Chunking: do not treat the first lonely dot special. CVE-2017-16944, Bug 2201

[exim.git] / src / src / receive.c

diff --git a/src/src/receive.c b/src/src/receive.c
index e535876191e067e59fcbf71ddfd83196c8b7deb2..2812ea2c87743f5db2171cbf86e00c6c24d08a9e 100644 (file)
--- a/src/src/receive.c
+++ b/src/src/receive.c
@@ -2,7 +2,7 @@
 *     Exim - an Internet mail transport agent    *
 *************************************************/
 
 *     Exim - an Internet mail transport agent    *
 *************************************************/
 

-/* Copyright (c) University of Cambridge 1995 - 2016 */
+/* Copyright (c) University of Cambridge 1995 - 2017 */

 /* See the file NOTICE for conditions of use and distribution. */
 
 /* Code for receiving a message and setting up spool files. */
 /* See the file NOTICE for conditions of use and distribution. */
 
 /* Code for receiving a message and setting up spool files. */


@@ -25,6 +25,7 @@ static FILE   *data_file = NULL;
 static int     data_fd = -1;
 static uschar *spool_name = US"";
 
 static int     data_fd = -1;
 static uschar *spool_name = US"";
 

+enum CH_STATE {LF_SEEN, MID_LINE, CR_SEEN};

 
 
 /*************************************************
 
 
 /*************************************************


@@ -37,7 +38,7 @@ the file. (When SMTP input is occurring, different functions are used by
 changing the pointer variables.) */
 
 int
 changing the pointer variables.) */
 
 int

-stdin_getc(void)
+stdin_getc(unsigned lim)

 {
 return getc(stdin);
 }
 {
 return getc(stdin);
 }


@@ -626,7 +627,7 @@ if (!dot_ends)
   {
   register int last_ch = '\n';
 
   {
   register int last_ch = '\n';
 

-  for (; (ch = (receive_getc)()) != EOF; last_ch = ch)
+  for (; (ch = (receive_getc)(GETC_BUFFER_UNLIMITED)) != EOF; last_ch = ch)

     {
     if (ch == 0) body_zerocount++;
     if (last_ch == '\r' && ch != '\n')
     {
     if (ch == 0) body_zerocount++;
     if (last_ch == '\r' && ch != '\n')


@@ -668,7 +669,7 @@ if (!dot_ends)
 
 ch_state = 1;
 
 
 ch_state = 1;
 

-while ((ch = (receive_getc)()) != EOF)
+while ((ch = (receive_getc)(GETC_BUFFER_UNLIMITED)) != EOF)

   {
   if (ch == 0) body_zerocount++;
   switch (ch_state)
   {
   if (ch == 0) body_zerocount++;
   switch (ch_state)


@@ -786,7 +787,7 @@ int ch_state = 0;
 int ch;
 int linelength = 0;
 
 int ch;
 int linelength = 0;
 

-while ((ch = (receive_getc)()) != EOF)
+while ((ch = (receive_getc)(GETC_BUFFER_UNLIMITED)) != EOF)

   {
   if (ch == 0) body_zerocount++;
   switch (ch_state)
   {
   if (ch == 0) body_zerocount++;
   switch (ch_state)


@@ -898,11 +899,15 @@ return END_EOF;
 
 
 /* Variant of the above read_message_data_smtp() specialised for RFC 3030
 
 
 /* Variant of the above read_message_data_smtp() specialised for RFC 3030

-CHUNKING.  We assume that the incoming has proper CRLF, so only have to scan
-for and strip CR.  On the downside there are more protocol reasons to stop.
+CHUNKING. Accept input lines separated by either CRLF or CR or LF and write
+LF-delimited spoolfile.  Until we have wireformat spoolfiles, we need the
+body_linecount accounting for proper re-expansion for the wire, so use
+a cut-down version of the state-machine above; we don't need to do leading-dot
+detection and unstuffing.

 
 Arguments:
 
 Arguments:

-  fout      a FILE to which to write the message; NULL if skipping
+  fout      a FILE to which to write the message; NULL if skipping;
+            must be open for both writing and reading.

