Adding upstream version 2.2.40.upstream/2.2.40

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

1 files changed, 1808 insertions, 0 deletions

diff --git a/g10/build-packet.c b/g10/build-packet.c
new file mode 100644
index 0000000..a40ed0d
--- /dev/null
+++ b/g10/build-packet.c
@@ -0,0 +1,1808 @@
+/* build-packet.c - assemble packets and write them
+ * Copyright (C) 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005,
+ *               2006, 2010, 2011  Free Software Foundation, Inc.
+ *
+ * This file is part of GnuPG.
+ *
+ * GnuPG 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 of the License, or
+ * (at your option) any later version.
+ *
+ * GnuPG 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; if not, see <https://www.gnu.org/licenses/>.
+ */
+
+#include <config.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <ctype.h>
+
+#include "gpg.h"
+#include "../common/util.h"
+#include "packet.h"
+#include "../common/status.h"
+#include "../common/iobuf.h"
+#include "../common/i18n.h"
+#include "options.h"
+#include "../common/host2net.h"
+
+static gpg_error_t do_ring_trust (iobuf_t out, PKT_ring_trust *rt);
+static int do_user_id( IOBUF out, int ctb, PKT_user_id *uid );
+static int do_key (iobuf_t out, int ctb, PKT_public_key *pk);
+static int do_symkey_enc( IOBUF out, int ctb, PKT_symkey_enc *enc );
+static int do_pubkey_enc( IOBUF out, int ctb, PKT_pubkey_enc *enc );
+static u32 calc_plaintext( PKT_plaintext *pt );
+static int do_plaintext( IOBUF out, int ctb, PKT_plaintext *pt );
+static int do_encrypted( IOBUF out, int ctb, PKT_encrypted *ed );
+static int do_encrypted_mdc( IOBUF out, int ctb, PKT_encrypted *ed );
+static int do_compressed( IOBUF out, int ctb, PKT_compressed *cd );
+static int do_signature( IOBUF out, int ctb, PKT_signature *sig );
+static int do_onepass_sig( IOBUF out, int ctb, PKT_onepass_sig *ops );
+
+static int calc_header_length( u32 len, int new_ctb );
+static int write_16(IOBUF inp, u16 a);
+static int write_32(IOBUF inp, u32 a);
+static int write_header( IOBUF out, int ctb, u32 len );
+static int write_sign_packet_header( IOBUF out, int ctb, u32 len );
+static int write_header2( IOBUF out, int ctb, u32 len, int hdrlen );
+static int write_new_header( IOBUF out, int ctb, u32 len, int hdrlen );
+
+/* Returns 1 if CTB is a new format ctb and 0 if CTB is an old format
+   ctb.  */
+static int
+ctb_new_format_p (int ctb)
+{
+  /* Bit 7 must always be set.  */
+  log_assert ((ctb & (1 << 7)));
+  /* Bit 6 indicates whether the packet is a new format packet.  */
+  return (ctb & (1 << 6));
+}
+
+/* Extract the packet type from a CTB.  */
+static int
+ctb_pkttype (int ctb)
+{
+  if (ctb_new_format_p (ctb))
+    /* Bits 0 through 5 are the packet type.  */
+    return (ctb & ((1 << 6) - 1));
+  else
+    /* Bits 2 through 5 are the packet type.  */
+    return (ctb & ((1 << 6) - 1)) >> 2;
+}
+
+
+/* Build a packet and write it to the stream OUT.
+ * Returns: 0 on success or on an error code.  */
+int
+build_packet (IOBUF out, PACKET *pkt)
+{
+  int rc = 0;
+  int new_ctb = 0;
+  int ctb, pkttype;
+
+  if (DBG_PACKET)
+    log_debug ("build_packet() type=%d\n", pkt->pkttype);
+  log_assert (pkt->pkt.generic);
+
+  switch ((pkttype = pkt->pkttype))
+    {
+    case PKT_PUBLIC_KEY:
+      if (pkt->pkt.public_key->seckey_info)
+        pkttype = PKT_SECRET_KEY;
+      break;
+    case PKT_PUBLIC_SUBKEY:
+      if (pkt->pkt.public_key->seckey_info)
+        pkttype = PKT_SECRET_SUBKEY;
+      break;
+    case PKT_PLAINTEXT:
+      new_ctb = pkt->pkt.plaintext->new_ctb;
+      break;
+    case PKT_ENCRYPTED:
+    case PKT_ENCRYPTED_MDC:
+      new_ctb = pkt->pkt.encrypted->new_ctb;
+      break;
+    case PKT_COMPRESSED:
+      new_ctb = pkt->pkt.compressed->new_ctb;
+      break;
+    case PKT_USER_ID:
+      if (pkt->pkt.user_id->attrib_data)
+        pkttype = PKT_ATTRIBUTE;
+      break;
+    default:
+      break;
+    }
+
+  if (new_ctb || pkttype > 15) /* new format */
+    ctb = (0xc0 | (pkttype & 0x3f));
+  else
+    ctb = (0x80 | ((pkttype & 15)<<2));
+  switch (pkttype)
+    {
+    case PKT_ATTRIBUTE:
+    case PKT_USER_ID:
+      rc = do_user_id (out, ctb, pkt->pkt.user_id);
+      break;
+    case PKT_OLD_COMMENT:
+    case PKT_COMMENT:
+      /* Ignore these.  Theoretically, this will never be called as we
+       * have no way to output comment packets any longer, but just in
+       * case there is some code path that would end up outputting a
+       * comment that was written before comments were dropped (in the
+       * public key?) this is a no-op. 	*/
+      break;
+    case PKT_PUBLIC_SUBKEY:
+    case PKT_PUBLIC_KEY:
+    case PKT_SECRET_SUBKEY:
+    case PKT_SECRET_KEY:
+      rc = do_key (out, ctb, pkt->pkt.public_key);
+      break;
+    case PKT_SYMKEY_ENC:
+      rc = do_symkey_enc (out, ctb, pkt->pkt.symkey_enc);
+      break;
+    case PKT_PUBKEY_ENC:
+      rc = do_pubkey_enc (out, ctb, pkt->pkt.pubkey_enc);
+      break;
+    case PKT_PLAINTEXT:
+      rc = do_plaintext (out, ctb, pkt->pkt.plaintext);
+      break;
+    case PKT_ENCRYPTED:
+      rc = do_encrypted (out, ctb, pkt->pkt.encrypted);
+      break;
+    case PKT_ENCRYPTED_MDC:
+      rc = do_encrypted_mdc (out, ctb, pkt->pkt.encrypted);
+      break;
+    case PKT_COMPRESSED:
+      rc = do_compressed (out, ctb, pkt->pkt.compressed);
+      break;
+    case PKT_SIGNATURE:
+      rc = do_signature (out, ctb, pkt->pkt.signature);
+      break;
+    case PKT_ONEPASS_SIG:
+      rc = do_onepass_sig (out, ctb, pkt->pkt.onepass_sig);
+      break;
+    case PKT_RING_TRUST:
+      /* Ignore it (only written by build_packet_and_meta)  */
+      break;
+    case PKT_MDC:
+      /* We write it directly, so we should never see it here. */
+    default:
+      log_bug ("invalid packet type in build_packet()\n");
+      break;
+    }
+
+  return rc;
+}
+
+
+/* Build a packet and write it to the stream OUT.  This variant also
+ * writes the meta data using ring trust packets.  Returns: 0 on
+ * success or on error code.  */
+gpg_error_t
+build_packet_and_meta (iobuf_t out, PACKET *pkt)
+{
+  gpg_error_t err;
+  PKT_ring_trust rt = {0};
+
+  err = build_packet (out, pkt);
+  if (err)
+    ;
+  else if (pkt->pkttype == PKT_SIGNATURE)
+    {
+      PKT_signature *sig = pkt->pkt.signature;
+
+      rt.subtype = RING_TRUST_SIG;
+      /* Note: trustval is not yet used.  */
+      if (sig->flags.checked)
+        {
+          rt.sigcache = 1;
+          if (sig->flags.valid)
+            rt.sigcache |= 2;
+        }
+      err = do_ring_trust (out, &rt);
+    }
+  else if (pkt->pkttype == PKT_USER_ID
+           || pkt->pkttype == PKT_ATTRIBUTE)
+    {
+      PKT_user_id *uid = pkt->pkt.user_id;
+
+      rt.subtype = RING_TRUST_UID;
+      rt.keyorg = uid->keyorg;
+      rt.keyupdate = uid->keyupdate;
+      rt.url = uid->updateurl;
+      err = do_ring_trust (out, &rt);
+      rt.url = NULL;
+    }
+  else if (pkt->pkttype == PKT_PUBLIC_KEY
+           || pkt->pkttype == PKT_SECRET_KEY)
+    {
+      PKT_public_key *pk = pkt->pkt.public_key;
+
+      rt.subtype = RING_TRUST_KEY;
+      rt.keyorg = pk->keyorg;
+      rt.keyupdate = pk->keyupdate;
+      rt.url = pk->updateurl;
+      err = do_ring_trust (out, &rt);
+      rt.url = NULL;
+
+    }
+
+  return err;
+}
+
+
+/*
+ * Write the mpi A to OUT.
