RFR: 8254695: G1: Next mark bitmap clear not cancelled after marking abort
Stefan Johansson
sjohanss at openjdk.java.net
Thu Oct 15 08:21:09 UTC 2020
On Wed, 14 Oct 2020 12:23:23 GMT, Thomas Schatzl <tschatzl at openjdk.org> wrote:
> Hi all,
>
> can I have reviews for this change that fixes aborting the concurrent next bitmap clear if aborted?
>
> The problems are that we can enter the clear loop with the marking already aborted, so
> - yielding does not return true any more (marking aborted while joining the STS) and we never abort in the loop if we
> join all conditions together
> - and G1 at least clears the first part(s) of the bitmap as the check is done afterwards
>
> Testing: tier1-5; no more reproducable crashes with other changes (JDK-8253600)
>
> Thanks,
> Thomas
Looks good.
-------------
Marked as reviewed by sjohanss (Reviewer).
PR: https://git.openjdk.java.net/jdk/pull/653
More information about the hotspot-gc-dev
mailing list