[
https://issues.apache.org/jira/browse/HADOOP-10758?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14118382#comment-14118382
]
Alejandro Abdelnur commented on HADOOP-10758:
---------------------------------------------
Looks good, just a few minor things:
*KeyAuthorizationKeyProvider.java*:
* shouldn’t {{getExtension()}} and {{getKeyProvider()}} return {{this}}? or is
the intention to return the unguarded entity? if the later, we should log a
warning on the GET call.
* {{doAccessCheck()}}, if the {{KEY_ACL_NAME}} attribute is NULL, shouldn’t we
pass the name of the key? by doing this you can key-acl existing keys via its
name (in the case you enable key-acl after the keys were created).
* {{authorizeCkreateKey()}}, the {{success =...}} predicate assignment could be
done once by doing a refactoring on how the name/attribute is assigned.
*KMSACLs.java*:
* {{setKeyACLs()}}, if name of the key has dots (can it?) then the logic here
will fail as you are expecting 4 elements after split. I think you should look
for postfix without assuming dots, you already filtered the prefix.
* it is not clear to me what is the behavior if no default ACLs are set. are we
assuming '*' or we are requiring explicit ACLs for every key? it seems the
later makes more sense, no? we should log a warning and put that in the docs.
*KMSConstants.java*:
* {{KEY_ACL_PREFIX}} does not seem used.
> KMS: add ACLs on per key basis.
> -------------------------------
>
> Key: HADOOP-10758
> URL: https://issues.apache.org/jira/browse/HADOOP-10758
> Project: Hadoop Common
> Issue Type: Improvement
> Components: security
> Affects Versions: 3.0.0
> Reporter: Alejandro Abdelnur
> Assignee: Arun Suresh
> Attachments: HADOOP-10758.1.patch, HADOOP-10758.2.patch,
> HADOOP-10758.3.patch, HADOOP-10758.4.patch, HADOOP-10758.5.patch,
> HADOOP-10758.6.patch
>
>
> The KMS server should enforce ACLs on per key basis.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