+ */
+gpg_error_t
+gpg_mpi_write (iobuf_t out, gcry_mpi_t a)
+{
+  int rc;
+
+  if (gcry_mpi_get_flag (a, GCRYMPI_FLAG_OPAQUE))
+    {
+      unsigned int nbits;
+      const unsigned char *p;
+      unsigned char lenhdr[2];
+
+      /* gcry_log_debugmpi ("a", a); */
+      p = gcry_mpi_get_opaque (a, &nbits);
+      if (p)
+        {
+          /* Strip leading zero bits.  */
+          for (; nbits >= 8 && !*p; p++, nbits -= 8)
+            ;
+          if (nbits >= 8 && !(*p & 0x80))
+            if (--nbits >= 7 && !(*p & 0x40))
+              if (--nbits >= 6 && !(*p & 0x20))
+                if (--nbits >= 5 && !(*p & 0x10))
+                  if (--nbits >= 4 && !(*p & 0x08))
+                    if (--nbits >= 3 && !(*p & 0x04))
+                      if (--nbits >= 2 && !(*p & 0x02))
+                        if (--nbits >= 1 && !(*p & 0x01))
+                          --nbits;
+        }
+      /* gcry_log_debug ("   [%u bit]\n", nbits); */
+      /* gcry_log_debughex (" ", p, (nbits+7)/8); */
+      lenhdr[0] = nbits >> 8;
+      lenhdr[1] = nbits;
+      rc = iobuf_write (out, lenhdr, 2);
+      if (!rc && p)
+        rc = iobuf_write (out, p, (nbits+7)/8);
+    }
+  else
+    {
+      char buffer[(MAX_EXTERN_MPI_BITS+7)/8+2]; /* 2 is for the mpi length. */
+      size_t nbytes;
+
+      nbytes = DIM(buffer);
+      rc = gcry_mpi_print (GCRYMPI_FMT_PGP, buffer, nbytes, &nbytes, a );
+      if( !rc )
+        rc = iobuf_write( out, buffer, nbytes );
+      else if (gpg_err_code(rc) == GPG_ERR_TOO_SHORT )
+        {
+          log_info ("mpi too large (%u bits)\n", gcry_mpi_get_nbits (a));
+          /* The buffer was too small. We better tell the user about the MPI. */
+          rc = gpg_error (GPG_ERR_TOO_LARGE);
+        }
+    }
+
+  return rc;
+}
+
+
+/*
+ * Write an opaque MPI to the output stream without length info.
+ */
+gpg_error_t
+gpg_mpi_write_nohdr (iobuf_t out, gcry_mpi_t a)
+{
+  int rc;
+
+  if (gcry_mpi_get_flag (a, GCRYMPI_FLAG_OPAQUE))
+    {
+      unsigned int nbits;
+      const void *p;
+
+      p = gcry_mpi_get_opaque (a, &nbits);
+      rc = p ? iobuf_write (out, p, (nbits+7)/8) : 0;
+    }
+  else
+    rc = gpg_error (GPG_ERR_BAD_MPI);
+
+  return rc;
+}
+
+
+/* Calculate the length of a packet described by PKT.  */
+u32
+calc_packet_length( PACKET *pkt )
+{
+  u32 n = 0;
+  int new_ctb = 0;
+
+  log_assert (pkt->pkt.generic);
+  switch (pkt->pkttype)
+    {
+    case PKT_PLAINTEXT:
+      n = calc_plaintext (pkt->pkt.plaintext);
+      new_ctb = pkt->pkt.plaintext->new_ctb;
+      break;
+    case PKT_ATTRIBUTE:
+    case PKT_USER_ID:
+    case PKT_COMMENT:
+    case PKT_PUBLIC_KEY:
+    case PKT_SECRET_KEY:
+    case PKT_SYMKEY_ENC:
+    case PKT_PUBKEY_ENC:
+    case PKT_ENCRYPTED:
+    case PKT_SIGNATURE:
+    case PKT_ONEPASS_SIG:
+    case PKT_RING_TRUST:
+    case PKT_COMPRESSED:
+    default:
+      log_bug ("invalid packet type in calc_packet_length()");
+      break;
+    }
+
+  n += calc_header_length (n, new_ctb);
+  return n;
+}
+
+
+static gpg_error_t
+write_fake_data (IOBUF out, gcry_mpi_t a)
+{
+  unsigned int n;
+  void *p;
+
+  if (!a)
+    return 0;
+  if (!gcry_mpi_get_flag (a, GCRYMPI_FLAG_OPAQUE))
+    return 0; /* e.g. due to generating a key with wrong usage.  */
+  p = gcry_mpi_get_opaque ( a, &n);
+  if (!p)
+    return 0; /* For example due to a read error in
+                 parse-packet.c:read_rest.  */
+  return iobuf_write (out, p, (n+7)/8 );
+}
+
+
+/* Write a ring trust meta packet.  */
+static gpg_error_t
+do_ring_trust (iobuf_t out, PKT_ring_trust *rt)
+{
+  unsigned int namelen = 0;
+  unsigned int pktlen = 6;
+
+  if (rt->subtype == RING_TRUST_KEY || rt->subtype == RING_TRUST_UID)
+    {
+      if (rt->url)
+        namelen = strlen (rt->url);
+      pktlen += 1 + 4 + 1 + namelen;
+    }
+
+  write_header (out, (0x80 | ((PKT_RING_TRUST & 15)<<2)), pktlen);
+  iobuf_put (out, rt->trustval);
+  iobuf_put (out, rt->sigcache);
+  iobuf_write (out, "gpg", 3);
+  iobuf_put (out, rt->subtype);
+  if (rt->subtype == RING_TRUST_KEY || rt->subtype == RING_TRUST_UID)
+    {
+      iobuf_put (out, rt->keyorg);
+      write_32 (out, rt->keyupdate);
+      iobuf_put (out, namelen);
+      if (namelen)
+        iobuf_write (out, rt->url, namelen);
+    }
+
+  return 0;
+}
+
+
+/* Serialize the user id (RFC 4880, Section 5.11) or the user
+ * attribute UID (Section 5.12) and write it to OUT.