 
 Returns:    One of the END_xxx values indicating why it stopped reading
 */
 
 Returns:    One of the END_xxx values indicating why it stopped reading
 */


@@ -910,43 +915,105 @@ Returns:    One of the END_xxx values indicating why it stopped reading
 static int
 read_message_bdat_smtp(FILE *fout)
 {
 static int
 read_message_bdat_smtp(FILE *fout)
 {

-int ch;
-int linelength = 0;
+int linelength = 0, ch;
+enum CH_STATE ch_state = LF_SEEN;
+BOOL fix_nl = FALSE;

 
 

-for (;;) switch (ch = bdat_getc())
+for(;;)

   {
   {

-  case EOF: return END_EOF;
-  case EOD: return END_DOT;
-  case ERR: return END_PROTOCOL;
+  switch ((ch = (bdat_getc)(GETC_BUFFER_UNLIMITED)))
+    {
+    case EOF:  return END_EOF;
+    case ERR:  return END_PROTOCOL;
+    case EOD:
+      /* Nothing to get from the sender anymore. We check the last
+      character written to the spool.
+
+      RFC 3030 states, that BDAT chunks are normal text, terminated by CRLF.
+      If we would be strict, we would refuse such broken messages.
+      But we are liberal, so we fix it.  It would be easy just to append
+      the "\n" to the spool.
+
+      But there are some more things (line counting, message size calculation and such),
+      that would need to be duplicated here.  So we simply do some ungetc
+      trickery.
+      */
+      if (fout)
+       {
+       if (fseek(fout, -1, SEEK_CUR) < 0)      return END_PROTOCOL;
+       if (fgetc(fout) == '\n')                return END_DOT;
+       }

 
 

-  case '\r':
-    body_linecount++;
-    if (linelength > max_received_linelength)
-      max_received_linelength = linelength;
-    linelength = -1;
-    break;
+      if (linelength == -1)    /* \r already seen (see below) */
+        {
+        DEBUG(D_receive) debug_printf("Add missing LF\n");
+        bdat_ungetc('\n');
+        continue;
+        }
+      DEBUG(D_receive) debug_printf("Add missing CRLF\n");
+      bdat_ungetc('\r');      /* not even \r was seen */
+      fix_nl = TRUE;

 
 

-  case 0:
-    body_zerocount++;
-    /*FALLTHROUGH*/
-  default:
-    message_size++;
-    linelength++;
-    if (fout)
-      {
-      if (fputc(ch, fout) == EOF) return END_WERROR;
-      if (message_size > thismessage_size_limit) return END_SIZE;
-      }
-#ifdef notyet
-    if(ch == '\n')
-      (void) cutthrough_put_nl();
-    else
-      {
-      uschar c = ch;
-      (void) cutthrough_puts(&c, 1);
-      }
-#endif
-    break;
+      continue;
+    case '\0':  body_zerocount++; break;
+    }
+  switch (ch_state)
+    {
+    case LF_SEEN:                             /* After LF or CRLF */
+      ch_state = MID_LINE;
+      /* fall through to handle as normal uschar. */
+
+    case MID_LINE:                            /* Mid-line state */
+      if (ch == '\n')
+       {
+       ch_state = LF_SEEN;
+       body_linecount++;
+       if (linelength > max_received_linelength)
+         max_received_linelength = linelength;
+       linelength = -1;
+       }
+      else if (ch == '\r')
+       {
+       ch_state = CR_SEEN;
+       if (fix_nl) bdat_ungetc('\n');
+       continue;                       /* don't write CR */
+       }
+      break;
+
+    case CR_SEEN:                       /* After (unwritten) CR */
+      body_linecount++;
+      if (linelength > max_received_linelength)
+       max_received_linelength = linelength;
+      linelength = -1;
+      if (ch == '\n')
+       ch_state = LF_SEEN;
+      else
+       {
+       message_size++;
+       if (fout && fputc('\n', fout) == EOF) return END_WERROR;
+       (void) cutthrough_put_nl();
+       if (ch == '\r') continue;       /* don't write CR */
+       ch_state = MID_LINE;
+       }
+      break;
+    }
+
+  /* Add the character to the spool file, unless skipping */
+
+  message_size++;
+  linelength++;
+  if (fout)
+    {
+    if (fputc(ch, fout) == EOF) return END_WERROR;
+    if (message_size > thismessage_size_limit) return END_SIZE;
+    }
+  if(ch == '\n')
+    (void) cutthrough_put_nl();
+  else
+    {
+    uschar c = ch;
+    (void) cutthrough_puts(&c, 1);
+    }

