msg_parser.c

sip协议栈

 * @relates msg_s
 *
 * When parsing a transport stream, only the first message in the stream is
 * created with msg_create(). The rest of the messages should be created
 * with msg_next() after previous message has been completely received and
 * parsed.
 *
 */
msg_t *msg_next(msg_t *msg)
{
  msg_t *next;
  unsigned n;

  if (msg && msg->m_next) {
    next = msg->m_next;
    msg->m_next = NULL;
    return next;
  }

  if ((n = msg_buf_committed(msg))) {
    if (msg_buf_move(next = msg_create(msg->m_class, msg->m_oflags), msg)) {
      msg_addr_copy(next, msg);
      return next;
    }
    /* How to indicate error? */
    msg_destroy(next);
  }

  return NULL;
}

/** Set next message of the stream */
int msg_set_next(msg_t *msg, msg_t *next)
{
  if (!msg || (next && next->m_next))
    return -1;

  if (msg->m_next && next)
    next->m_next = msg->m_next;

  msg->m_next = next;

  return 0;
}

/** Clear committed data */
void msg_clear_committed(msg_t *msg)
{
  if (msg) {
    unsigned n = msg_buf_committed(msg);

    if (n)
      msg_buf_used(msg, n);
  }
}

#if 0
struct sigcomp_udvm;

struct sigcomp_udvm *msg_get_udvm(msg_t *msg);
struct sigcomp_udvm *msg_set_udvm(msg_t *msg, struct sigcomp_udvm *);

/** Save UDVM. */
struct sigcomp_udvm *msg_set_udvm(msg_t *msg, struct sigcomp_udvm *udvm)
{
  struct sigcomp_udvm *prev = NULL;

  if (msg) {
    prev = msg->m_udvm;
    msg->m_udvm = udvm;
  }

  return prev;
}

/** Get saved UDVM */
struct sigcomp_udvm *msg_get_udvm(msg_t *msg)
{
  return msg ? msg->m_udvm : NULL;
}

#endif

/** Mark message as complete.
 *
 * @relates msg_s
 */
inline
unsigned msg_mark_as_complete(msg_t *msg, unsigned mask)
{
  if (msg) {
    msg->m_streaming = 0;
    return msg->m_object->msg_flags |= mask | MSG_FLG_COMPLETE;
  }
  else {
    return 0;
  }
}

/** Return true if message is complete.
 *
 * @relates msg_s
 */
int msg_is_complete(msg_t const *msg)
{
  return msg && MSG_IS_COMPLETE(msg->m_object);
}

/** Return true if message has parsing errors.
 *
 * @relates msg_s
*/
int msg_has_error(msg_t const *msg)
{
  return msg->m_object->msg_flags & MSG_FLG_ERROR;
}

/**Total size of message.
 *
 * @relates msg_s
 */
unsigned msg_size(msg_t const *msg)
{
  return msg ? msg->m_size : 0;
}

/** Set the maximum size of a message.
 *
 * @relates msg_s
 *
 * The function msg_maxsize() sets the maximum buffer size of a message. It
 * returns the previous maximum size. If the @a maxsize is 0, maximum size
 * is not set, but the current maximum size is returned.
 *
 * If the message size exceeds maxsize, msg_errno() returns ENOBUFS,
 * MSG_FLG_TOOLARGE and MSG_FLG_ERROR flags are set.
 */
unsigned msg_maxsize(msg_t *msg, unsigned maxsize)
{
  unsigned retval = 0;

  if (msg) {
    retval = msg->m_maxsize;
    if (maxsize)
      msg->m_maxsize = maxsize;
  }

  return retval;
}

/**Set the size of next fragment.
 *
 * @relates msg_s
 *
 * The function msg_streaming_size() sets the size of the message body for
 * streaming.
 */
unsigned msg_streaming_size(msg_t *msg, unsigned ssize)
{
  if (!msg)
    return -1;

