[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-hotspot-gc-dev
Subject:    Re: RFR: 8068971: A heap region being cleared should not belong to the cset
From:       Erik Helin <erik.helin () oracle ! com>
Date:       2015-01-16 10:53:21
Message-ID: 20150116105321.GB6476 () ehelin ! jrpg ! bea ! com
[Download RAW message or body]

On 2015-01-16, Bengt Rutisson wrote:
> 
> Hi Erik,
> 
> On 2015-01-16 08:51, Thomas Schatzl wrote:
> >Hi Erik,
> >
> >On Fri, 2015-01-16 at 08:47 +0100, Erik Helin wrote:
> >>Hi all,
> >>
> >>this tiny patch ensures that a heap region being cleared via
> >>HeapRegion::hr_clear does not belong to the collection set. This has
> >>never been the case, because we are only clearing heap regions upon
> >>initialization (and then they don't belong to any cset) or from the
> >>cleanup phase of the concurrent marking. The cleanup phase will only
> >>call hr_clear on an old heap region, and an old heap region is only part
> >>of the cset during a mixed GC (which can't happen concurrently with the
> >>cleanup phase of concurrent marking).
> >>
> >>Webrev:
> >>http://cr.openjdk.java.net/~ehelin/8068971/webrev.00/
> >   could you change the assert to something like:
> >
> >+  assert(!_in_collection_set,
> >+         err_msg("Should not clear heap region %u in the collection set", hrm_index()));
> >
> >to improve the error message?
> 
> Looks good to me too modulo Thomas' error message comment. No need for a new
> webrev from my side.

Thanks!
Erik

> Thanks,
> Bengt
> 
> >
> >Thanks,
> >   Thomas
> >
> >
> 
[prev in list] [next in list] [prev in thread] [next in thread]