUFSIZE. */
size_t length; /* Length of the container (ti.length). */
int in_ndef; /* Saved IN_NDEF flag (ti.ndef). */
} stack[TLV_MAX_DEPTH];
};
static unsigned char *cram_octet_string (const unsigned char *input,
size_t length, size_t *r_newlength);
static int need_octet_string_cramming (const unsigned char *input,
size_t length);
void
_tlv_parser_dump_tag (const char *text, int lno, tlv_parser_t tlv)
{
struct tag_info *ti;
if (!tlv || tlv->verbosity < 2)
return;
ti = &tlv->ti;
log_debug ("p12_parse:%s:%d: @%04zu class=%d tag=%lu len=%zu nhdr=%zu %s%s\n",
text, lno,
(size_t)(tlv->buffer - tlv->origbuffer) - ti->nhdr,
ti->class, ti->tag, ti->length, ti->nhdr,
ti->is_constructed?" cons":"",
ti->ndef?" ndef":"");
}
void
_tlv_parser_dump_state (const char *text, const char *text2,
int lno, tlv_parser_t tlv)
{
if (!tlv || tlv->verbosity < 2)
return;
log_debug ("p12_parse:%s%s%s:%d: @%04zu lvl=%u %s\n",
text,
text2? "/":"", text2? text2:"",
lno,
(size_t)(tlv->buffer - tlv->origbuffer),
tlv->stacklen,
tlv->in_ndef? " in-ndef":"");
}
/* Parse the buffer at the address BUFFER which is of SIZE and return
* the tag and the length part from the TLV triplet. Update BUFFER
* and SIZE on success. Checks that the encoded length does not
* exhaust the length of the provided buffer. */
static int
parse_tag (unsigned char const **buffer, size_t *size, struct tag_info *ti)
{
gpg_error_t err;
int tag;
err = parse_ber_header (buffer, size,
&ti->class, &tag,
&ti->is_constructed, &ti->ndef,
&ti->length, &ti->nhdr);
if (err)
return err;
if (tag < 0)
return gpg_error (GPG_ERR_EOVERFLOW);
ti->tag = tag;
if (ti->length > *size)
return gpg_error (GPG_ERR_BUFFER_TOO_SHORT); /* data larger than buffer. */
return 0;
}
/* Public version of parse_tag. */
gpg_error_t
tlv_parse_tag (unsigned char const **buffer, size_t *size, struct tag_info *ti)
{
return parse_tag (buffer, size, ti);
}
/* Create a new TLV object. */
tlv_parser_t
tlv_parser_new (const unsigned char *buffer, size_t bufsize, int verbosity)
{
tlv_parser_t tlv;
tlv = xtrycalloc (1, sizeof *tlv);
if (tlv)
{
tlv->origbuffer = buffer;
tlv->origbufsize = bufsize;
tlv->buffer = buffer;
tlv->bufsize = bufsize;
tlv->verbosity = verbosity;
}
return tlv;
}
/* This function can be used to store a malloced buffer into the TLV
* object. Ownership of BUFFER is thus transferred to TLV. This
* buffer will then only be released by tlv_release. */
static gpg_error_t
register_buffer (tlv_parser_t tlv, char *buffer)
{
struct bufferlist_s *item;
item = xtrycalloc (1, sizeof *item);
if (!item)
return gpg_error_from_syserror ();
item->buffer = buffer;
item->next = tlv->bufferlist;
tlv->bufferlist = item;
return 0;
}
void
tlv_parser_release (tlv_parser_t tlv)
{
if (!tlv)
return;
while (tlv->bufferlist)
{
struct bufferlist_s *save = tlv->bufferlist->next;
xfree (tlv->bufferlist->buffer);
xfree (tlv->bufferlist);
tlv->bufferlist = save;
}
xfree (tlv);
}
/* Helper for the tlv_peek functions. */
static gpg_error_t
_tlv_peek (tlv_parser_t tlv, struct tag_info *ti)
{
const unsigned char *p;
size_t n;
/* Note that we want to peek ahead of any current container but of
* course not beyond our entire buffer. */
p = tlv->buffer;
if ((p - tlv->origbuffer) > tlv->origbufsize)
return gpg_error (GPG_ERR_BUG);
n = tlv->origbufsize - (p - tlv->origbuffer);
return parse_tag (&p, &n, ti);
}
/* Look for the next tag and return true if it matches CLASS and TAG.
* Otherwise return false. No state is changed. */
int
_tlv_parser_peek (tlv_parser_t tlv, int class, int tag)
{
struct tag_info ti;
return (!_tlv_peek (tlv, &ti)
&& ti.class == class && ti.tag == tag);
}
/* Look for the next tag and return true if it is the Null tag.