+ *
+ * CTB is the serialization's CTB.  It specifies the header format and
+ * the packet's type.  The header length must not be set.  */
+static int
+do_user_id( IOBUF out, int ctb, PKT_user_id *uid )
+{
+  int rc;
+  int hdrlen;
+
+  log_assert (ctb_pkttype (ctb) == PKT_USER_ID
+              || ctb_pkttype (ctb) == PKT_ATTRIBUTE);
+
+  /* We need to take special care of a user ID with a length of 0:
+   * Without forcing HDRLEN to 2 in this case an indeterminate length
+   * packet would be written which is not allowed.  Note that we are
+   * always called with a CTB indicating an old packet header format,
+   * so that forcing a 2 octet header works.  We also check for the
+   * maximum allowed packet size by the parser using an arbitrary
+   * extra 10 bytes for header data. */
+  if (uid->attrib_data)
+    {
+      if (uid->attrib_len > MAX_ATTR_PACKET_LENGTH - 10)
+        return gpg_error (GPG_ERR_TOO_LARGE);
+      hdrlen = uid->attrib_len? 0 : 2;
+      write_header2 (out, ctb, uid->attrib_len, hdrlen);
+      rc = iobuf_write( out, uid->attrib_data, uid->attrib_len );
+    }
+  else
+    {
+      if (uid->len > MAX_UID_PACKET_LENGTH - 10)
+        return gpg_error (GPG_ERR_TOO_LARGE);
+      hdrlen = uid->len? 0 : 2;
+      write_header2 (out, ctb, uid->len, hdrlen);
+      rc = iobuf_write( out, uid->name, uid->len );
+    }
+  return rc;
+}
+
+
+/* Serialize the key (RFC 4880, Section 5.5) described by PK and write
+ * it to OUT.
+ *
+ * This function serializes both primary keys and subkeys with or
+ * without a secret part.
+ *
+ * CTB is the serialization's CTB.  It specifies the header format and
+ * the packet's type.  The header length must not be set.
+ *
+ * PK->VERSION specifies the serialization format.  A value of 0 means
+ * to use the default version.  Currently, only version 4 packets are
+ * supported.
+ */
+static int
+do_key (iobuf_t out, int ctb, PKT_public_key *pk)
+{
+  gpg_error_t err = 0;
+  /* The length of the body is stored in the packet's header, which
+     occurs before the body.  Unfortunately, we don't know the length
+     of the packet's body until we've written all of the data!  To
+     work around this, we first write the data into this temporary
+     buffer, then generate the header, and finally copy the contents
+     of this buffer to OUT.  */
+  iobuf_t a = iobuf_temp();
+  int i, nskey, npkey;
+
+  log_assert (pk->version == 0 || pk->version == 4);
+  log_assert (ctb_pkttype (ctb) == PKT_PUBLIC_KEY
+              || ctb_pkttype (ctb) == PKT_PUBLIC_SUBKEY
+              || ctb_pkttype (ctb) == PKT_SECRET_KEY
+              || ctb_pkttype (ctb) == PKT_SECRET_SUBKEY);
+
+  /* Write the version number - if none is specified, use 4 */
+  if ( !pk->version )
+    iobuf_put ( a, 4 );
+  else
+    iobuf_put ( a, pk->version );
+  write_32 (a, pk->timestamp );
+
+  iobuf_put (a, pk->pubkey_algo );
+
+  /* Get number of secret and public parameters.  They are held in one
+     array: the public ones followed by the secret ones.  */
+  nskey = pubkey_get_nskey (pk->pubkey_algo);
+  npkey = pubkey_get_npkey (pk->pubkey_algo);
+
+  /* If we don't have any public parameters - which is for example the
+     case if we don't know the algorithm used - the parameters are
+     stored as one blob in a faked (opaque) MPI. */
+  if (!npkey)
+    {
+      write_fake_data (a, pk->pkey[0]);
+      goto leave;
+    }
+  log_assert (npkey < nskey);
+
+  for (i=0; i < npkey; i++ )
+    {
+      if (   (pk->pubkey_algo == PUBKEY_ALGO_ECDSA && (i == 0))
+          || (pk->pubkey_algo == PUBKEY_ALGO_EDDSA && (i == 0))
+          || (pk->pubkey_algo == PUBKEY_ALGO_ECDH  && (i == 0 || i == 2)))
+        err = gpg_mpi_write_nohdr (a, pk->pkey[i]);
+      else
+        err = gpg_mpi_write (a, pk->pkey[i]);
+      if (err)
+        goto leave;
+    }
+
+
+  if (pk->seckey_info)
+    {
+      /* This is a secret key packet.  */
+      struct seckey_info *ski = pk->seckey_info;
+
+      /* Build the header for protected (encrypted) secret parameters.  */
+      if (ski->is_protected)
+        {
+          /* OpenPGP protection according to rfc2440. */
+          iobuf_put (a, ski->sha1chk? 0xfe : 0xff);
+          iobuf_put (a, ski->algo);
+          if (ski->s2k.mode >= 1000)
+            {
+              /* These modes are not possible in OpenPGP, we use them
+                 to implement our extensions, 101 can be viewed as a
+                 private/experimental extension (this is not specified
+                 in rfc2440 but the same scheme is used for all other
+                 algorithm identifiers). */
+              iobuf_put (a, 101);
+              iobuf_put (a, ski->s2k.hash_algo);
+              iobuf_write (a, "GNU", 3 );
+              iobuf_put (a, ski->s2k.mode - 1000);
+            }
+          else
+            {
+              iobuf_put (a, ski->s2k.mode);
+              iobuf_put (a, ski->s2k.hash_algo);
+            }
+
+          if (ski->s2k.mode == 1 || ski->s2k.mode == 3)
+            iobuf_write (a, ski->s2k.salt, 8);
+
+          if (ski->s2k.mode == 3)
+            iobuf_put (a, ski->s2k.count);
+
+          /* For our special modes 1001, 1002 we do not need an IV. */
+          if (ski->s2k.mode != 1001 && ski->s2k.mode != 1002)
+            iobuf_write (a, ski->iv, ski->ivlen);
+
+        }
+      else /* Not protected. */
+        iobuf_put (a, 0 );
+
+      if (ski->s2k.mode == 1001)
+        ; /* GnuPG extension - don't write a secret key at all. */
+      else if (ski->s2k.mode == 1002)
+        {
+          /* GnuPG extension - divert to OpenPGP smartcard. */
+          /* Length of the serial number or 0 for no serial number. */
+          iobuf_put (a, ski->ivlen );
+          /* The serial number gets stored in the IV field.  */
+          iobuf_write (a, ski->iv, ski->ivlen);
+        }
+      else if (ski->is_protected)
+        {
+          /* The secret key is protected - write it out as it is.  */
+          byte *p;
+          unsigned int ndatabits;
+
+          log_assert (gcry_mpi_get_flag (pk->pkey[npkey], GCRYMPI_FLAG_OPAQUE));
+          p = gcry_mpi_get_opaque (pk->pkey[npkey], &ndatabits);
+          if (p)
+            iobuf_write (a, p, (ndatabits+7)/8 );
+        }
+      else
+        {
+          /* Non-protected key. */
+          for ( ; i < nskey; i++ )
+            if ( (err = gpg_mpi_write (a, pk->pkey[i])))
+              goto leave;
+          write_16 (a, ski->csum );
+        }
+    }
+
+ leave:
+  if (!err)
+    {
+      /* Build the header of the packet - which we must do after
+         writing all the other stuff, so that we know the length of
+         the packet */
+      write_header2 (out, ctb, iobuf_get_temp_length(a), 0);
+       /* And finally write it out to the real stream. */
+      err = iobuf_write_temp (out, a);
+    }
+
+  iobuf_close (a); /* Close the temporary buffer */
+  return err;
+}
+
+
+/* Serialize the symmetric-key encrypted session key packet (RFC 4880,
+ * 5.3) described by ENC and write it to OUT.