   }
 /*NOTREACHED*/
 }
   }
 /*NOTREACHED*/
 }


@@ -1083,7 +1150,7 @@ switch(where)
 
 if (acl_removed_headers != NULL)
   {
 
 if (acl_removed_headers != NULL)
   {

-  DEBUG(D_receive|D_acl) debug_printf(">>Headers removed by %s ACL:\n", acl_name);
+  DEBUG(D_receive|D_acl) debug_printf_indent(">>Headers removed by %s ACL:\n", acl_name);

 
   for (h = header_list; h != NULL; h = h->next) if (h->type != htype_old)
     {
 
   for (h = header_list; h != NULL; h = h->next) if (h->type != htype_old)
     {


@@ -1096,15 +1163,15 @@ if (acl_removed_headers != NULL)
       if (header_testname(h, s, Ustrlen(s), FALSE))
        {
        h->type = htype_old;
       if (header_testname(h, s, Ustrlen(s), FALSE))
        {
        h->type = htype_old;

-        DEBUG(D_receive|D_acl) debug_printf("  %s", h->text);
+        DEBUG(D_receive|D_acl) debug_printf_indent("  %s", h->text);

        }
     }
   acl_removed_headers = NULL;
        }
     }
   acl_removed_headers = NULL;

-  DEBUG(D_receive|D_acl) debug_printf(">>\n");
+  DEBUG(D_receive|D_acl) debug_printf_indent(">>\n");

   }
 
 if (acl_added_headers == NULL) return;
   }
 
 if (acl_added_headers == NULL) return;

-DEBUG(D_receive|D_acl) debug_printf(">>Headers added by %s ACL:\n", acl_name);
+DEBUG(D_receive|D_acl) debug_printf_indent(">>Headers added by %s ACL:\n", acl_name);

 
 for (h = acl_added_headers; h != NULL; h = next)
   {
 
 for (h = acl_added_headers; h != NULL; h = next)
   {


@@ -1115,7 +1182,7 @@ for (h = acl_added_headers; h != NULL; h = next)
     case htype_add_top:
     h->next = header_list;
     header_list = h;
     case htype_add_top:
     h->next = header_list;
     header_list = h;

-    DEBUG(D_receive|D_acl) debug_printf("  (at top)");
+    DEBUG(D_receive|D_acl) debug_printf_indent("  (at top)");

     break;
 
     case htype_add_rec:
     break;
 
     case htype_add_rec:


@@ -1130,7 +1197,7 @@ for (h = acl_added_headers; h != NULL; h = next)
       }
     h->next = last_received->next;
     last_received->next = h;
       }
     h->next = last_received->next;
     last_received->next = h;

-    DEBUG(D_receive|D_acl) debug_printf("  (after Received:)");
+    DEBUG(D_receive|D_acl) debug_printf_indent("  (after Received:)");

     break;
 
     case htype_add_rfc:
     break;
 
     case htype_add_rfc:


@@ -1145,7 +1212,7 @@ for (h = acl_added_headers; h != NULL; h = next)
        of all headers. Our current header must follow it. */
     h->next = last_received->next;
     last_received->next = h;
        of all headers. Our current header must follow it. */
     h->next = last_received->next;
     last_received->next = h;

-    DEBUG(D_receive|D_acl) debug_printf("  (before any non-Received: or Resent-*: header)");
+    DEBUG(D_receive|D_acl) debug_printf_indent("  (before any non-Received: or Resent-*: header)");

     break;
 
     default:
     break;
 
     default:


