SECURITY: fix Qualys CVE-2020-PFPZA

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

diff --git a/src/src/parse.c b/src/src/parse.c
index 5d50d6862181876289bec20ebcf38b5c642abaca..8d689e88adc349b59cafd110813d64779fc562fa 100644 (file)
--- a/src/src/parse.c
+++ b/src/src/parse.c
@@ -3,6 +3,7 @@
 *************************************************/
 
 /* Copyright (c) University of Cambridge 1995 - 2018 */
 *************************************************/
 
 /* Copyright (c) University of Cambridge 1995 - 2018 */

+/* Copyright (c) The Exim Maintainers 2020 */

 /* See the file NOTICE for conditions of use and distribution. */
 
 /* Functions for parsing addresses */
 /* See the file NOTICE for conditions of use and distribution. */
 
 /* Functions for parsing addresses */


@@ -197,10 +198,10 @@ last_comment_position = s;
 while (*s)
   {
   int c, level;
 while (*s)
   {
   int c, level;

-  while (isspace(*s)) s++;
-  if (*s != '(') break;
+
+  if (Uskip_whitespace(&s) != '(') break;

   level = 1;
   level = 1;

-  while((c = *(++s)) != 0)
+  while((c = *(++s)))

     {
     if (c == '(') level++;
     else if (c == ')') { if (--level <= 0) { s++; break; } }
     {
     if (c == '(') level++;
     else if (c == ')') { if (--level <= 0) { s++; break; } }


@@ -807,11 +808,11 @@ while (isspace(endptr[-1])) endptr--;
 *end = endptr - US mailbox;
 
 /* Although this code has no limitation on the length of address extracted,
 *end = endptr - US mailbox;
 
 /* Although this code has no limitation on the length of address extracted,

-other parts of Exim may have limits, and in any case, RFC 2821 limits local
-parts to 64 and domains to 255, so we do a check here, giving an error if the
-address is ridiculously long. */
+other parts of Exim may have limits, and in any case, RFC 5321 limits email
+addresses to 256, so we do a check here, giving an error if the address is
+ridiculously long. */

 
 

-if (*end - *start > ADDRESS_MAXLENGTH)
+if (*end - *start > EXIM_EMAILADDR_MAX)

   {
   *errorptr = string_sprintf("address is ridiculously long: %.64s...", yield);
   return NULL;
   {
   *errorptr = string_sprintf("address is ridiculously long: %.64s...", yield);
   return NULL;


@@ -842,8 +843,7 @@ return NULL;
 
 /* This function is used for quoting text in headers according to RFC 2047.
 If the only characters that strictly need quoting are spaces, we return the
 
 /* This function is used for quoting text in headers according to RFC 2047.
 If the only characters that strictly need quoting are spaces, we return the

-original string, unmodified. If a quoted string is too long for the buffer, it
-is truncated. (This shouldn't happen: this is normally handling short strings.)
+original string, unmodified.

 
 Hmmph. As always, things get perverted for other uses. This function was
 originally for the "phrase" part of addresses. Now it is being used for much
 
 Hmmph. As always, things get perverted for other uses. This function was
 originally for the "phrase" part of addresses. Now it is being used for much


@@ -855,77 +855,62 @@ Arguments:
                  chars
   len          the length of the string
   charset      the name of the character set; NULL => iso-8859-1
                  chars
   len          the length of the string
   charset      the name of the character set; NULL => iso-8859-1

-  buffer       the buffer to put the answer in
-  buffer_size  the size of the buffer

   fold         if TRUE, a newline is inserted before the separating space when
                  more than one encoded-word is generated
 
 Returns:       pointer to the original string, if no quoting needed, or
   fold         if TRUE, a newline is inserted before the separating space when
                  more than one encoded-word is generated
 
 Returns:       pointer to the original string, if no quoting needed, or

-               pointer to buffer containing the quoted string, or
-               a pointer to "String too long" if the buffer can't even hold
-               the introduction
+               pointer to allocated memory containing the quoted string

 */
 
 const uschar *
 */
 
 const uschar *

-parse_quote_2047(const uschar *string, int len, uschar *charset, uschar *buffer,
-  int buffer_size, BOOL fold)
+parse_quote_2047(const uschar *string, int len, uschar *charset, BOOL fold)