  msg->m_ssize = ssize;

  return 0;
}

/**Allocate a list of external buffers.
 *
 * @relates msg_s
 *
 * The function msg_buf_external() allocates at most msg_n_fragments
 * external buffers for the message body.
 *
 * @return The function msg_buf_external() returns number of allocated
 * buffers, or -1 upon an error.
 */
int msg_buf_external(msg_t *msg,
		     unsigned N,
		     unsigned blocksize)
{
  msg_buffer_t *ext, *b, **bb;
  int i, I;

  assert(N <= 128 * 1024);

  if (msg == NULL)
    return -1;
  if (blocksize == 0)
    blocksize = msg_min_block;
  if (N == 0)
    N = blocksize;
  if (N > blocksize * msg_n_fragments)
    N = blocksize * msg_n_fragments;
  if (N > msg->m_ssize)
    N = msg->m_ssize;

  I = (N + blocksize - 1) / blocksize; assert(I <= msg_n_fragments);

  for (i = 0, bb = &ext; i < I; i++) {
    *bb = su_zalloc(msg_home(msg), sizeof **bb);
    if (!*bb)
      break;
    bb = &(*bb)->b_next;
  }

  if (i == I)
    for (b = ext, i = 0; b; b = b->b_next, i++) {
      b->b_data = su_alloc(msg_home(msg), b->b_size = blocksize);
      if (!b->b_data)
	break;
    }

  if (i == I) {
    /* Successful return */
    for (bb = &msg->m_stream; *bb; bb = &(*bb)->b_next)
      ;

    *bb = ext;

    if (msg->m_ssize != MSG_SSIZE_MAX)
      for (b = ext; b; b = b->b_next) {
	if (msg->m_ssize < b->b_size) {
	  b->b_size = msg->m_ssize;
	}
	msg->m_ssize -= b->b_size;
      }

    return i;
  }

  for (b = ext; b; b = ext) {
    ext = b->b_next;
    su_free(msg_home(msg), b->b_data);
    su_free(msg_home(msg), b);
  }

  return -1;

}

int msg_unref_external(msg_t *msg, msg_buffer_t *b)
{
  if (msg && b) {
    su_free(msg_home(msg), b->b_data);
    su_free(msg_home(msg), b);
    return 0;
  }
  errno = EINVAL;
  return -1;
}

/* ====================================================================== */
/* Parsing messages */

static inline int extract_incomplete_chunks(msg_t *, int eos);
static int extract_first(msg_t *, msg_pub_t *, char b[], int bsiz, int eos);
static inline int extract_next(msg_t *, msg_pub_t *, char *, int, int, int);
static int extract_header(msg_t *, msg_pub_t*,
			  char b[], int bsiz, int eos, int copy);
static msg_header_t *header_parse(msg_t *, msg_pub_t *, msg_href_t const *,
				  char s[], int slen, int copy_buffer);
static inline int
extract_trailers(msg_t *msg, msg_pub_t *mo,
		 char *b, int bsiz, int eos, int copy);

/** Calculate length of line ending (0, 1 or 2) */
#define CRLF_TEST(b) ((b)[0] == '\r' ? ((b)[1] == '\n') + 1 : (b)[0] =='\n')

static inline void
append_parsed(msg_t *msg, msg_pub_t *mo, msg_header_t **hh, msg_header_t *h,
	      int always_into_chain);