+ *
+ * CTB is the serialization's CTB.  It specifies the header format and
+ * the packet's type.  The header length must not be set.  */
+static int
+do_symkey_enc( IOBUF out, int ctb, PKT_symkey_enc *enc )
+{
+  int rc = 0;
+  IOBUF a = iobuf_temp();
+
+  log_assert (ctb_pkttype (ctb) == PKT_SYMKEY_ENC);
+
+  /* The only acceptable version.  */
+  log_assert( enc->version == 4 );
+
+  /* RFC 4880, Section 3.7.  */
+  switch (enc->s2k.mode)
+    {
+    case 0: /* Simple S2K.  */
+    case 1: /* Salted S2K.  */
+    case 3: /* Iterated and salted S2K.  */
+      break; /* Reasonable values.  */
+
+    default:
+      log_bug ("do_symkey_enc: s2k=%d\n", enc->s2k.mode);
+    }
+    iobuf_put( a, enc->version );
+    iobuf_put( a, enc->cipher_algo );
+    iobuf_put( a, enc->s2k.mode );
+    iobuf_put( a, enc->s2k.hash_algo );
+    if( enc->s2k.mode == 1 || enc->s2k.mode == 3 ) {
+	iobuf_write(a, enc->s2k.salt, 8 );
+	if( enc->s2k.mode == 3 )
+	    iobuf_put(a, enc->s2k.count);
+    }
+    if( enc->seskeylen )
+	iobuf_write(a, enc->seskey, enc->seskeylen );
+
+    write_header(out, ctb, iobuf_get_temp_length(a) );
+    rc = iobuf_write_temp( out, a );
+
+    iobuf_close(a);
+    return rc;
+}
+
+
+/* Serialize the public-key encrypted session key packet (RFC 4880,
+   5.1) described by ENC and write it to OUT.
+
+   CTB is the serialization's CTB.  It specifies the header format and
+   the packet's type.  The header length must not be set.  */
+static int
+do_pubkey_enc( IOBUF out, int ctb, PKT_pubkey_enc *enc )
+{
+  int rc = 0;
+  int n, i;
+  IOBUF a = iobuf_temp();
+
+  log_assert (ctb_pkttype (ctb) == PKT_PUBKEY_ENC);
+
+  iobuf_put (a, 3); /* Version.  */
+
+  if ( enc->throw_keyid )
+    {
+      write_32(a, 0 );  /* Don't tell Eve who can decrypt the message.  */
+      write_32(a, 0 );
+    }
+  else
+    {
+      write_32(a, enc->keyid[0] );
+      write_32(a, enc->keyid[1] );
+    }
+  iobuf_put(a,enc->pubkey_algo );
+  n = pubkey_get_nenc( enc->pubkey_algo );
+  if ( !n )
+    write_fake_data( a, enc->data[0] );
+
+  for (i=0; i < n && !rc ; i++ )
+    {
+      if (enc->pubkey_algo == PUBKEY_ALGO_ECDH && i == 1)
+        rc = gpg_mpi_write_nohdr (a, enc->data[i]);
+      else
+        rc = gpg_mpi_write (a, enc->data[i]);
+    }
+
+  if (!rc)
+    {
+      write_header (out, ctb, iobuf_get_temp_length(a) );
+      rc = iobuf_write_temp (out, a);
+    }
+  iobuf_close(a);
+  return rc;
+}
+
+
+/* Calculate the length of the serialized plaintext packet PT (RFC
+   4480, Section 5.9).  */
+static u32
+calc_plaintext( PKT_plaintext *pt )
+{
+  /* Truncate namelen to the maximum 255 characters.  Note this means
+     that a function that calls build_packet with an illegal literal
+     packet will get it back legalized. */
+
+  if(pt->namelen>255)
+    pt->namelen=255;
+
+  return pt->len? (1 + 1 + pt->namelen + 4 + pt->len) : 0;
+}
+
+/* Serialize the plaintext packet (RFC 4880, 5.9) described by PT and
+   write it to OUT.
+
+   The body of the message is stored in PT->BUF.  The amount of data
+   to write is PT->LEN.  (PT->BUF should be configured to return EOF
+   after this much data has been read.)  If PT->LEN is 0 and CTB
+   indicates that this is a new format packet, then partial block mode
+   is assumed to have been enabled on OUT.  On success, partial block
+   mode is disabled.
+
+   If PT->BUF is NULL, the caller must write out the data.  In
+   this case, if PT->LEN was 0, then partial body length mode was
+   enabled and the caller must disable it by calling
+   iobuf_set_partial_body_length_mode (out, 0).  */
+static int
+do_plaintext( IOBUF out, int ctb, PKT_plaintext *pt )
+{
+    int rc = 0;
+    size_t nbytes;
+
+    log_assert (ctb_pkttype (ctb) == PKT_PLAINTEXT);
+
+    write_header(out, ctb, calc_plaintext( pt ) );
+    log_assert (pt->mode == 'b' || pt->mode == 't' || pt->mode == 'u'
+                || pt->mode == 'm'
+                || pt->mode == 'l' || pt->mode == '1');
+    iobuf_put(out, pt->mode );
+    iobuf_put(out, pt->namelen );
+    iobuf_write (out, pt->name, pt->namelen);
+    rc = write_32(out, pt->timestamp );
+    if (rc)
+      return rc;
+
+    if (pt->buf)
+      {
+        nbytes = iobuf_copy (out, pt->buf);
+        if (nbytes == (size_t)(-1)
+            && (iobuf_error (out) || iobuf_error (pt->buf)))
+            return iobuf_error (out)? iobuf_error (out):iobuf_error (pt->buf);
+        if(ctb_new_format_p (ctb) && !pt->len)
+          /* Turn off partial body length mode.  */
+          iobuf_set_partial_body_length_mode (out, 0);
+        if( pt->len && nbytes != pt->len )
+          log_error("do_plaintext(): wrote %lu bytes but expected %lu bytes\n",
+                    (ulong)nbytes, (ulong)pt->len );
+      }
+
+    return rc;
+}
+
+
+
+/* Serialize the symmetrically encrypted data packet (RFC 4880,
+   Section 5.7) described by ED and write it to OUT.
+
+   Note: this only writes the packets header!  The call must then
+   follow up and write the initial random data and the body to OUT.
+   (If you use the encryption iobuf filter (cipher_filter), then this
+   is done automatically.)  */
+static int
+do_encrypted( IOBUF out, int ctb, PKT_encrypted *ed )
+{
+    int rc = 0;
+    u32 n;
+
+    log_assert (! ed->mdc_method);
+    log_assert (ctb_pkttype (ctb) == PKT_ENCRYPTED);
+
+    n = ed->len ? (ed->len + ed->extralen) : 0;
+    write_header(out, ctb, n );
+
+    /* This is all. The caller has to write the real data */
+
+    return rc;
+}
+
+/* Serialize the symmetrically encrypted integrity protected data
+   packet (RFC 4880, Section 5.13) described by ED and write it to
+   OUT.
