From: Simon Arlott <bugzilla.exim.simon@arlott.org>
Date: Sun, 22 Sep 2019 17:45:18 +0000 (+0100)
Subject: DNS: do not skip initial two components of SRV & TLSA lookups before checking name... 
X-Git-Url: https://git.exim.org/users/jgh/exim.git/commitdiff_plain/6219e0ec4a59a06b84eaabb6b3ae5d9e8f166672

DNS: do not skip initial two components of SRV & TLSA lookups before checking name syntax.
The introduction of DKIM added _ to the permitted chars, so those components will pass.
---

diff --git a/doc/doc-docbook/spec.xfpt b/doc/doc-docbook/spec.xfpt
index c2adc9ea6..5acdce0a6 100644
--- a/doc/doc-docbook/spec.xfpt
+++ b/doc/doc-docbook/spec.xfpt
@@ -14691,13 +14691,20 @@ recommended, except when you have no other choice.
 .cindex "UTF-8" "in domain name"
 Lots of discussion is going on about internationalized domain names. One
 camp is strongly in favour of just using UTF-8 characters, and it seems
-that at least two other MTAs permit this. This option allows Exim users to
-experiment if they wish.
+that at least two other MTAs permit this.
+This option allows Exim users to experiment if they wish.
 
 If it is set true, Exim's domain parsing function allows valid
 UTF-8 multicharacters to appear in domain name components, in addition to
-letters, digits, and hyphens. However, just setting this option is not
-enough; if you want to look up these domain names in the DNS, you must also
+letters, digits, and hyphens.
+
+.new
+If Exim is built with internationalization support
+and the SMTPUTF8 ESMTP option is in use (see chapter &<<CHAPi18n>>&)
+this option can be left as default.
+.wen
+Without that,
+if you want to look up such domain names in the DNS, you must also
 adjust the value of &%dns_check_names_pattern%& to match the extended form. A
 suitable setting is:
 .code
diff --git a/src/src/dns.c b/src/src/dns.c
index 44654353c..4750f1b52 100644
--- a/src/src/dns.c
+++ b/src/src/dns.c
@@ -818,34 +818,17 @@ regex has substrings that are used - the default uses a conditional.
 
 This test is omitted for PTR records. These occur only in calls from the dnsdb
 lookup, which constructs the names itself, so they should be OK. Besides,
-bitstring labels don't conform to normal name syntax. (But the aren't used any
-more.)
-
-For SRV records, we omit the initial _smtp._tcp. components at the start.
-The check has been seen to bite on the destination of a SRV lookup that
-initiall hit a CNAME, for which the next name had only two components.
-RFC2782 makes no mention of the possibiility of CNAMES, but the Wikipedia
-article on SRV says they are not a valid configuration. */
+bitstring labels don't conform to normal name syntax. (But they aren't used any
+more.) */
 
 #ifndef STAND_ALONE   /* Omit this for stand-alone tests */
 
 if (check_dns_names_pattern[0] != 0 && type != T_PTR && type != T_TXT)
   {
-  const uschar *checkname = name;
   int ovector[3*(EXPAND_MAXN+1)];
 
   dns_pattern_init();
-
-  /* For an SRV lookup, skip over the first two components (the service and
-  protocol names, which both start with an underscore). */
-
-  if (type == T_SRV || type == T_TLSA)
-    {
-    while (*checkname && *checkname++ != '.') ;
-    while (*checkname && *checkname++ != '.') ;
-    }
-
-  if (pcre_exec(regex_check_dns_names, NULL, CCS checkname, Ustrlen(checkname),
+  if (pcre_exec(regex_check_dns_names, NULL, CCS name, Ustrlen(name),
       0, PCRE_EOPT, ovector, nelem(ovector)) < 0)
     {
     DEBUG(D_dns)