Changeset 315708 in indico

Timestamp:

08/23/11 15:38:51 (21 months ago)

Author:

Jose Benito <jose.benito.gonzalez@…>

Branches:

master, hello-world-walkthrough, ipv6, v0.98-series, v0.98.2, v0.98.3, v0.98b2, v0.99, 051b2622c51afb171a1dedb46a0df4fbb0cbd02e, 0da0c1403bae8e51d8229f460181c71b9e6dda72

Children:

fcdc60

Parents:

1ea4c5

git-author:

Adrian Moennich <jerome.ernst.monnich@…> (05/12/11 14:06:08)

git-committer:

Jose Benito <jose.benito.gonzalez@…> (08/23/11 15:38:51)

Message:

[IMP] Improve URL format for caching/signing

• in both cases the query strign is sorted
• the signature is now passed via &signature=...
• the timestamp is included as &timestamp=...
• the cache key now uses SHA-256 instead of hash()

Location:

indico/web/http_api

Files:

• cache.py (modified) (2 diffs)
• handlers.py (modified) (6 diffs)

indico/web/http_api/cache.py

@@ -20 +20 @@
 from MaKaC.common.cache import MultiLevelCacheEntry, MultiLevelCache
 import datetime
+import hashlib
 import time
 import os
@@ -51 +52 @@
         return ['requests']
 
-    def _generateKey(self, path, qdata):
-        queryHash = hash(frozenset((key, frozenset(values)) for key, values in qdata.iteritems()))
-        return '.'.join(map(str, (hash(path), queryHash)))
+    def cacheObject(self, key, obj):
+        key = hashlib.sha256(key).hexdigest()
+        return super(RequestCache, self).cacheObject(key, obj)
 
-    def cacheObject(self, path, qdata, obj):
-        return super(RequestCache, self).cacheObject(self._generateKey(path, qdata), obj)
-
-    def loadObject(self, path, qdata):
-        return super(RequestCache, self).loadObject(self._generateKey(path, qdata))
+    def loadObject(self, key):
+        key = hashlib.sha256(key).hexdigest()
+        return super(RequestCache, self).loadObject(key)
 
     def isDirty(self, path, object):

indico/web/http_api/handlers.py

@@ -28 +28 @@
 import re
 import time
+import urllib
 from urlparse import parse_qs
 from ZODB.POSException import ConflictError
@@ -61 +62 @@
 }
 
-# Compile regexps
+# Compile url regexps
 EXPORT_URL_MAP = dict((re.compile(pathRe), handlerFunc) for pathRe, handlerFunc in EXPORT_URL_MAP.iteritems())
-RE_REMOVE_EXTENSION = re.compile(r'\.(\w+)$')
+# Remove the extension at the end or before the querystring
+RE_REMOVE_EXTENSION = re.compile(r'\.(\w+)(?:$|(?=\?))')
+
+
+def normalizeQuery(path, query, ts=None, remove=('timestamp', 'signature')):
+    """Normalize request path and query so it can be used for caching and signing
+
+    Returns a string consisting of path and sorted query string.
+    Dynamic arguments like signature and timestamp are removed from the query string.
+    """
+    qdata = remove_lists(parse_qs(query))
+    if remove:
+        for key in remove:
+            qdata.pop(key, None)
+    if ts is not None:
+        qdata['timestamp'] = ts
+    sortedQuery = sorted(qdata.items(), key=lambda x: x[0].lower())
+    if sortedQuery:
+        return '%s?%s' % (path, urllib.urlencode(sortedQuery))
+    else:
+        return path
 
 
@@ -72 +93 @@
     candidates = []
     for i in xrange(-1, 2):
-        h = hmac.new(key, '%s?%s&%d' % (path, query, ts + i), hashlib.sha1)
+        h = hmac.new(key, normalizeQuery(path, query, ts + i), hashlib.sha1)
         candidates.append(h.hexdigest())
     if signature not in candidates:
@@ -172 +193 @@
 def handler(req, **params):
     path, query = req.URLFields['PATH_INFO'], req.URLFields['QUERY_STRING']
-    # Extract HMAC signature
-    signature = None
-    m = re.search(r'&([0-9a-fA-F]{40})$', query)
-    if m:
-        signature = m.group(1).lower()
-        query = query[:-41]
-
     # Parse the actual query string
     qdata = parse_qs(query)
-    no_cache = get_query_parameter(qdata, ['nc', 'nocache'], 'no') == 'yes'
 
     dbi = DBMgr.getInstance()
     dbi.startRequest()
 
-    # Copy qdata for the cache key
-    qdata_copy = dict(qdata)
     cache = RequestCache(HelperMaKaCInfo.getMaKaCInfoInstance().getAPICacheTTL())
 
+    apiKey = get_query_parameter(qdata, ['ak', 'apikey'], None)
+    signature = get_query_parameter(qdata, ['signature'])
+    no_cache = get_query_parameter(qdata, ['nc', 'nocache'], 'no') == 'yes'
     pretty = get_query_parameter(qdata, ['p', 'pretty'], 'no') == 'yes'
-    apiKey = get_query_parameter(qdata, ['ak', 'apikey'], None)
     onlyPublic = get_query_parameter(qdata, ['op', 'onlypublic'], 'no') == 'yes'
 
@@ -210 +223 @@
         # Get rid of API key in cache key if we did not impersonate a user
         if ak and aw.getUser() is None:
-            qdata_copy.pop('ak', None)
-            qdata_copy.pop('apikey', None)
+            cache_key = normalizeQuery(path, query, remove=('ak', 'apiKey', 'signature', 'timestamp'))
+        else:
+            cache_key = normalizeQuery(path, query, remove=('signature', 'timestamp'))
 
         obj = None
         add_to_cache = True
-        cache_path = RE_REMOVE_EXTENSION.sub('', path)
+        cache_key = RE_REMOVE_EXTENSION.sub('', cache_key)
         if not no_cache:
-            obj = cache.loadObject(cache_path, qdata_copy)
+            obj = cache.loadObject(cache_key)
             if obj is not None:
                 result, complete = obj.getContent()
@@ -226 +240 @@
             result, complete = func(dbi, aw, qdata, *args)
         if result is not None and add_to_cache:
-            cache.cacheObject(cache_path, qdata_copy, (result, complete))
+            cache.cacheObject(cache_key, (result, complete))
     except HTTPAPIError, e:
         error = e