+
+   Note: this only writes the packet's header!  The caller must then
+   follow up and write the initial random data, the body and the MDC
+   packet to OUT.  (If you use the encryption iobuf filter
+   (cipher_filter), then this is done automatically.)  */
+static int
+do_encrypted_mdc( IOBUF out, int ctb, PKT_encrypted *ed )
+{
+    int rc = 0;
+    u32 n;
+
+    log_assert (ed->mdc_method);
+    log_assert (ctb_pkttype (ctb) == PKT_ENCRYPTED_MDC);
+
+    /* Take version number and the following MDC packet in account. */
+    n = ed->len ? (ed->len + ed->extralen + 1 + 22) : 0;
+    write_header(out, ctb, n );
+    iobuf_put(out, 1 );  /* version */
+
+    /* This is all. The caller has to write the real data */
+
+    return rc;
+}
+
+
+/* Serialize the compressed packet (RFC 4880, Section 5.6) described
+   by CD and write it to OUT.
+
+   Note: this only writes the packet's header!  The caller must then
+   follow up and write the body to OUT.  */
+static int
+do_compressed( IOBUF out, int ctb, PKT_compressed *cd )
+{
+    int rc = 0;
+
+    log_assert (ctb_pkttype (ctb) == PKT_COMPRESSED);
+
+    /* We must use the old convention and don't use blockmode for the
+       sake of PGP 2 compatibility.  However if the new_ctb flag was
+       set, CTB is already formatted as new style and write_header2
+       does create a partial length encoding using new the new
+       style. */
+    write_header2(out, ctb, 0, 0);
+    iobuf_put(out, cd->algorithm );
+
+    /* This is all. The caller has to write the real data */
+
+    return rc;
+}
+
+
+/****************
+ * Delete all subpackets of type REQTYPE and return a bool whether a packet
+ * was deleted.
+ */
+int
+delete_sig_subpkt (subpktarea_t *area, sigsubpkttype_t reqtype )
+{
+    int buflen;
+    sigsubpkttype_t type;
+    byte *buffer, *bufstart;
+    size_t n;
+    size_t unused = 0;
+    int okay = 0;
+
+    if( !area )
+	return 0;
+    buflen = area->len;
+    buffer = area->data;
+    for(;;) {
+	if( !buflen ) {
+            okay = 1;
+            break;
+        }
+	bufstart = buffer;
+	n = *buffer++; buflen--;
+	if( n == 255 ) {
+	    if( buflen < 4 )
+		break;
+	    n = buf32_to_size_t (buffer);
+	    buffer += 4;
+	    buflen -= 4;
+	}
+	else if( n >= 192 ) {
+	    if( buflen < 2 )
+		break;
+	    n = (( n - 192 ) << 8) + *buffer + 192;
+	    buffer++;
+	    buflen--;
+	}
+	if( buflen < n )
+	    break;
+
+	type = *buffer & 0x7f;
+	if( type == reqtype ) {
+	    buffer++;
+            buflen--;
+	    n--;
+	    if( n > buflen )
+		break;
+            buffer += n; /* point to next subpkt */
+            buflen -= n;
+            memmove (bufstart, buffer, buflen); /* shift */
+            unused +=  buffer - bufstart;
+            buffer = bufstart;
+	}
+        else {
+            buffer += n; buflen -=n;
+        }
+    }
+
+    if (!okay)
+        log_error ("delete_subpkt: buffer shorter than subpacket\n");
+    log_assert (unused <= area->len);
+    area->len -= unused;
+    return !!unused;
+}
+
+
+/****************
+ * Create or update a signature subpacket for SIG of TYPE.  This
+ * functions knows where to put the data (hashed or unhashed).  The
+ * function may move data from the unhashed part to the hashed one.
+ * Note: All pointers into sig->[un]hashed (e.g. returned by
+ * parse_sig_subpkt) are not valid after a call to this function.  The
+ * data to put into the subpaket should be in a buffer with a length
+ * of buflen.
+ */
+void
+build_sig_subpkt (PKT_signature *sig, sigsubpkttype_t type,
+		  const byte *buffer, size_t buflen )
+{
+    byte *p;
+    int critical, hashed;
+    subpktarea_t *oldarea, *newarea;
+    size_t nlen, n, n0;
+
+    critical = (type & SIGSUBPKT_FLAG_CRITICAL);
+    type &= ~SIGSUBPKT_FLAG_CRITICAL;
+
+    /* Sanity check buffer sizes */
+    if(parse_one_sig_subpkt(buffer,buflen,type)<0)
+      BUG();
+
+    switch(type)
+      {
+      case SIGSUBPKT_NOTATION:
+      case SIGSUBPKT_POLICY:
+      case SIGSUBPKT_REV_KEY:
+      case SIGSUBPKT_SIGNATURE:
+	/* we do allow multiple subpackets */
+	break;
+
+      default:
+	/* we don't allow multiple subpackets */
+	delete_sig_subpkt(sig->hashed,type);
+	delete_sig_subpkt(sig->unhashed,type);
+	break;
+      }
+
+    /* Any special magic that needs to be done for this type so the
+       packet doesn't need to be reparsed? */
+    switch(type)
+      {
+      case SIGSUBPKT_NOTATION:
+	sig->flags.notation=1;
+	break;
+
+      case SIGSUBPKT_POLICY:
+	sig->flags.policy_url=1;
+	break;
+
+      case SIGSUBPKT_PREF_KS:
+	sig->flags.pref_ks=1;
+	break;
+
+      case SIGSUBPKT_EXPORTABLE:
+	if(buffer[0])
+	  sig->flags.exportable=1;
+	else
+	  sig->flags.exportable=0;
+	break;
+
+      case SIGSUBPKT_REVOCABLE:
+	if(buffer[0])
+	  sig->flags.revocable=1;
+	else
+	  sig->flags.revocable=0;
+	break;
+
+      case SIGSUBPKT_TRUST:
+	sig->trust_depth=buffer[0];
+	sig->trust_value=buffer[1];
+	break;
+
+      case SIGSUBPKT_REGEXP:
+	sig->trust_regexp=buffer;
+	break;
+
+	/* This should never happen since we don't currently allow
+	   creating such a subpacket, but just in case... */
+      case SIGSUBPKT_SIG_EXPIRE:
+	if(buf32_to_u32(buffer)+sig->timestamp<=make_timestamp())
+	  sig->flags.expired=1;
+	else
+	  sig->flags.expired=0;
+	break;
+
+      default:
+	break;
+      }
+
+    if( (buflen+1) >= 8384 )
+	nlen = 5; /* write 5 byte length header */
+    else if( (buflen+1) >= 192 )
+	nlen = 2; /* write 2 byte length header */
+    else
+	nlen = 1; /* just a 1 byte length header */
+
+    switch( type )
+      {
+	/* The issuer being unhashed is a historical oddity.  It
+	   should work equally as well hashed.  Of course, if even an
+	   unhashed issuer is tampered with, it makes it awfully hard
+	   to verify the sig... */
+      case SIGSUBPKT_ISSUER:
+      case SIGSUBPKT_SIGNATURE:
+        hashed = 0;
+        break;
+      default:
+        hashed = 1;
+        break;
+      }
+
+    if( critical )
+	type |= SIGSUBPKT_FLAG_CRITICAL;
+
+    oldarea = hashed? sig->hashed : sig->unhashed;
+
+    /* Calculate new size of the area and allocate */
+    n0 = oldarea? oldarea->len : 0;
+    n = n0 + nlen + 1 + buflen; /* length, type, buffer */
+    if (oldarea && n <= oldarea->size) { /* fits into the unused space */
+        newarea = oldarea;
+        /*log_debug ("updating area for type %d\n", type );*/
+    }
+    else if (oldarea) {
+        newarea = xrealloc (oldarea, sizeof (*newarea) + n - 1);
+        newarea->size = n;
+        /*log_debug ("reallocating area for type %d\n", type );*/
+    }
+    else {
+        newarea = xmalloc (sizeof (*newarea) + n - 1);
+        newarea->size = n;
+        /*log_debug ("allocating area for type %d\n", type );*/
+    }
+    newarea->len = n;
+
+    p = newarea->data + n0;
+    if (nlen == 5) {
+	*p++ = 255;
+	*p++ = (buflen+1) >> 24;
+	*p++ = (buflen+1) >> 16;
+	*p++ = (buflen+1) >>  8;
+	*p++ = (buflen+1);
+	*p++ = type;
+	memcpy (p, buffer, buflen);
+    }
+    else if (nlen == 2) {
+	*p++ = (buflen+1-192) / 256 + 192;
+	*p++ = (buflen+1-192) % 256;
+	*p++ = type;
+	memcpy (p, buffer, buflen);
+    }
+    else {
+	*p++ = buflen+1;
+	*p++ = type;
+	memcpy (p, buffer, buflen);
+    }
+
+    if (hashed)
+	sig->hashed = newarea;
+    else
+	sig->unhashed = newarea;
+}
+
+/*
+ * Put all the required stuff from SIG into subpackets of sig.
