[ https://issues.apache.org/jira/browse/SSHD-634?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15130098#comment-15130098 ]
Eugene Petrenko commented on SSHD-634: -------------------------------------- Thanks! Watched/Voted for related issue. My idea was to use 2 pre-allocated buffers per session. Those pre-allocated buffers may grow in size once necessary. So there is no need to have more than one extra buffer for decoding, once decoding can only run on one thread. Eventually, this means zero memory traffic at all, no matter how many classes/interfaces we use to represent a buffer > Suspicious read buffer re-use in write command > ---------------------------------------------- > > Key: SSHD-634 > URL: https://issues.apache.org/jira/browse/SSHD-634 > Project: MINA SSHD > Issue Type: Bug > Affects Versions: 1.0.0 > Reporter: Eugene Petrenko > Assignee: Goldstein Lyor > Fix For: 1.2.0 > > > From time to time I saw a {{Bad packet length ###}} errors from SSH clients > that calls commands to the SSHD server running with Apache SSHD. > It may turn out there are some issues in the implementation. > I found one suspicious usage of buffers around > {{org.apache.sshd.common.channel.AbstractChannel#sendResponse}} method. As > far as I see from the code it uses the read buffer as a place to generate a > write packet command. Maybe it's worth to create a fresh buffer here to avoid > it be updated from the other thread while it is put to a write queue -- This message was sent by Atlassian JIRA (v6.3.4#6332)