* Otherwise return false. No state is changed. */
int
_tlv_parser_peek_null (tlv_parser_t tlv)
{
struct tag_info ti;
return (!_tlv_peek (tlv, &ti)
&& ti.class == CLASS_UNIVERSAL && ti.tag == TAG_NULL
&& !ti.is_constructed && !ti.length);
}
/* Helper for tlv_expect_sequence and tlv_expect_context_tag. */
static gpg_error_t
_tlv_push (tlv_parser_t tlv)
{
/* Right now our pointer is at the value of the current container.
* We push that info onto the stack. */
if (tlv->stacklen >= TLV_MAX_DEPTH)
return (tlv->lasterr = gpg_error (GPG_ERR_TOO_MANY));
tlv->stack[tlv->stacklen].buffer = tlv->buffer;
tlv->stack[tlv->stacklen].bufsize = tlv->bufsize;
tlv->stack[tlv->stacklen].in_ndef = tlv->in_ndef;
tlv->stack[tlv->stacklen].length = tlv->ti.length;
tlv->stacklen++;
tlv->in_ndef = tlv->ti.ndef;
/* We set the size of the buffer to the TLV length if it is known or
* else to the size of the remaining entire buffer. */
if (tlv->in_ndef)
{
if ((tlv->buffer - tlv->origbuffer) > tlv->origbufsize)
return (tlv->lasterr = gpg_error (GPG_ERR_BUG));
tlv->bufsize = tlv->origbufsize - (tlv->buffer - tlv->origbuffer);
}
else
tlv->bufsize = tlv->ti.length;
_tlv_parser_dump_state (__func__, NULL, 0, tlv);
return 0;
}
/* Helper for tlv_next. */
static gpg_error_t
_tlv_pop (tlv_parser_t tlv)
{
size_t lastlen;
/* We reached the end of a container, either due to the size limit
* or due to an end tag. Now we pop the last container so that we
* are positioned at the value of the last container. */
if (!tlv->stacklen)
return gpg_error (GPG_ERR_EOF);
tlv->stacklen--;
tlv->in_ndef = tlv->stack[tlv->stacklen].in_ndef;
if (tlv->in_ndef)
{
/* We keep buffer but adjust bufsize to the end of the origbuffer. */
if ((tlv->buffer - tlv->origbuffer) > tlv->origbufsize)
return (tlv->lasterr = gpg_error (GPG_ERR_BUG));
tlv->bufsize = tlv->origbufsize - (tlv->buffer - tlv->origbuffer);
}
else
{
lastlen = tlv->stack[tlv->stacklen].length;
tlv->buffer = tlv->stack[tlv->stacklen].buffer;
tlv->bufsize = tlv->stack[tlv->stacklen].bufsize;
if (lastlen > tlv->bufsize)
{
log_debug ("%s: container length larger than buffer (%zu/%zu)\n",
__func__, lastlen, tlv->bufsize);
return gpg_error (GPG_ERR_INV_BER);
}
tlv->buffer += lastlen;
tlv->bufsize -= lastlen;
}
_tlv_parser_dump_state (__func__, NULL, 0, tlv);
return 0;
}
/* Parse the next tag and value. Also detect the end of a
* container. The caller should use the tlv_next macro. */
gpg_error_t
_tlv_parser_next (tlv_parser_t tlv, int lno)
{
gpg_error_t err;
tlv->lasterr = 0;
tlv->lastfunc = __func__;
if (tlv->pending)
{
tlv->pending = 0;
if (tlv->verbosity > 1)
log_debug ("%s: skipped\n", __func__);
return 0;
}
if (tlv->verbosity > 1)
log_debug ("%s: called\n", __func__);
/* If we are at the end of an ndef container pop the stack. */
if (!tlv->in_ndef && !tlv->bufsize)
{
do
err = _tlv_pop (tlv);
while (!err && !tlv->in_ndef && !tlv->bufsize);
if (err)
return (tlv->lasterr = err);
if (tlv->verbosity > 1)
log_debug ("%s: container(s) closed due to size\n", __func__);
}
again:
/* Get the next tag. */
err = parse_tag (&tlv->buffer, &tlv->bufsize, &tlv->ti);
if (err)
{
if (tlv->verbosity > 1)
log_debug ("%s: reading tag returned err=%d\n", __func__, err);
return err;
}
/* If there is an end tag in an ndef container pop the stack. Also
* pop other containers which are fully consumed. */
if (tlv->in_ndef && (tlv->ti.class == CLASS_UNIVERSAL
&& !tlv->ti.tag && !tlv->ti.is_constructed))
{
do
err = _tlv_pop (tlv);
while (!err && !tlv->in_ndef && !tlv->bufsize);
if (err)
return (tlv->lasterr = err);
if (tlv->verbosity > 1)
log_debug ("%s: container(s) closed due to end tag\n", __func__);
goto again;
}
_tlv_parser_dump_tag (__func__, lno, tlv);
return 0;
}
/* Return the current neting level of the TLV object. */