+ * PKSK is the signing key.
+ * Hmmm, should we delete those subpackets which are in a wrong area?
+ */
+void
+build_sig_subpkt_from_sig (PKT_signature *sig, PKT_public_key *pksk)
+{
+    u32  u;
+    byte buf[1+MAX_FINGERPRINT_LEN];
+    size_t fprlen;
+
+    /* For v4 keys we need to write the ISSUER subpacket.  We do not
+     * want that for a future v5 format.  */
+    if (pksk->version < 5)
+      {
+        u = sig->keyid[0];
+        buf[0] = (u >> 24) & 0xff;
+        buf[1] = (u >> 16) & 0xff;
+        buf[2] = (u >>  8) & 0xff;
+        buf[3] = u & 0xff;
+        u = sig->keyid[1];
+        buf[4] = (u >> 24) & 0xff;
+        buf[5] = (u >> 16) & 0xff;
+        buf[6] = (u >>  8) & 0xff;
+        buf[7] = u & 0xff;
+        build_sig_subpkt (sig, SIGSUBPKT_ISSUER, buf, 8);
+      }
+
+    /* Write the new ISSUER_FPR subpacket.  */
+    fingerprint_from_pk (pksk, buf+1, &fprlen);
+    if (fprlen == 20)
+      {
+        buf[0] = pksk->version;
+        build_sig_subpkt (sig, SIGSUBPKT_ISSUER_FPR, buf, 21);
+      }
+
+    /* Write the timestamp.  */
+    u = sig->timestamp;
+    buf[0] = (u >> 24) & 0xff;
+    buf[1] = (u >> 16) & 0xff;
+    buf[2] = (u >>  8) & 0xff;
+    buf[3] = u & 0xff;
+    build_sig_subpkt( sig, SIGSUBPKT_SIG_CREATED, buf, 4 );
+
+    if(sig->expiredate)
+      {
+	if(sig->expiredate>sig->timestamp)
+	  u=sig->expiredate-sig->timestamp;
+	else
+	  u=1; /* A 1-second expiration time is the shortest one
+		  OpenPGP has */
+
+	buf[0] = (u >> 24) & 0xff;
+	buf[1] = (u >> 16) & 0xff;
+	buf[2] = (u >>  8) & 0xff;
+	buf[3] = u & 0xff;
+
+	/* Mark this CRITICAL, so if any implementation doesn't
+           understand sigs that can expire, it'll just disregard this
+           sig altogether. */
+
+	build_sig_subpkt( sig, SIGSUBPKT_SIG_EXPIRE | SIGSUBPKT_FLAG_CRITICAL,
+			  buf, 4 );
+      }
+}
+
+void
+build_attribute_subpkt(PKT_user_id *uid,byte type,
+		       const void *buf,u32 buflen,
+		       const void *header,u32 headerlen)
+{
+  byte *attrib;
+  int idx;
+
+  if(1+headerlen+buflen>8383)
+    idx=5;
+  else if(1+headerlen+buflen>191)
+    idx=2;
+  else
+    idx=1;
+
+  /* realloc uid->attrib_data to the right size */
+
+  uid->attrib_data=xrealloc(uid->attrib_data,
+			     uid->attrib_len+idx+1+headerlen+buflen);
+
+  attrib=&uid->attrib_data[uid->attrib_len];
+
+  if(idx==5)
+    {
+      attrib[0]=255;
+      attrib[1]=(1+headerlen+buflen) >> 24;
+      attrib[2]=(1+headerlen+buflen) >> 16;
+      attrib[3]=(1+headerlen+buflen) >> 8;
+      attrib[4]=1+headerlen+buflen;
+    }
+  else if(idx==2)
+    {
+      attrib[0]=(1+headerlen+buflen-192) / 256 + 192;
+      attrib[1]=(1+headerlen+buflen-192) % 256;
+    }
+  else
+    attrib[0]=1+headerlen+buflen; /* Good luck finding a JPEG this small! */
+
+  attrib[idx++]=type;
+
+  /* Tack on our data at the end */
+
+  if(headerlen>0)
+    memcpy(&attrib[idx],header,headerlen);
+  memcpy(&attrib[idx+headerlen],buf,buflen);
+  uid->attrib_len+=idx+headerlen+buflen;
+}
+
+/* Returns a human-readable string corresponding to the notation.
+   This ignores notation->value.  The caller must free the result.  */
+static char *
+notation_value_to_human_readable_string (struct notation *notation)
+{
+  if(notation->bdat)
+    /* Binary data.  */
+    {
+      size_t len = notation->blen;
+      int i;
+      char preview[20];
+
+      for (i = 0; i < len && i < sizeof (preview) - 1; i ++)
+        if (isprint (notation->bdat[i]))
+          preview[i] = notation->bdat[i];
+        else
+          preview[i] = '?';
+      preview[i] = 0;
+
+      return xasprintf (_("[ not human readable (%zu bytes: %s%s) ]"),
+                        len, preview, i < len ? "..." : "");
+    }
+  else
+    /* The value is human-readable.  */
+    return xstrdup (notation->value);
+}
+
+/* Turn the notation described by the string STRING into a notation.
+
+   STRING has the form:
+
+     - -name - Delete the notation.
+     - name@domain.name=value - Normal notation
+     - !name@domain.name=value - Notation with critical bit set.
+
+   The caller must free the result using free_notation().  */
+struct notation *
+string_to_notation(const char *string,int is_utf8)
+{
+  const char *s;
+  int saw_at=0;
+  struct notation *notation;
+
+  notation=xmalloc_clear(sizeof(*notation));
+
+  if(*string=='-')
+    {
+      notation->flags.ignore=1;
+      string++;
+    }
+
+  if(*string=='!')
