"Broken pipe" error when using VirtualMachine API emulation

We provide a utility for remote attachment of our agent that relies on ByteBuddy’s VirtualMachine emulation feature. The current version we use is greater than 1.10.1, where there is a mention of “broken pipe” in the release notes.

We have a user reporting getting the broken pipe error using the following environment:

• Oracle JVM 1.8.0_74 (using the same JRE to run the app and the attachment utility)
• Centos Linux release 7.4
• SELinux is disabled
• Running both the app and the attachment utility as root

Here is the relevant part of the stack trace:

Caused by: com.sun.jna.LastErrorException: [32] Broken Pipe
at com.sun.jna.Native.invokeInt(Native Method)
at com.sun.jna.Function.invoke(Function.java:426)
at com.sun.jna.Function.invoke(Function.java:361)
at com.sun.jna.Library$Handler.invoke(Library.java:265)
at net.bytebuddy.agent.$Proxy0.write(Unknown Source)
at net.bytebuddy.agent.VirtualMachine$ForHotSpot$Connection$ForJnaPosixSocket.write(VirtualMachine.java:735)
at net.bytebuddy.agent.VirtualMachine$ForHotSpot$Connection$ForJnaPosixSocket.write(VirtualMachine.java:686)
at net.bytebuddy.agent.VirtualMachine$ForHotSpot$Connection$OnPersistentByteChannel.execute(VirtualMachine.java:592)
at net.bytebuddy.agent.VirtualMachine$ForHotSpot.load(VirtualMachine.java:361)
at net.bytebuddy.agent.VirtualMachine$ForHotSpot.loadAgent(VirtualMachine.java:335)

Do you know what could cause this kind of errors ?

• from the stack trace, creating the temporary file in working directory and opening the socket is not the issue
• issue is triggered when trying to write to the socket using JNA, is broken pipe a generic issue for JNA, or does it provide any meaningful indication about the root cause ?

[raphw]

Yes, would be interesting to check its chmod for being 600.

[eyalkoren]

They could also try to set the permissions of the socket file manually.

Is that the .attach_pid<pid> file? They looked at the .java_pid<pid> permissions and they seemed correct.