unsigned int
tlv_parser_level (tlv_parser_t tlv)
{
return tlv? tlv->stacklen : 0;
}
/* Returns the current offset of the parser. */
size_t
tlv_parser_offset (tlv_parser_t tlv)
{
return tlv? (size_t)(tlv->buffer - tlv->origbuffer) : 0;
}
/* Return a string with the last function used. If TLV is NULL an
* empty string is returned. */
const char *
tlv_parser_lastfunc (tlv_parser_t tlv)
{
return tlv? tlv->lastfunc:"";
}
const char *
tlv_parser_lasterrstr (tlv_parser_t tlv)
{
return tlv? gpg_strerror (tlv->lasterr) : "tlv parser not yet initialized";
}
/* Set a flag to indicate that the last tlv_next has not yet been
* consumed. */
void
tlv_parser_set_pending (tlv_parser_t tlv)
{
tlv->pending = 1;
}
/* Return the length of the last read tag. If with_header is 1 the
* lengtb of the header is added to the returned length. */
size_t
tlv_parser_tag_length (tlv_parser_t tlv, int with_header)
{
if (with_header)
return tlv->ti.length + tlv->ti.nhdr;
else
return tlv->ti.length;
}
/* Skip over the value of the current tag. Does not yet work for ndef
* containers. */
void
tlv_parser_skip (tlv_parser_t tlv)
{
tlv->lastfunc = __func__;
log_assert (tlv->bufsize >= tlv->ti.length);
tlv->buffer += tlv->ti.length;
tlv->bufsize -= tlv->ti.length;
}
/* Expect that the current tag is a sequence and setup the context for
* processing. */
gpg_error_t
tlv_expect_sequence (tlv_parser_t tlv)
{
tlv->lastfunc = __func__;
if (!(tlv->ti.class == CLASS_UNIVERSAL && tlv->ti.tag == TAG_SEQUENCE
&& tlv->ti.is_constructed))
return (tlv->lasterr = gpg_error (GPG_ERR_INV_OBJ));
return _tlv_push (tlv);
}
/* Expect that the current tag is a context tag and setup the context
* for processing. The tag of the context is returned at R_TAG. */
gpg_error_t
tlv_expect_context_tag (tlv_parser_t tlv, int *r_tag)
{
tlv->lastfunc = __func__;
if (!(tlv->ti.class == CLASS_CONTEXT && tlv->ti.is_constructed))
return (tlv->lasterr = gpg_error (GPG_ERR_INV_OBJ));
*r_tag = tlv->ti.tag;
return _tlv_push (tlv);
}
/* Expect that the current tag is a SET and setup the context for
* processing. */
gpg_error_t
tlv_expect_set (tlv_parser_t tlv)
{
tlv->lastfunc = __func__;
if (!(tlv->ti.class == CLASS_UNIVERSAL && tlv->ti.tag == TAG_SET
&& tlv->ti.is_constructed))
return (tlv->lasterr = gpg_error (GPG_ERR_INV_OBJ));
return _tlv_push (tlv);
}
/* Expect an object of CLASS with TAG and store its value at
* (R_DATA,R_DATALEN). Then skip over its value to the next tag.
* Note that the stored value is not allocated but points into
* TLV. */
gpg_error_t
tlv_expect_object (tlv_parser_t tlv, int class, int tag,
unsigned char const **r_data, size_t *r_datalen)
{
gpg_error_t err;
const unsigned char *p;
size_t n;
int needpush = 0;
tlv->lastfunc = __func__;
if (!(tlv->ti.class == class && tlv->ti.tag == tag))
return (tlv->lasterr = gpg_error (GPG_ERR_INV_OBJ));
p = tlv->buffer;
n = tlv->ti.length;
if (!n && tlv->ti.ndef)
{
n = tlv->bufsize;
needpush = 1;
}
else if (!tlv->ti.length)
return (tlv->lasterr = gpg_error (GPG_ERR_TOO_SHORT));
if (class == CLASS_CONTEXT && tag == 0 && tlv->ti.is_constructed
&& need_octet_string_cramming (p, n))
{
char *newbuffer;
newbuffer = cram_octet_string (p, n, r_datalen);
if (!newbuffer)
return (tlv->lasterr = gpg_error (GPG_ERR_BAD_BER));
err = register_buffer (tlv, newbuffer);
if (err)
{
xfree (newbuffer);
return (tlv->lasterr = err);
}
*r_data = newbuffer;
}
else
{
*r_data = p;
*r_datalen = n;
}
if (needpush)
return _tlv_push (tlv);
if (!(tlv->bufsize >= tlv->ti.length))
return (tlv->lasterr = gpg_error (GPG_ERR_TOO_SHORT));
tlv->buffer += tlv->ti.length;
tlv->bufsize -= tlv->ti.length;
return 0;
}
/* Expect that the current tag is an object string and store its value
* at (R_DATA,R_DATALEN). Then skip over its value to the next tag.