/**Extract and parse a message from internal buffer.
 *
 * @relates msg_s
 *
 * This function parses the internal buffer and adds the parsed fragments to
 * the message object. It marks the successfully parsed data as extracted.
 *
 * @param msg message to be parsed
 *
 * @retval positive if a complete message was parsed
 * @retval 0 if message was incomplete
 * @retval negative if an error occurred
 */
int msg_extract(msg_t *msg)
{
  msg_pub_t *mo = msg_object(msg);
  msg_mclass_t const *mc;
  char *b;
  int m, bsiz, eos;

  if (!msg || !msg->m_buffer->mb_data)
    return -1;

  assert(mo);

  mc = msg->m_class;
  mo = msg->m_object;
  eos = msg->m_buffer->mb_eos;

  if (msg->m_chunk) {
    m = extract_incomplete_chunks(msg, eos);
    if (m < 1 || MSG_IS_COMPLETE(mo))
      return m;
  }

  if (mo->msg_flags & MSG_FLG_TRAILERS)
    msg_set_streaming(msg, 0);

  if (msg->m_buffer->mb_used + msg->m_buffer->mb_commit ==
      msg->m_buffer->mb_size)
    /* Why? When? */
    return 0;

  assert(msg->m_buffer->mb_used + msg->m_buffer->mb_commit <
	 msg->m_buffer->mb_size);

  m = 0;

  b = msg->m_buffer->mb_data + msg->m_buffer->mb_used;
  bsiz = msg->m_buffer->mb_commit;
  b[bsiz] = '\0';

  while (msg->m_buffer->mb_commit > 0) {
    int flags = mo->msg_flags;
    int copy = MSG_IS_EXTRACT_COPY(flags);

    if (flags & MSG_FLG_COMPLETE)
      break;

    if (flags & MSG_FLG_TRAILERS)
      m = extract_trailers(msg, mo, b, bsiz, eos, copy);
    else if (flags & MSG_FLG_BODY)
      m = mc->mc_extract_body(msg, mo, b, bsiz, eos);
    else if (flags & MSG_FLG_HEADERS)
      m = extract_next(msg, mo, b, bsiz, eos, copy);
    else
      m = extract_first(msg, mo, b, bsiz, eos);

    if (m <= 0 || msg->m_chunk)
      break;

    b += m;
    bsiz -= m;

    msg_buf_used(msg, m);
  }

  if (eos && bsiz == 0)
    msg_mark_as_complete(msg, 0);

  if (m < 0 || (mo->msg_flags & MSG_FLG_ERROR)) {
    msg_mark_as_complete(msg, MSG_FLG_ERROR);
    return -1;
  }
  else if (!MSG_IS_COMPLETE(mo))
    return 0;
  else if (!(mo->msg_flags & MSG_FLG_HEADERS)) {
    msg_mark_as_complete(msg, MSG_FLG_ERROR);
    return -1;
  }
  else
    return 1;
}

static
int extract_first(msg_t *msg, msg_pub_t *mo, char b[], int bsiz, int eos)
{
  /* First line */
  int k, l, m, n, crlf, xtra;
  msg_header_t *h, **hh;
  msg_href_t const *hr;
  msg_mclass_t const *mc = msg->m_class;

  for (k = 0; IS_LWS(b[k]); k++) /* Skip whitespace */
    ;
  if (!b[k]) return k;

  /* If first token contains no /, this is request, otherwise status line */
  l = span_token(b + k) + k;
  if (b[l] != '/')
    hr = mc->mc_request;
  else
    hr = mc->mc_status;

  n = span_non_crlf(b + l) + l;
  if (!b[n])
    return eos ? -1 : 0;
  crlf = CRLF_TEST(b + n);

  for (m = n + crlf; IS_WS(b[m]); m++)
    ;
  /* In order to skip possible whitespace after first line, we don't parse
     first line until first non-ws char from next one has been received */
  if (!b[m] && !eos)
    return 0;

  xtra = MSG_IS_EXTRACT_COPY(mo->msg_flags) ? n + 1 - k : 0;
  if (!(h = msg_header_alloc(msg_home(msg), hr->hr_class, xtra)))
    return -1;

  if (xtra) {
    char *bb = memcpy(MSG_HEADER_DATA(h), b, xtra - 1);
    h->sh_data = b, h->sh_len = n + crlf;
    b = bb; n = xtra - 1;
  }
  else {
    b = b + k; n = n - k;
  }

  b[n] = 0;

