[
https://issues.apache.org/jira/browse/HDFS-6136?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Liang Xie updated HDFS-6136:
----------------------------
Attachment: HDFS-6136.txt
I am looking the related code these days, i agree with Nick's finding.
Seems Nick didn't make the patch under root dir and could not let QA robot run
successful.
I just made HDFS-6136.txt per original patch and added a test verifier to
expose this issue.
We works fine before, per my understanding, since when we generate a new
"curHeader", we use the real interesting stuff only in later code flow, like:
offset into block,pipeline seq no,data length...
I think it's still a potential bomb here, we do need to fix it.
> PacketReceiver#doRead calls setFieldsFromData with wrong argument
> -----------------------------------------------------------------
>
> Key: HDFS-6136
> URL: https://issues.apache.org/jira/browse/HDFS-6136
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: datanode
> Affects Versions: 2.3.0
> Reporter: Nick White
> Labels: patch
> Attachments: HDFS-6136.patch, HDFS-6136.txt
>
>
> PacketHeader#setFieldsFromData takes the packet length as the first argument,
> but PacketReceiver#doRead passes the dataPlusChecksumLen (which is 4 less).
--
This message was sent by Atlassian JIRA
(v6.2#6252)