+    {
+      notation->flags.critical=1;
+      string++;
+    }
+
+  /* If and when the IETF assigns some official name tags, we'll have
+     to add them here. */
+
+  for( s=string ; *s != '='; s++ )
+    {
+      if( *s=='@')
+	saw_at++;
+
+      /* -notationname is legal without an = sign */
+      if(!*s && notation->flags.ignore)
+	break;
+
+      if( !*s || !isascii (*s) || (!isgraph(*s) && !isspace(*s)) )
+	{
+	  log_error(_("a notation name must have only printable characters"
+		      " or spaces, and end with an '='\n") );
+	  goto fail;
+	}
+    }
+
+  notation->name=xmalloc((s-string)+1);
+  memcpy(notation->name,string,s-string);
+  notation->name[s-string]='\0';
+
+  if(!saw_at && !opt.expert)
+    {
+      log_error(_("a user notation name must contain the '@' character\n"));
+      goto fail;
+    }
+
+  if (saw_at > 1)
+    {
+      log_error(_("a notation name must not contain more than"
+		  " one '@' character\n"));
+      goto fail;
+    }
+
+  if(*s)
+    {
+      const char *i=s+1;
+      int highbit=0;
+
+      /* we only support printable text - therefore we enforce the use
+	 of only printable characters (an empty value is valid) */
+      for(s++; *s ; s++ )
+	{
+	  if ( !isascii (*s) )
+	    highbit=1;
+	  else if (iscntrl(*s))
+	    {
+	      log_error(_("a notation value must not use any"
+			  " control characters\n"));
+	      goto fail;
+	    }
+	}
+
+      if(!highbit || is_utf8)
+	notation->value=xstrdup(i);
+      else
+	notation->value=native_to_utf8(i);
+    }
+
+  return notation;
+
+ fail:
+  free_notation(notation);
+  return NULL;
+}
+
+/* Like string_to_notation, but store opaque data rather than human
+   readable data.  */
+struct notation *
+blob_to_notation(const char *name, const char *data, size_t len)
+{
+  const char *s;
+  int saw_at=0;
+  struct notation *notation;
+
+  notation=xmalloc_clear(sizeof(*notation));
+
+  if(*name=='-')
+    {
+      notation->flags.ignore=1;
+      name++;
+    }
+
+  if(*name=='!')
+    {
+      notation->flags.critical=1;
+      name++;
+    }
+
+  /* If and when the IETF assigns some official name tags, we'll have
+     to add them here. */
+
+  for( s=name ; *s; s++ )
+    {
+      if( *s=='@')
+	saw_at++;
+
+      /* -notationname is legal without an = sign */
+      if(!*s && notation->flags.ignore)
+	break;
+
+      if (*s == '=')
+        {
+          log_error(_("a notation name may not contain an '=' character\n"));
+          goto fail;
+        }
+
+      if (!isascii (*s) || (!isgraph(*s) && !isspace(*s)))
+	{
+	  log_error(_("a notation name must have only printable characters"
+		      " or spaces\n") );
+	  goto fail;
+	}
+    }
+
+  notation->name=xstrdup (name);
+
+  if(!saw_at && !opt.expert)
+    {
+      log_error(_("a user notation name must contain the '@' character\n"));
+      goto fail;
+    }
+
+  if (saw_at > 1)
+    {
+      log_error(_("a notation name must not contain more than"
+		  " one '@' character\n"));
+      goto fail;
+    }
+
+  notation->bdat = xmalloc (len);
+  memcpy (notation->bdat, data, len);
+  notation->blen = len;
+
+  notation->value = notation_value_to_human_readable_string (notation);
+
+  return notation;
+
+ fail:
+  free_notation(notation);
+  return NULL;
+}
+
+struct notation *
+sig_to_notation(PKT_signature *sig)
+{
+  const byte *p;
+  size_t len;
+  int seq = 0;
+  int crit;
+  notation_t list = NULL;
+
+  /* See RFC 4880, 5.2.3.16 for the format of notation data.  In
+     short, a notation has:
+
+       - 4 bytes of flags
+       - 2 byte name length (n1)
+       - 2 byte value length (n2)
+       - n1 bytes of name data
+       - n2 bytes of value data
+   */
+  while((p=enum_sig_subpkt(sig->hashed,SIGSUBPKT_NOTATION,&len,&seq,&crit)))
+    {
+      int n1,n2;
+      struct notation *n=NULL;
+
+      if(len<8)
+	{
+	  log_info(_("WARNING: invalid notation data found\n"));
+	  continue;
+	}
+
+      /* name length.  */
+      n1=(p[4]<<8)|p[5];
+      /* value length.  */
+      n2=(p[6]<<8)|p[7];
+
+      if(8+n1+n2!=len)
+	{
+	  log_info(_("WARNING: invalid notation data found\n"));
+	  continue;
+	}
+
+      n=xmalloc_clear(sizeof(*n));
+      n->name=xmalloc(n1+1);
+
+      memcpy(n->name,&p[8],n1);
+      n->name[n1]='\0';
+
+      if(p[0]&0x80)
+        /* The value is human-readable.  */
+	{
+	  n->value=xmalloc(n2+1);
+	  memcpy(n->value,&p[8+n1],n2);
+	  n->value[n2]='\0';
+          n->flags.human = 1;
+	}
+      else
+        /* Binary data.  */
+	{
+	  n->bdat=xmalloc(n2);
+	  n->blen=n2;
+	  memcpy(n->bdat,&p[8+n1],n2);
+
+          n->value = notation_value_to_human_readable_string (n);
+	}
+
+      n->flags.critical=crit;
+
+      n->next=list;
+      list=n;
+    }
+
+  return list;
+}
+
+/* Release the resources associated with the *list* of notations.  To
+   release a single notation, make sure that notation->next is
+   NULL.  */
+void
+free_notation(struct notation *notation)
+{
+  while(notation)
+    {
+      struct notation *n=notation;
+
+      xfree(n->name);
+      xfree(n->value);
+      xfree(n->altvalue);
+      xfree(n->bdat);
+      notation=n->next;
+      xfree(n);
+    }
+}
+
+/* Serialize the signature packet (RFC 4880, Section 5.2) described by
+   SIG and write it to OUT.  */
+static int
+do_signature( IOBUF out, int ctb, PKT_signature *sig )
+{
+  int rc = 0;
+  int n, i;
+  IOBUF a = iobuf_temp();
+
+  log_assert (ctb_pkttype (ctb) == PKT_SIGNATURE);
+
+  if ( !sig->version || sig->version == 3)
+    {
+      iobuf_put( a, 3 );
+
+      /* Version 3 packets don't support subpackets.  Actually we
+       * should never get to here but real life is different and thus
+       * we now use a log_fatal instead of a log_assert here. */
+      if (sig->hashed || sig->unhashed)
+        log_fatal ("trying to write a subpacket to a v3 signature (%d,%d)\n",
+                   !!sig->hashed, !!sig->unhashed);
+    }
+  else
+    iobuf_put( a, sig->version );
+  if ( sig->version < 4 )
+    iobuf_put (a, 5 ); /* Constant */
+  iobuf_put (a, sig->sig_class );
+  if ( sig->version < 4 )
+    {
+      write_32(a, sig->timestamp );
+      write_32(a, sig->keyid[0] );
+      write_32(a, sig->keyid[1] );
+    }
+  iobuf_put(a, sig->pubkey_algo );
+  iobuf_put(a, sig->digest_algo );
+  if ( sig->version >= 4 )
+    {
+      size_t nn;
+      /* Timestamp and keyid must have been packed into the subpackets
+	 prior to the call of this function, because these subpackets
+	 are hashed. */
+      nn = sig->hashed? sig->hashed->len : 0;
+      write_16(a, nn);
+      if (nn)
+        iobuf_write( a, sig->hashed->data, nn );
+      nn = sig->unhashed? sig->unhashed->len : 0;
+      write_16(a, nn);
+      if (nn)
+        iobuf_write( a, sig->unhashed->data, nn );
+    }
+  iobuf_put(a, sig->digest_start[0] );
+  iobuf_put(a, sig->digest_start[1] );
+  n = pubkey_get_nsig( sig->pubkey_algo );
+  if ( !n )
+    write_fake_data( a, sig->data[0] );
+  for (i=0; i < n && !rc ; i++ )
+    rc = gpg_mpi_write (a, sig->data[i] );
+
+  if (!rc)
+    {
+      if ( is_RSA(sig->pubkey_algo) && sig->version < 4 )
+        write_sign_packet_header(out, ctb, iobuf_get_temp_length(a) );
+      else
+        write_header(out, ctb, iobuf_get_temp_length(a) );
+      rc = iobuf_write_temp( out, a );
+    }
+
+  iobuf_close(a);
+  return rc;
+}
+
+
+/* Serialize the one-pass signature packet (RFC 4880, Section 5.4)
+   described by OPS and write it to OUT.  */
+static int
+do_onepass_sig( IOBUF out, int ctb, PKT_onepass_sig *ops )
+{
+    log_assert (ctb_pkttype (ctb) == PKT_ONEPASS_SIG);
+
+    write_header(out, ctb, 4 + 8 + 1);
+
+    iobuf_put (out, 3);  /* Version.  */
+    iobuf_put(out, ops->sig_class );
+    iobuf_put(out, ops->digest_algo );
+    iobuf_put(out, ops->pubkey_algo );
+    write_32(out, ops->keyid[0] );
+    write_32(out, ops->keyid[1] );
+    iobuf_put(out, ops->last );
+
+    return 0;
+}
+
+
+/* Write a 16-bit quantity to OUT in big endian order.  */
+static int
+write_16(IOBUF out, u16 a)
+{
+    iobuf_put(out, a>>8);
+    if( iobuf_put(out,a) )
+	return -1;
+    return 0;
+}
+
+/* Write a 32-bit quantity to OUT in big endian order.  */
+static int
+write_32(IOBUF out, u32 a)
+{
+    iobuf_put(out, a>> 24);
+    iobuf_put(out, a>> 16);
+    iobuf_put(out, a>> 8);
+    return iobuf_put(out, a);
+}
+
+
+/****************
+ * calculate the length of a header.
