Request for review 6583275: Hotspot crash in vm_perform_shutdown_actions due to uninitialized TLS duri,ng out of memory handling (original) (raw)

David Holmes David.Holmes at oracle.com
Wed Jan 5 18:36:46 PST 2011


Looks good to me.

David

Coleen Phillimore said the following on 01/06/11 12:30:

Thanks, John and David for the same comment. I found the threadindex() != 1 check in code above so changed them both to isinitialized() and reversed the ?: see:

http://cr.openjdk.java.net/~coleenp/65832752/ Thanks! Coleen On 1/5/2011 6:26 PM, John Coomes wrote: Coleen Phillimore (coleen.phillimore at oracle.com) wrote:

Summary: Call getthreadslow() in vmperformshutdown actions and add null check.

Tested by forcing vmexitoutofmemory in currentstackregion (if isinitcompleted) to reproduce the bug. open webrev at http://cr.openjdk.java.net/~coleenp/6583275/ bug link at http://bugs.sun.com/viewbug.do?bugid=6583275 You should use ThreadLocalStorage::isinitialized() instead of checking for -1. Other than that, looks good. -John



More information about the hotspot-runtime-dev mailing list