* Note that the stored value are not allocated but point into TLV.
* If ENCAPSULATES is set the octet string is used as a new
* container. R_DATA and R_DATALEN are optional. */
gpg_error_t
tlv_expect_octet_string (tlv_parser_t tlv, int encapsulates,
unsigned char const **r_data, size_t *r_datalen)
{
gpg_error_t err;
const unsigned char *p;
size_t n;
tlv->lastfunc = __func__;
if (!(tlv->ti.class == CLASS_UNIVERSAL && tlv->ti.tag == TAG_OCTET_STRING
&& (!tlv->ti.is_constructed || encapsulates)))
return (tlv->lasterr = gpg_error (GPG_ERR_INV_OBJ));
p = tlv->buffer;
if (!(n=tlv->ti.length) && !tlv->ti.ndef)
return (tlv->lasterr = gpg_error (GPG_ERR_TOO_SHORT));
if (encapsulates && tlv->ti.is_constructed
&& need_octet_string_cramming (p, n))
{
char *newbuffer;
newbuffer = cram_octet_string (p, n, r_datalen);
if (!newbuffer)
return (tlv->lasterr = gpg_error (GPG_ERR_BAD_BER));
err = register_buffer (tlv, newbuffer);
if (err)
{
xfree (newbuffer);
return (tlv->lasterr = err);
}
*r_data = newbuffer;
}
else
{
if (r_data)
*r_data = p;
if (r_datalen)
*r_datalen = tlv->ti.length;
}
if (encapsulates)
return _tlv_push (tlv);
if (!(tlv->bufsize >= tlv->ti.length))
return (tlv->lasterr = gpg_error (GPG_ERR_TOO_SHORT));
tlv->buffer += tlv->ti.length;
tlv->bufsize -= tlv->ti.length;
return 0;
}
/* Expect that the current tag is an integer and return its value at
* R_VALUE. Then skip over its value to the next tag. */
gpg_error_t
tlv_expect_integer (tlv_parser_t tlv, int *r_value)
{
const unsigned char *p;
size_t n;
int value;
tlv->lastfunc = __func__;
if (!(tlv->ti.class == CLASS_UNIVERSAL && tlv->ti.tag == TAG_INTEGER
&& !tlv->ti.is_constructed))
return (tlv->lasterr = gpg_error (GPG_ERR_INV_OBJ));
p = tlv->buffer;
if (!(n=tlv->ti.length))
return (tlv->lasterr = gpg_error (GPG_ERR_TOO_SHORT));
/* We currently support only positive values. */
if ((*p & 0x80))
return (tlv->lasterr = gpg_error (GPG_ERR_ERANGE));
for (value = 0; n; n--)
{
value <<= 8;
value |= (*p++) & 0xff;
if (value < 0)
return (tlv->lasterr = gpg_error (GPG_ERR_EOVERFLOW));
}
*r_value = value;
if (!(tlv->bufsize >= tlv->ti.length))
return (tlv->lasterr = gpg_error (GPG_ERR_TOO_SHORT));
tlv->buffer += tlv->ti.length;
tlv->bufsize -= tlv->ti.length;
return 0;
}
/* Variant of tlv_expect_integer which returns an MPI. If IGNORE_ZERO
* is set a value of 0 is ignored and R_VALUE not changed and the
* function returns GPG_ERR_FALSE. No check for negative encoded
* integers is done because the old code here worked the same and we
* can't foreclose invalid encoded PKCS#12 stuff - after all it is
* PKCS#12 see https://www.cs.auckland.ac.nz/~pgut001/pubs/pfx.html */
#ifdef GCRYPT_VERSION
gpg_error_t
tlv_expect_mpinteger (tlv_parser_t tlv, int ignore_zero,
gcry_mpi_t *r_value)
{
const unsigned char *p;
size_t n;
tlv->lastfunc = __func__;
if (!(tlv->ti.class == CLASS_UNIVERSAL && tlv->ti.tag == TAG_INTEGER
&& !tlv->ti.is_constructed))
return (tlv->lasterr = gpg_error (GPG_ERR_INV_OBJ));
p = tlv->buffer;
if (!(n=tlv->ti.length))
return (tlv->lasterr = gpg_error (GPG_ERR_TOO_SHORT));
if (!(tlv->bufsize >= tlv->ti.length))