+ *
+ * LEN is the length of the packet's body.  NEW_CTB is whether we are
+ * using a new or old format packet.
+ *
+ * This function does not handle indeterminate lengths or partial body
+ * lengths.  (If you pass LEN as 0, then this function assumes you
+ * really mean an empty body.)
+ */
+static int
+calc_header_length( u32 len, int new_ctb )
+{
+    if( new_ctb ) {
+	if( len < 192 )
+	    return 2;
+	if( len < 8384 )
+	    return 3;
+	else
+	    return 6;
+    }
+    if( len < 256 )
+	return 2;
+    if( len < 65536 )
+	return 3;
+
+    return 5;
+}
+
+/****************
+ * Write the CTB and the packet length
+ */
+static int
+write_header( IOBUF out, int ctb, u32 len )
+{
+    return write_header2( out, ctb, len, 0 );
+}
+
+
+static int
+write_sign_packet_header (IOBUF out, int ctb, u32 len)
+{
+  (void)ctb;
+
+  /* Work around a bug in the pgp read function for signature packets,
+     which are not correctly coded and silently assume at some point 2
+     byte length headers.*/
+  iobuf_put (out, 0x89 );
+  iobuf_put (out, len >> 8 );
+  return iobuf_put (out, len) == -1 ? -1:0;
+}
+
+/****************
+ * Write a packet header to OUT.
+ *
+ * CTB is the ctb.  It determines whether a new or old format packet
+ * header should be written.  The length field is adjusted, but the
+ * CTB is otherwise written out as is.
+ *
+ * LEN is the length of the packet's body.
+ *
+ * If HDRLEN is set, then we don't necessarily use the most efficient
+ * encoding to store LEN, but the specified length.  (If this is not
+ * possible, this is a bug.)  In this case, LEN=0 means a 0 length
+ * packet.  Note: setting HDRLEN is only supported for old format
+ * packets!
+ *
+ * If HDRLEN is not set, then the shortest encoding is used.  In this
+ * case, LEN=0 means the body has an indeterminate length and a
+ * partial body length header (if a new format packet) or an
+ * indeterminate length header (if an old format packet) is written
+ * out.  Further, if using partial body lengths, this enables partial
+ * body length mode on OUT.
+ */
+static int
+write_header2( IOBUF out, int ctb, u32 len, int hdrlen )
+{
+  if (ctb_new_format_p (ctb))
+    return write_new_header( out, ctb, len, hdrlen );
+
+  /* An old format packet.  Refer to RFC 4880, Section 4.2.1 to
+     understand how lengths are encoded in this case.  */
+
+  /* The length encoding is stored in the two least significant bits.
+     Make sure they are cleared.  */
+  log_assert ((ctb & 3) == 0);
+
+  log_assert (hdrlen == 0 || hdrlen == 2 || hdrlen == 3 || hdrlen == 5);
+
+  if (hdrlen)
+    /* Header length is given.  */
+    {
+      if( hdrlen == 2 && len < 256 )
+        /* 00 => 1 byte length.  */
+	;
+      else if( hdrlen == 3 && len < 65536 )
+        /* 01 => 2 byte length.  If len < 256, this is not the most
+           compact encoding, but it is a correct encoding.  */
+	ctb |= 1;
+      else if (hdrlen == 5)
+        /* 10 => 4 byte length.  If len < 65536, this is not the most
+           compact encoding, but it is a correct encoding.  */
+	ctb |= 2;
+      else
+        log_bug ("Can't encode length=%d in a %d byte header!\n",
+                 len, hdrlen);
+    }
+  else
+    {
+      if( !len )
+        /* 11 => Indeterminate length.  */
+	ctb |= 3;
+      else if( len < 256 )
+        /* 00 => 1 byte length.  */
+	;
+      else if( len < 65536 )
+        /* 01 => 2 byte length.  */
+	ctb |= 1;
+      else
+        /* 10 => 4 byte length.  */
+	ctb |= 2;
+    }
+
+  if( iobuf_put(out, ctb ) )
+    return -1;
+
+  if( len || hdrlen )
+    {
+      if( ctb & 2 )
+	{
+	  if(iobuf_put(out, len >> 24 ))
+	    return -1;
+	  if(iobuf_put(out, len >> 16 ))
+	    return -1;
+	}
+
+      if( ctb & 3 )
+	if(iobuf_put(out, len >> 8 ))
+	  return -1;
+
+      if( iobuf_put(out, len ) )
+	return -1;
+    }
+
+  return 0;
+}
+
+
+/* Write a new format header to OUT.
+
+   CTB is the ctb.
+
+   LEN is the length of the packet's body.  If LEN is 0, then enables
+   partial body length mode (i.e., the body is of an indeterminant
+   length) on OUT.  Note: this function cannot be used to generate a
+   header for a zero length packet.
+
+   HDRLEN is the length of the packet's header.  If HDRLEN is 0, the
+   shortest encoding is chosen based on the length of the packet's
+   body.  Currently, values other than 0 are not supported.
+
+   Returns 0 on success.  */
+static int
+write_new_header( IOBUF out, int ctb, u32 len, int hdrlen )
+{
+    if( hdrlen )
+	log_bug("can't cope with hdrlen yet\n");
+
+    if( iobuf_put(out, ctb ) )
+	return -1;
+    if( !len ) {
+	iobuf_set_partial_body_length_mode(out, 512 );
+    }
+    else {
+	if( len < 192 ) {
+	    if( iobuf_put(out, len ) )
+		return -1;
+	}
+	else if( len < 8384 ) {
+	    len -= 192;
+	    if( iobuf_put( out, (len / 256) + 192) )
+		return -1;
+	    if( iobuf_put( out, (len % 256) )  )
+		return -1;
+	}
+	else {
+	    if( iobuf_put( out, 0xff ) )
+		return -1;
+	    if( iobuf_put( out, (len >> 24)&0xff ) )
+		return -1;
+	    if( iobuf_put( out, (len >> 16)&0xff ) )
+		return -1;
+	    if( iobuf_put( out, (len >> 8)&0xff )  )
+		return -1;
+	    if( iobuf_put( out, len & 0xff ) )
+		return -1;
+	}
+    }
+    return 0;
+}