[
https://issues.apache.org/jira/browse/SSHD-911?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16829954#comment-16829954
]
Goldstein Lyor commented on SSHD-911:
-------------------------------------
I am a bit confused - in the new deadlock you report it seems to revolve only
around the {{encodeLock}} so I don't see how a deadlock can occur. On the other
hand, you mention {{AbstractCloseable#lock}} as the participant in the
deadlock. I am therefore not sure who is locking what and why is it causing a
deadlock.
Nevertheless, I took a closer look at the {{AbstractCloseable#lock}} and it
seems that code rather "abuses" it. However, figuring out where is it really
needed and where it isn't is quite an upheaval, so I have opened a *separate*
issue for it (SSHD-912).
That being said, the only way I can figure out a deadlock between the
{{encodeLock}} and the {{AbstractCloseable#lock}} is if thread A somehow lock
{{AbstractCloseable#lock}} and tries to write a packet (which seems a possible
scenario) while thread B tries to lock {{AbstractCloseable#lock}} while already
holding {{encodeLock}} - which my current code analysis does not seem to
indicate is possible (but I'll keep looking).
> Deadlock in Apache SshClient used for local port forwarding
> -----------------------------------------------------------
>
> Key: SSHD-911
> URL: https://issues.apache.org/jira/browse/SSHD-911
> Project: MINA SSHD
> Issue Type: Bug
> Affects Versions: 2.2.0
> Reporter: Roberto Deandrea
> Assignee: Goldstein Lyor
> Priority: Major
> Attachments: spxp.usrdump.20382...@aix72.dmz.20190417205240026.log,
> spxp.usrdump.33095...@aix72.dmz.20190423123751496.log
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Hi Lyor,
> We have an Apache SSHD 2.2.0 client and server SSH pair used for local port
> forwarding of different protocols.
> In our lab environment we found that sometimes when the SSH client for any
> reason is disconnecting from the SSH server, a DEADLOCK happens on the
> sshClient regarding the threads in the thread pool.
> We are using builtin nio2 implementation for our SshClient.
> The deadlock is regarding the ClientChannelPendingMessagesQueue class added
> to fix routing problems for local port forwarding in the past.
> Attached you can find the thread dump taken after the deadlock happend.
>
> Can you please troubleshoot this problem ?
> [^spxp.usrdump.20382...@aix72.dmz.20190417205240026.log]
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
