JVM hangs beyond recovery

David Holmes David.Holmes at oracle.com
Wed Jul 7 21:03:30 PDT 2010


Stas Oskin said the following on 06/30/10 00:58:
> 
>     The main issue is that the stack trace always contains "??", which
>     make it impossible to find out what caused the deadlock.
> 
> Just wanted to update that I tried kill -QUIT (or kill -3) and there 
> wasn't any thread dump as a result.
> It seems the deadlock appears so deeply, that it basically hangs even 
> the basic routines.

When the VM hangs during a safepoint there is no functionality available 
to you. The signal can't trigger the start of the thread to perform the 
dump because that thread must wait for the safepoint to end.

> Again, if there any way to convert these ?? symbols in GDB dump to 
> actual function names, it would greatly help discovering the culprit.

Some of the ??? may be coming from JIT compiled frames - these have no 
name but aren't responsible for the problem anyway. Other ??? may be 
coming from libraries with no symbol information and I'm afraid that 
trying to correlate addresses with names is an OS/loader issue not a 
Java/JVM one.

> Will appreciate any idea on subject.

Is there any LD debugging/tracing that might help?

David Holmes

> Thanks.


More information about the hotspot-runtime-dev mailing list