return (tlv->lasterr = gpg_error (GPG_ERR_TOO_SHORT));
tlv->buffer += tlv->ti.length;
tlv->bufsize -= tlv->ti.length;
if (ignore_zero && n == 1 && !*p)
return gpg_error (GPG_ERR_FALSE);
return gcry_mpi_scan (r_value, GCRYMPI_FMT_USG, p, n, NULL);
}
#endif /*GCRYPT_VERSION*/
/* Expect that the current tag is an object id and store its value at
* (R_OID,R_OIDLEN). Then skip over its value to the next tag. Note
* that the stored value is not allocated but points into TLV. */
gpg_error_t
tlv_expect_object_id (tlv_parser_t tlv,
unsigned char const **r_oid, size_t *r_oidlen)
{
const unsigned char *p;
size_t n;
tlv->lastfunc = __func__;
if (!(tlv->ti.class == CLASS_UNIVERSAL && tlv->ti.tag == TAG_OBJECT_ID
&& !tlv->ti.is_constructed))
return (tlv->lasterr = gpg_error (GPG_ERR_INV_OBJ));
p = tlv->buffer;
if (!(n=tlv->ti.length))
return (tlv->lasterr = gpg_error (GPG_ERR_TOO_SHORT));
*r_oid = p;
*r_oidlen = tlv->ti.length;
if (!(tlv->bufsize >= tlv->ti.length))
return (tlv->lasterr = gpg_error (GPG_ERR_TOO_SHORT));
tlv->buffer += tlv->ti.length;
tlv->bufsize -= tlv->ti.length;
return 0;
}
/* Given an ASN.1 chunk of a structure like:
*
* 24 NDEF: OCTET STRING -- This is not passed to us
* 04 1: OCTET STRING -- INPUT point s to here
* : 30
* 04 1: OCTET STRING
* : 80
* [...]
* 04 2: OCTET STRING
* : 00 00
* : } -- This denotes a Null tag and are the last
* -- two bytes in INPUT.
*
* The example is from Mozilla Firefox 1.0.4 which actually exports
* certs as single byte chunks of octet strings.
*
* Create a new buffer with the content of that octet string. INPUT
* is the original buffer with a LENGTH. Returns
* NULL on error or a new malloced buffer with its actual used length
* stored at R_NEWLENGTH. */
static unsigned char *
cram_octet_string (const unsigned char *input, size_t length,
size_t *r_newlength)
{
const unsigned char *s = input;
size_t n = length;
unsigned char *output, *d;
struct tag_info ti;
/* Allocate output buf. We know that it won't be longer than the
input buffer. */
d = output = xtrymalloc (length);
if (!output)
goto bailout;
while (n)
{
if (parse_tag (&s, &n, &ti))
goto bailout;
if (ti.class == CLASS_UNIVERSAL && ti.tag == TAG_OCTET_STRING
&& !ti.ndef && !ti.is_constructed)
{
memcpy (d, s, ti.length);
s += ti.length;
d += ti.length;
n -= ti.length;
}
else if (ti.class == CLASS_UNIVERSAL && !ti.tag && !ti.is_constructed)
break; /* Ready */
else
goto bailout;
}
*r_newlength = d - output;
return output;
bailout:
xfree (output);
return NULL;
}
/* Return true if (INPUT,LENGTH) is a structure which should be passed
* to cram_octet_string. This is basically the same loop as in
* cram_octet_string but without any actual copying. */
static int
need_octet_string_cramming (const unsigned char *input, size_t length)
{
const unsigned char *s = input;
size_t n = length;
struct tag_info ti;
if (!length)
return 0;
while (n)
{
if (parse_tag (&s, &n, &ti))
return 0;
if (ti.class == CLASS_UNIVERSAL && ti.tag == TAG_OCTET_STRING
&& !ti.ndef && !ti.is_constructed)
{
s += ti.length;
n -= ti.length;
}
else if (ti.class == CLASS_UNIVERSAL && !ti.tag && !ti.is_constructed)
break; /* Ready */
else
return 0;
}
return 1;
}