@@ -1165,11 +1232,11 @@ for (h = acl_added_headers; h != NULL; h = next)
   h->type = header_checkname(h, FALSE);
   if (h->type >= 'a') h->type = htype_other;
 
   h->type = header_checkname(h, FALSE);
   if (h->type >= 'a') h->type = htype_other;
 

-  DEBUG(D_receive|D_acl) debug_printf("  %s", header_last->text);
+  DEBUG(D_receive|D_acl) debug_printf_indent("  %s", header_last->text);

   }
 
 acl_added_headers = NULL;
   }
 
 acl_added_headers = NULL;

-DEBUG(D_receive|D_acl) debug_printf(">>\n");
+DEBUG(D_receive|D_acl) debug_printf_indent(">>\n");

 }
 
 
 }
 
 


@@ -1321,7 +1388,7 @@ if (rc == OK)
       {
       (void) string_format(rfc822_file_path, sizeof(rfc822_file_path),
        "%s/scan/%s/%s", spool_directory, message_id, entry->d_name);
       {
       (void) string_format(rfc822_file_path, sizeof(rfc822_file_path),
        "%s/scan/%s/%s", spool_directory, message_id, entry->d_name);

-      debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n",
+      DEBUG(D_receive) debug_printf("RFC822 attachment detected: running MIME ACL for '%s'\n",

        rfc822_file_path);
       break;
       }
        rfc822_file_path);
       break;
       }


@@ -1682,7 +1749,7 @@ next->text. */
 
 for (;;)
   {
 
 for (;;)
   {

-  int ch = (receive_getc)();
+  int ch = (receive_getc)(GETC_BUFFER_UNLIMITED);

 
   /* If we hit EOF on a SMTP connection, it's an error, since incoming
   SMTP must have a correct "." terminator. */
 
   /* If we hit EOF on a SMTP connection, it's an error, since incoming
   SMTP must have a correct "." terminator. */


@@ -1705,8 +1772,8 @@ for (;;)
   (and sometimes lunatic messages can have ones that are 100s of K long) we
   call store_release() for strings that have been copied - if the string is at
   the start of a block (and therefore the only thing in it, because we aren't
   (and sometimes lunatic messages can have ones that are 100s of K long) we
   call store_release() for strings that have been copied - if the string is at
   the start of a block (and therefore the only thing in it, because we aren't

-  doing any other gets), the block gets freed. We can only do this because we
-  know there are no other calls to store_get() going on. */
+  doing any other gets), the block gets freed. We can only do this release if
+  there were no allocations since the once that we want to free. */

 
   if (ptr >= header_size - 4)
     {
 
   if (ptr >= header_size - 4)
     {


@@ -1715,9 +1782,10 @@ for (;;)
     header_size *= 2;
     if (!store_extend(next->text, oldsize, header_size))
       {
     header_size *= 2;
     if (!store_extend(next->text, oldsize, header_size))
       {

+      BOOL release_ok = store_last_get[store_pool] == next->text;

       uschar *newtext = store_get(header_size);
       memcpy(newtext, next->text, ptr);
       uschar *newtext = store_get(header_size);
       memcpy(newtext, next->text, ptr);

-      store_release(next->text);
+      if (release_ok) store_release(next->text);

       next->text = newtext;
       }
     }
       next->text = newtext;
       }
     }


@@ -1759,12 +1827,12 @@ for (;;)
   prevent further reading), and break out of the loop, having freed the
   empty header, and set next = NULL to indicate no data line. */
 
   prevent further reading), and break out of the loop, having freed the
   empty header, and set next = NULL to indicate no data line. */
 

-  if (ptr == 0 && ch == '.' && (smtp_input || dot_ends))
+  if (ptr == 0 && ch == '.' && dot_ends)

