From: Abhishek Lekshmanan <abhishek@suse.com>
Date: Mon, 20 Aug 2018 15:45:04 +0000 (+0200)
Subject: rgw: s3: awsv4 drop special handling for x-amz-credential
X-Git-Tag: v14.1.0~894^2
X-Git-Url: http://git-server-git.apps.pok.os.sepia.ceph.com/?a=commitdiff_plain;h=25e1d2a32074254e9ebd780f3a41b2ff6469370f;p=ceph.git

rgw: s3: awsv4 drop special handling for x-amz-credential

While s3 docs mention that every byte must be urlencoded, they are relaxed in
its implementation, when testing this behaviour on aws s3 itself, they seem to
be relaxed in handling aws credentials of the form

X-Amz-Credential=access1/20180817T123456Z/us-east-1...

wherein

X-Amz-Credential=access1%2F2018...

is expected. This allows for clients requesting the urls in both forms to
succeed as aws s3 does at the moment

Fixes: http://tracker.ceph.com/issues/26965
Signed-off-by: Abhishek Lekshmanan <abhishek@suse.com>
---

diff --git a/src/rgw/rgw_auth_s3.cc b/src/rgw/rgw_auth_s3.cc
index 55893b6fe38e..81416849d7af 100644
--- a/src/rgw/rgw_auth_s3.cc
+++ b/src/rgw/rgw_auth_s3.cc
@@ -489,14 +489,10 @@ std::string get_v4_canonical_qs(const req_info& info, const bool using_qs)
       continue;
     }
 
-    if (key == "X-Amz-Credential") {
-      /* FIXME(rzarzynski): I can't find any comment in the previously linked
-       * Amazon's docs saying that X-Amz-Credential should be handled in this
-       * way. */
-      canonical_qs_map[key.to_string()] = val.to_string();
-    } else {
-      canonical_qs_map[aws4_uri_recode(key, true)] = aws4_uri_recode(val, true);
-    }
+    // while awsv4 specs ask for all slashes to be encoded, s3 itself is relaxed
+    // in its implementation allowing non-url-encoded slashes to be present in
+    // presigned urls for instance
+    canonical_qs_map[aws4_uri_recode(key, true)] = aws4_uri_recode(val, true);
   }
 
   /* Thanks to the early exist we have the guarantee that canonical_qs_map has