 {
 {

-const uschar *s = string;
-uschar *p, *t;
-int hlen;
+const uschar * s = string;
+int hlen, l;

 BOOL coded = FALSE;
 BOOL first_byte = FALSE;
 BOOL coded = FALSE;
 BOOL first_byte = FALSE;

+gstring * g =
+  string_fmt_append(NULL, "=?%s?Q?", charset ? charset : US"iso-8859-1");

 
 

-if (!charset) charset = US"iso-8859-1";
-
-/* We don't expect this to fail! */
-
-if (!string_format(buffer, buffer_size, "=?%s?Q?", charset))
-  return US"String too long";
+hlen = l = g->ptr;

 
 

-hlen = Ustrlen(buffer);
-t = buffer + hlen;
-p = buffer;
-
-for (; len > 0; len--)
+for (s = string; len > 0; s++, len--)

   {
   {

-  int ch = *s++;
-  if (t > buffer + buffer_size - hlen - 8) break;
+  int ch = *s;

 
 

-  if ((t - p > 67) && !first_byte)
+  if (g->ptr - l > 67 && !first_byte)

     {
     {

-    *t++ = '?';
-    *t++ = '=';
-    if (fold) *t++ = '\n';
-    *t++ = ' ';
-    p = t;
-    Ustrncpy(p, buffer, hlen);
-    t += hlen;
+    g = fold ? string_catn(g, US"?=\n ", 4) : string_catn(g, US"?= ", 3);
+    l = g->ptr;
+    g = string_catn(g, g->s, hlen);

     }
 
     }
 

-  if (ch < 33 || ch > 126 ||
-      Ustrchr("?=()<>@,;:\\\".[]_", ch) != NULL)
+  if (  ch < 33 || ch > 126
+     || Ustrchr("?=()<>@,;:\\\".[]_", ch) != NULL)

     {
     if (ch == ' ')
       {
     {
     if (ch == ' ')
       {

-      *t++ = '_';
+      g = string_catn(g, US"_", 1);

       first_byte = FALSE;
       }
     else
       {
       first_byte = FALSE;
       }
     else
       {

-      t += sprintf(CS t, "=%02X", ch);
+      g = string_fmt_append(g, "=%02X", ch);

       coded = TRUE;
       first_byte = !first_byte;
       }
     }
       coded = TRUE;
       first_byte = !first_byte;
       }
     }

-  else { *t++ = ch; first_byte = FALSE; }
+  else
+    { g = string_catn(g, s, 1); first_byte = FALSE; }

   }
 
   }
 

-*t++ = '?';
-*t++ = '=';
-*t = 0;
+if (coded)
+  string = string_from_gstring(g = string_catn(g, US"?=", 2));
+else
+  g->ptr = -1;

 
 

-return coded ? buffer : string;
+gstring_release_unused(g);
+return string;

 }
 
 
 }
 
 


@@ -968,32 +953,25 @@ August 2000: Additional code added:
   We *could* use this for all cases, getting rid of the messy original code,
   but leave it for now. It would complicate simple cases like "John Q. Smith".
 
   We *could* use this for all cases, getting rid of the messy original code,
   but leave it for now. It would complicate simple cases like "John Q. Smith".
 

-The result is passed back in the buffer; it is usually going to be added to
-some other string. In order to be sure there is going to be no overflow,
-restrict the length of the input to 1/4 of the buffer size - this allows for
-every single character to be quoted or encoded without overflowing, and that
-wouldn't happen because of amalgamation. If the phrase is too long, return a
-fixed string.
+The result is passed back in allocated memory.

 
 Arguments:
   phrase       an RFC822 phrase
   len          the length of the phrase
 
 Arguments:
   phrase       an RFC822 phrase
   len          the length of the phrase

-  buffer       a buffer to put the result in
-  buffer_size  the size of the buffer

 
 Returns:       the fixed RFC822 phrase
 */
 
 const uschar *
 
 Returns:       the fixed RFC822 phrase
 */
 
 const uschar *

-parse_fix_phrase(const uschar *phrase, int len, uschar *buffer, int buffer_size)
+parse_fix_phrase(const uschar *phrase, int len)

 {
 int ch, i;
 BOOL quoted = FALSE;
 const uschar *s, *end;
 {
 int ch, i;
 BOOL quoted = FALSE;
 const uschar *s, *end;