     {
     {

-    ch = (receive_getc)();
+    ch = (receive_getc)(GETC_BUFFER_UNLIMITED);

     if (ch == '\r')
       {
     if (ch == '\r')
       {

-      ch = (receive_getc)();
+      ch = (receive_getc)(GETC_BUFFER_UNLIMITED);

       if (ch != '\n')
         {
         receive_ungetc(ch);
       if (ch != '\n')
         {
         receive_ungetc(ch);


@@ -1795,7 +1863,7 @@ for (;;)
 
   if (ch == '\r')
     {
 
   if (ch == '\r')
     {

-    ch = (receive_getc)();
+    ch = (receive_getc)(GETC_BUFFER_UNLIMITED);

     if (ch == '\n')
       {
       if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;
     if (ch == '\n')
       {
       if (first_line_ended_crlf == TRUE_UNSET) first_line_ended_crlf = TRUE;


@@ -1890,7 +1958,7 @@ for (;;)
 
   if (ch != EOF)
     {
 
   if (ch != EOF)
     {

-    int nextch = (receive_getc)();
+    int nextch = (receive_getc)(GETC_BUFFER_UNLIMITED);

     if (nextch == ' ' || nextch == '\t')
       {
       next->text[ptr++] = nextch;
     if (nextch == ' ' || nextch == '\t')
       {
       next->text[ptr++] = nextch;


@@ -2084,6 +2152,21 @@ for (;;)
       }
     }
 
       }
     }
 

+  /* Reject CHUNKING messages that do not CRLF their first header line */
+
+  if (!first_line_ended_crlf && chunking_state > CHUNKING_OFFERED)
+    {
+    log_write(L_size_reject, LOG_MAIN|LOG_REJECT, "rejected from <%s>%s%s%s%s: "
+      "Non-CRLF-terminated header, under CHUNKING: message abandoned",
+      sender_address,
+      sender_fullhost ? " H=" : "", sender_fullhost ? sender_fullhost : US"",
+      sender_ident ? " U=" : "",    sender_ident ? sender_ident : US"");
+    smtp_printf("552 Message header not CRLF terminated\r\n");
+    bdat_flush_data();
+    smtp_reply = US"";
+    goto TIDYUP;                             /* Skip to end of function */
+    }
+

   /* The line has been handled. If we have hit EOF, break out of the loop,
   indicating no pending data line. */
 
   /* The line has been handled. If we have hit EOF, break out of the loop,
   indicating no pending data line. */
 


@@ -2108,7 +2191,7 @@ normal case). */
 DEBUG(D_receive)
   {
   debug_printf(">>Headers received:\n");
 DEBUG(D_receive)
   {
   debug_printf(">>Headers received:\n");

-  for (h = header_list->next; h != NULL; h = h->next)
+  for (h = header_list->next; h; h = h->next)

     debug_printf("%s", h->text);
   debug_printf("\n");
   }
     debug_printf("%s", h->text);
   debug_printf("\n");
   }


@@ -2135,7 +2218,7 @@ if (filter_test != FTEST_NONE && header_list->next == NULL)
 /* Scan the headers to identify them. Some are merely marked for later
 processing; some are dealt with here. */
 
 /* Scan the headers to identify them. Some are merely marked for later
 processing; some are dealt with here. */
 

-for (h = header_list->next; h != NULL; h = h->next)
+for (h = header_list->next; h; h = h->next)

   {
   BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
   if (is_resent) contains_resent_headers = TRUE;
   {
   BOOL is_resent = strncmpic(h->text, US"resent-", 7) == 0;
   if (is_resent) contains_resent_headers = TRUE;


@@ -2351,7 +2434,7 @@ if (extract_recip)
 
   /* Now scan the headers */
 
 
   /* Now scan the headers */
 

-  for (h = header_list->next; h != NULL; h = h->next)
+  for (h = header_list->next; h; h = h->next)

     {
     if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
         (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))
     {
     if ((h->type == htype_to || h->type == htype_cc || h->type == htype_bcc) &&
         (!contains_resent_headers || strncmpic(h->text, US"resent-", 7) == 0))


