[
https://issues.apache.org/jira/browse/HDFS-6451?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14010503#comment-14010503
]
Brandon Li commented on HDFS-6451:
----------------------------------
>From [~zhongyi-altiscale]:
{quote}Hi Jing Zhao, it's definitely good to have a single exception handler
instead of replicating the same code everywhere, but since each server
procedure (ACCESS, GETATTR, FSSTAT, etc) might have their private data that
needs to be written out, the child NFS3Response class still need to overload
the writeHeaderAndResponse anyways
for AccessControlException, do you mean we need to catch it together with
AuthorizationException in RpcProgramNfs3.java?
or do you mean we need to examine the whole codebase looking for every function
that could potentially throw AccessControlException,
and make sure the error code is set correctly in the catch clause?{quote}
> NFS should not return NFS3ERR_IO for AccessControlException
> ------------------------------------------------------------
>
> Key: HDFS-6451
> URL: https://issues.apache.org/jira/browse/HDFS-6451
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: nfs
> Reporter: Brandon Li
>
> As [~jingzhao] pointed out in HDFS-6411, we need to catch the
> AccessControlException from the HDFS calls, and return NFS3ERR_PERM instead
> of NFS3ERR_IO for it.
> Another possible improvement is to have a single class/method for the common
> exception handling process, instead of repeating the same exception handling
> process in different NFS methods.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
