Move reverse DNS core management into iwla.py + Add robot_domains configuration

iwla.py

@@ -32,6 +32,7 @@ import logging
 import gettext
 from calendar import monthrange
 from datetime import date, datetime
+import socket
 
 import default_conf as conf
 
@@ -53,6 +54,7 @@ Conf values needed :
     compress_output_files
     excluded_ip
     excluded_domain_name
+    reverse_dns_timeout*
 
 Output files :
     DB_ROOT/meta.db
@@ -133,7 +135,8 @@ class IWLA(object):
 
     ANALYSIS_CLASS = 'HTTP'
     API_VERSION = 1
-    IWLA_VERSION = '0.7'
+    IWLA_VERSION = '0.8'
+    DEFAULT_DNS_TIMEOUT = 0.5
 
     def __init__(self, logLevel, args):
         self.meta_infos = {}
@@ -145,6 +148,9 @@ class IWLA(object):
         self.valid_visitors = None
         self.args = args
 
+        self.reverse_dns_timeout = self.getConfValue('reverse_dns_timeout',
+                                                     IWLA.DEFAULT_DNS_TIMEOUT)
+       
         self.log_format_extracted = re.sub(r'([^\$?\w])', r'\\\g<1>', conf.log_format)
         self.log_format_extracted = re.sub(r'\$(\w+)', '(?P<\g<1>>.+)', self.log_format_extracted)
         self.http_request_extracted = re.compile(r'(?P<http_method>\S+) (?P<http_uri>\S+) (?P<http_version>\S+)')
@@ -242,6 +248,26 @@ class IWLA(object):
     def getCSSPath(self):
         return conf.css_path
 
+    def reverseDNS(self, hit):
+        if hit.get('dns_name_replaced', False):
+            return hit['remote_addr']
+        
+        try:
+            timeout = socket.getdefaulttimeout()
+            if timeout != self.reverse_dns_timeout:
+                socket.setdefaulttimeout(self.reverse_dns_timeout)
+            name, _, _ = socket.gethostbyaddr(hit['remote_ip'])
+            if timeout != self.reverse_dns_timeout:
+                socket.setdefaulttimeout(timeout)
+            hit['remote_addr'] = name.lower()
+            hit['dns_name_replaced'] = True
+        except socket.herror:
+            pass
+        finally:
+            hit['dns_analysed'] = True
+
+        return hit['remote_addr']
+    
     def _clearMeta(self):
         self.meta_infos = {
             'last_time' : None,

plugins/display/feeds.py

@@ -59,7 +59,7 @@ class IWLADisplayFeeds(IPlugin):
         return True
 
     def hook(self):
-        from plugins.post_analysis.feeds import IWLAPostAnalysisFeeds
+        from plugins.pre_analysis.feeds import IWLAPostAnalysisFeeds
 
         display = self.iwla.getDisplay()
         hits = self.iwla.getCurrentVisits()

plugins/pre_analysis/feeds.py

@@ -25,7 +25,7 @@ from iwla import IWLA
 from iplugin import IPlugin
 
 """
-Post analysis hook
+Pre analysis hook
 
 Find feeds parsers (first hit in feeds conf value and no viewed pages if it's a robot)
 If merge_feeds_parsers is set to True, merge feeds parsers with the same user agent
@@ -77,6 +77,7 @@ class IWLAPostAnalysisFeeds(IPlugin):
         feeds_agents = self.iwla.getConfValue('feeds_agents', [])
         self.merge_feeds_parsers = self.iwla.getConfValue('merge_feeds_parsers', False)
         _merge_feeds_parsers_list = self.iwla.getConfValue('merge_feeds_parsers_list', [])
+        _no_merge_feeds_parsers_list = self.iwla.getConfValue('no_merge_feeds_parsers_list', [])
 
         if feeds is None: return False
 
@@ -104,6 +105,10 @@ class IWLAPostAnalysisFeeds(IPlugin):
         for f in _merge_feeds_parsers_list:
             self.merge_feeds_parsers_list.append(re.compile(f))
 
+        self.no_merge_feeds_parsers_list = []
+        for f in _no_merge_feeds_parsers_list:
+            self.no_merge_feeds_parsers_list.append(re.compile(f))
+
         self.merged_feeds = {}
 
         return True
@@ -134,8 +139,11 @@ class IWLAPostAnalysisFeeds(IPlugin):
     
     def mergeFeedsParsers(self, isFeedParser, hit):
         if isFeedParser in (self.FEED_PARSER, self.MERGED_FEED_PARSER):
+            for r in self.no_merge_feeds_parsers_list:
+                if r.match(hit['remote_addr']) or r.match(hit['remote_ip']) or r.match(hit['requests'][0]['http_user_agent']):
+                    return
             for r in self.merge_feeds_parsers_list:
-                if r.match(hit['remote_addr']) or r.match(hit['remote_ip']):
+                if r.match(hit['remote_addr']) or r.match(hit['remote_ip']) or r.match(hit['requests'][0]['http_user_agent']):
                     # One group can view multiple different feeds
                     key = r.pattern + hit.get('feed_domain', '') + hit.get('feed_uri', '')
                     self._appendToMergeCache(isFeedParser, key, hit)
@@ -159,15 +167,6 @@ class IWLAPostAnalysisFeeds(IPlugin):
                 if hit['last_access'] > hit.get('feed_parser_last_access', time.gmtime(0)):
                     hit['feed_parser_last_access'] = hit['last_access']
                                             