@@ -2500,8 +2583,9 @@ letter and it is not used internally.
 NOTE: If ever the format of message ids is changed, the regular expression for
 checking that a string is in this format must be updated in a corresponding
 way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH
 NOTE: If ever the format of message ids is changed, the regular expression for
 checking that a string is in this format must be updated in a corresponding
 way. It appears in the initializing code in exim.c. The macro MESSAGE_ID_LENGTH

-must also be changed to reflect the correct string length. Then, of course,
-other programs that rely on the message id format will need updating too. */
+must also be changed to reflect the correct string length. The queue-sort code
+needs to know the layout. Then, of course, other programs that rely on the
+message id format will need updating too. */

 
 Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
 message_id[6] = '-';
 
 Ustrncpy(message_id, string_base62((long int)(message_id_tv.tv_sec)), 6);
 message_id[6] = '-';


@@ -2845,11 +2929,11 @@ We start at the second header, skipping our own Received:. This rewriting is
 documented as happening *after* recipient addresses are taken from the headers
 by the -t command line option. An added Sender: gets rewritten here. */
 
 documented as happening *after* recipient addresses are taken from the headers
 by the -t command line option. An added Sender: gets rewritten here. */
 

-for (h = header_list->next; h != NULL; h = h->next)
+for (h = header_list->next; h; h = h->next)

   {
   header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
     rewrite_existflags, TRUE);
   {
   header_line *newh = rewrite_header(h, NULL, NULL, global_rewrite_rules,
     rewrite_existflags, TRUE);

-  if (newh != NULL) h = newh;
+  if (newh) h = newh;

   }
 
 
   }
 
 


@@ -3738,7 +3822,7 @@ if (bmi_run == 1)
   }
 #endif
 
   }
 #endif
 

-/* Update the timstamp in our Received: header to account for any time taken by
+/* Update the timestamp in our Received: header to account for any time taken by

 an ACL or by local_scan(). The new time is the time that all reception
 processing is complete. */
 
 an ACL or by local_scan(). The new time is the time that all reception
 processing is complete. */
 


@@ -3823,7 +3907,7 @@ string as required. Since we commonly want to add two items at a time, use a
 macro to simplify the coding. We log the arrival of a new message while the
 file is still locked, just in case the machine is *really* fast, and delivers
 it first! Include any message id that is in the message - since the syntax of a
 macro to simplify the coding. We log the arrival of a new message while the
 file is still locked, just in case the machine is *really* fast, and delivers
 it first! Include any message id that is in the message - since the syntax of a

-message id is actually an addr-spec, we can use the parse routine to canonicize
+message id is actually an addr-spec, we can use the parse routine to canonicalize

 it. */
 
 size = 256;
 it. */
 
 size = 256;


@@ -4024,7 +4108,7 @@ if (smtp_input && sender_host_address != NULL && !sender_host_notsocket &&
 
   if (select(fileno(smtp_in) + 1, &select_check, NULL, NULL, &tv) != 0)
     {
 
   if (select(fileno(smtp_in) + 1, &select_check, NULL, NULL, &tv) != 0)
     {

-    int c = (receive_getc)();
+    int c = (receive_getc)(GETC_BUFFER_UNLIMITED);

     if (c != EOF) (receive_ungetc)(c); else
       {
       smtp_notquit_exit(US"connection-lost", NULL, NULL);
     if (c != EOF) (receive_ungetc)(c); else
       {
       smtp_notquit_exit(US"connection-lost", NULL, NULL);


@@ -4058,7 +4142,7 @@ for this message. */
 
    Send dot onward.  If accepted, wipe the spooled files, log as delivered and accept
    the sender's dot (below).
 
    Send dot onward.  If accepted, wipe the spooled files, log as delivered and accept
    the sender's dot (below).

-   If rejected: copy response to sender, wipe the spooled files, log approriately.
+   If rejected: copy response to sender, wipe the spooled files, log appropriately.

    If temp-reject: normally accept to sender, keep the spooled file - unless defer=pass
    in which case pass temp-reject back to initiator and dump the files.
 
    If temp-reject: normally accept to sender, keep the spooled file - unless defer=pass
    in which case pass temp-reject back to initiator and dump the files.