+uschar * buffer;

 uschar *t, *yield;
 
 while (len > 0 && isspace(*phrase)) { phrase++; len--; }
 uschar *t, *yield;
 
 while (len > 0 && isspace(*phrase)) { phrase++; len--; }

-if (len > buffer_size/4) return US"Name too long";

 
 /* See if there are any non-printing characters, and if so, use the RFC 2047
 encoding for the whole thing. */
 
 /* See if there are any non-printing characters, and if so, use the RFC 2047
 encoding for the whole thing. */


@@ -1001,11 +979,18 @@ encoding for the whole thing. */
 for (i = 0, s = phrase; i < len; i++, s++)
   if ((*s < 32 && *s != '\t') || *s > 126) break;
 
 for (i = 0, s = phrase; i < len; i++, s++)
   if ((*s < 32 && *s != '\t') || *s > 126) break;
 

-if (i < len) return parse_quote_2047(phrase, len, headers_charset, buffer,
-  buffer_size, FALSE);
+if (i < len)
+  return parse_quote_2047(phrase, len, headers_charset, FALSE);

 
 /* No non-printers; use the RFC 822 quoting rules */
 
 
 /* No non-printers; use the RFC 822 quoting rules */
 

+if (!len)
+  {
+  return string_copy_taint_function("", is_tainted(phrase));
+  }
+
+buffer = store_get(len*4, is_tainted(phrase));
+

 s = phrase;
 end = s + len;
 yield = t = buffer + 1;
 s = phrase;
 end = s + len;
 yield = t = buffer + 1;


@@ -1149,9 +1134,17 @@ while (s < end)
             {
             if (ss >= end) ss--;
             *t++ = '(';
             {
             if (ss >= end) ss--;
             *t++ = '(';

-            Ustrncpy(t, s, ss-s);
-            t += ss-s;
-            s = ss;
+            if (ss < s)
+              {
+              /* Someone has ended the string with "<punct>(". */
+              ss = s;
+              }
+            else
+              {
+              Ustrncpy(t, s, ss-s);
+              t += ss-s;
+              s = ss;
+              }

             }
           }
 
             }
           }
 


@@ -1172,6 +1165,7 @@ while (s < end)
   }
 
 *t = 0;
   }
 
 *t = 0;

+store_release_above(t+1);

 return yield;
 }
 
 return yield;
 }
 


@@ -1524,7 +1518,7 @@ for (;;)
 
     if (!f)
       {
 
     if (!f)
       {

-      *error = string_open_failed(errno, "included file %s", filename);
+      *error = string_open_failed("included file %s", filename);

       return FF_INCLUDEFAIL;
       }
 
       return FF_INCLUDEFAIL;
       }
 


@@ -2101,8 +2095,8 @@ int main(void)
 {
 int start, end, domain;
 uschar buffer[1024];
 {
 int start, end, domain;
 uschar buffer[1024];

-uschar outbuff[1024];

 
 

+store_init();

 big_buffer = store_malloc(big_buffer_size);
 
 /* strip_trailing_dot = TRUE; */
 big_buffer = store_malloc(big_buffer_size);
 
 /* strip_trailing_dot = TRUE; */


@@ -2114,8 +2108,7 @@ while (Ufgets(buffer, sizeof(buffer), stdin) != NULL)
   {
   buffer[Ustrlen(buffer)-1] = 0;
   if (buffer[0] == 0) break;
   {
   buffer[Ustrlen(buffer)-1] = 0;
   if (buffer[0] == 0) break;

-  printf("%s\n", CS parse_fix_phrase(buffer, Ustrlen(buffer), outbuff,
-    sizeof(outbuff)));
+  printf("%s\n", CS parse_fix_phrase(buffer, Ustrlen(buffer)));

   }
 
 printf("Testing parse_extract_address without group syntax and without UTF-8\n");
   }
 
 printf("Testing parse_extract_address without group syntax and without UTF-8\n");


@@ -2190,7 +2183,7 @@ while (Ufgets(buffer, sizeof(buffer), stdin) != NULL)
       }
 
     s = ss + (terminator? 1:0);
       }
 
     s = ss + (terminator? 1:0);

-    while (isspace(*s)) s++;
+    Uskip_whitespace(&s);

     }
   }
 
     }
   }