-                if not hit.get('feed_name_analyzed', False) and\
-                   hit.get('dns_name_replaced', False):
-                    hit['feed_name_analyzed'] = True
-                    addr = hit.get('remote_addr', None)
-                    for r in self.bad_feeds_re:
-                        if r.match(addr):
-                            hit['feed_parser'] = self.NOT_A_FEED_PARSER
-                            break
-                        
                 # Register already tagged feed parser in merged_feeds
                 if self.merge_feeds_parsers:
                     self.mergeFeedsParsers(isFeedParser, hit)
@@ -177,12 +176,12 @@ class IWLAPostAnalysisFeeds(IPlugin):
             isFeedParser = self.NOT_A_FEED_PARSER
             uri = request['extract_request']['extract_uri'].lower()
             for regexp in self.feeds_re:
-                if regexp.match(uri):
+                if regexp.match(uri) and self.iwla.hasBeenViewed(request):
                     isFeedParser = self.FEED_PARSER
-                    # Robot that views pages -> bot
-                    if hit['robot']:
-                        if hit['not_viewed_pages'][0]:
-                            isFeedParser = self.NOT_A_FEED_PARSER
+                    # # Robot that views pages -> bot
+                    # if hit['robot']:
+                    #     if hit['not_viewed_pages'][0]:
+                    #         isFeedParser = self.NOT_A_FEED_PARSER
                     break
 
             user_agent = request['http_user_agent'].lower()
@@ -199,14 +198,17 @@ class IWLAPostAnalysisFeeds(IPlugin):
                         isFeedParser = self.NOT_A_FEED_PARSER
                         break
 
-            if not hit.get('feed_name_analyzed', False) and\
-               hit.get('dns_name_replaced', False):
-                hit['feed_name_analyzed'] = True
-                addr = hit.get('remote_addr', None)
-                for r in self.bad_feeds_re:
-                    if r.match(addr):
-                        isFeedParser = hit['feed_parser'] = self.NOT_A_FEED_PARSER
-                        break
+            if isFeedParser == self.FEED_PARSER:
+                if not hit.get('dns_name_replaced', False):
+                    self.iwla.reverseDNS(hit)
+                    
+                if not hit.get('feed_name_analyzed', False):
+                    hit['feed_name_analyzed'] = True
+                    addr = hit.get('remote_addr', None)
+                    for r in self.bad_feeds_re:
+                        if r.match(addr):
+                            isFeedParser = self.NOT_A_FEED_PARSER
+                            break
 
             if isFeedParser == self.FEED_PARSER:
                 hit['feed_domain'] = request['server_name']
@@ -217,6 +219,7 @@ class IWLAPostAnalysisFeeds(IPlugin):
                 if subscribers:
                     hit['feed_subscribers'] = int(subscribers.groups()[0])
             
+                hit['robot'] = True
             hit['feed_parser'] = isFeedParser
             if self.merge_feeds_parsers:
                 self.mergeFeedsParsers(isFeedParser, hit)

plugins/pre_analysis/reverse_dns.py

@@ -19,12 +19,13 @@
 #
 
 import socket
+import re
 
 from iwla import IWLA
 from iplugin import IPlugin
 
 """
-Post analysis hook
+Pre analysis hook
 
 Replace IP by reverse DNS names
 
@@ -32,7 +33,7 @@ Plugin requirements :
     None
 
 Conf values needed :
-    reverse_dns_timeout*
+    robot_domains*
 
 Output files :
     None
@@ -51,12 +52,13 @@ Statistics deletion :
 """
 
 class IWLAPostAnalysisReverseDNS(IPlugin):
-    DEFAULT_DNS_TIMEOUT = 0.5
 
     def load(self):
-        timeout = self.iwla.getConfValue('reverse_dns_timeout',
-                                         IWLAPostAnalysisReverseDNS.DEFAULT_DNS_TIMEOUT)
-        socket.setdefaulttimeout(timeout)
+        self.robot_domains_re = []
+        robot_domains = self.iwla.getConfValue('robot_domains', [])
+        for domain in robot_domains:
+            self.robot_domains_re.append(re.compile(domain))
+
         return True
         
     def hook(self):
@@ -65,15 +67,13 @@ class IWLAPostAnalysisReverseDNS(IPlugin):
             if hit.get('dns_analysed', False): continue
             # Do reverse for feed parser even if they're not
             # valid visitors
-            if not hit.get('feed_parser', False) and\
-               not self.iwla.isValidVisitor(hit):
+            if hit.get('robot', False) and not hit.get('feed_parser', False):
                 continue
-            try:
-                name, _, _ = socket.gethostbyaddr(k)
-                hit['remote_addr'] = name.lower()
-                hit['dns_name_replaced'] = True
-            except:
-                pass
-            finally:
-                hit['dns_analysed'] = True
+
+            res = self.iwla.reverseDNS(hit)
+            
+            for r in self.robot_domains_re:
+                if r.match(hit['remote_addr']):
+                    hit['robot'] = True
+                    break