  if (hr->hr_class->hc_parse(msg_home(msg), h, b, n) < 0)
    return -1;

  assert(hr->hr_offset);

  hh = (msg_header_t**)((char *)mo + hr->hr_offset);

  append_parsed(msg, mo, hh, h, 1);

  mo->msg_flags |= MSG_FLG_HEADERS;

  return m;
}

/* Extract header or message body */
static inline int
extract_next(msg_t *msg, msg_pub_t *mo, char *b, int bsiz, int eos, int copy)
{
  if (IS_CRLF(b[0]))
    return msg->m_class->mc_extract_body(msg, mo, b, bsiz, eos);
  else
    return extract_header(msg, mo, b, bsiz, eos, copy);
}

/** Extract a header. */
int msg_extract_header(msg_t *msg, msg_pub_t *mo, char b[], int bsiz, int eos)
{
  return extract_header(msg, mo, b, bsiz, eos, 0);
}

/** Extract a header from buffer @a b.
 */
static
int
extract_header(msg_t *msg, msg_pub_t *mo, char *b, int bsiz, int eos,
	       int copy_buffer)
{
  int len, n = 0, m, crlf = 0;
  int name_len = 0, name_len_set = 0;
  int error = 0, xtra;
  msg_header_t *h;
  msg_href_t const *hr;
  msg_mclass_t const *mc = msg->m_class;

  hr = msg_find_hclass(mc, b, &n); /* Get header name */
  error = n == 0;
  if (hr == NULL)		/* Panic */
    return -1;

  xtra = span_ws(b + n);

  /* Find next crlf which is not followed by whitespace */
  do {
    n += xtra + crlf;
    if (!eos && bsiz == n)
      return 0;
    m = span_non_crlf(b + n);
    if (!name_len_set && m)
      name_len = n, name_len_set = 1; /* First non-ws after COLON */
    n += m;
    crlf = CRLF_TEST(b + n);
    xtra = span_ws(b + n + crlf);
  }
  while (xtra);

  if (!eos && bsiz == n + crlf)
    return 0;

  if (hr->hr_class->hc_hash == msg_unknown_hash)
    name_len = 0, name_len_set = 1;

  if (error) {
    msg->m_extract_err |= hr->hr_flags;
    if (hr->hr_class->hc_critical)
      mo->msg_flags |= MSG_FLG_ERROR;
    hr = mc->mc_error;
    copy_buffer = 1;
    h = header_parse(msg, mo, hr, NULL, 0, 0);
  }
  else {
    if (!name_len_set)
      /* Empty header - nothing but name, COLON and LWS */
      name_len = n;
    else
      /* Strip extra whitespace at the end of header */
      while (n > name_len && IS_LWS(b[n - 1]))
	n--, crlf++;

    h = header_parse(msg, mo, hr, b + name_len, n - name_len, copy_buffer);
  }

  if (h == NULL)
    return -1;

  len = n + crlf;

  /*
   * If the header contains multiple header fields, set the pointer to the
   * encodeded data correctly
   */
  while (h) {
    if (copy_buffer)
      h->sh_data = b, h->sh_len = len;
    b += len, len = 0;
    if (h->sh_succ)
      assert(&h->sh_succ == h->sh_succ->sh_prev);
    h = h->sh_next;
  }

  return n + crlf;
}

static
msg_header_t *header_parse(msg_t *msg, msg_pub_t *mo,
			   msg_href_t const *hr,
			   char s[], int slen,
			   int copy_buffer)
{
  su_home_t *home = msg_home(msg);
  msg_header_t *h, **hh;
  msg_header_t *h0, *h_next;
  msg_hclass_t *hc = hr->hr_class;
  int n;
  int add_to_list, clear = 0;

  hh = (msg_header_t **)((char *)mo + hr->hr_offset);

  add_to_list = (hc->hc_kind == msg_kind_list && !copy_buffer && *hh);

  if (add_to_list)
    h = *hh;